WEBVTT

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/31-0
00:00:04.337 --> 00:00:05.217
[noise]

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/47-0
00:00:05.997 --> 00:00:09.970
Thank you all for attending the
Earthquake Science Center Weekly

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/47-1
00:00:09.970 --> 00:00:10.887
Seminar Series.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/53-0
00:00:10.897 --> 00:00:12.207
If you are new, welcome.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/68-0
00:00:12.277 --> 00:00:14.798
If you'd like to be added to our
email distribution group, please

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/68-1
00:00:14.798 --> 00:00:15.447
send us an email.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/94-0
00:00:16.377 --> 00:00:19.974
Seminars are recorded and mostly
all talks are posted on the USGS

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/94-1
00:00:19.974 --> 00:00:21.997
Earthquake Science Center web
page.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/118-0
00:00:22.077 --> 00:00:25.014
Closed caption can be turned on
by clicking on the CC icon on

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/118-1
00:00:25.014 --> 00:00:26.767
the more tab at the top of the
page.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/167-0
00:00:27.187 --> 00:00:30.027
Attendees, please mute your mics
and turn off your cameras, until

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/167-1
00:00:30.027 --> 00:00:32.644
the Q&A session at the
end of the talk. Submit your

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/167-2
00:00:32.644 --> 00:00:35.306
questions via the chat at any
time, or wait to turn on your

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/167-3
00:00:35.306 --> 00:00:36.637
camera and ask your questions 

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/186-0
00:00:36.647 --> 00:00:42.003
during the Q&A session.
Alright, so Ms. Linda McCrory

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/186-1
00:00:42.003 --> 00:00:45.717
and announcements for today,
October 25th.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/203-0
00:00:45.967 --> 00:00:50.545
Linda McCrory and Patricia
Elliott will be in Menlo Park on

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/203-1
00:00:50.545 --> 00:00:51.537
October 25th.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/280-0
00:00:51.747 --> 00:00:55.588
And they have some open slots if
you'd like to speak with them

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/280-1
00:00:55.588 --> 00:00:59.185
and sign up, please see the
email from Kristina Hearn, the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/280-2
00:00:59.185 --> 00:01:02.599
ESC Administrative Officer and
the USGS Town Hall is on

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/280-3
00:01:02.599 --> 00:01:06.256
Thursday, November 2nd from 2:30
to 4:00 PM Eastern Time on

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/280-4
00:01:06.256 --> 00:01:10.158
Microsoft Teams and that is live
joined USGS senior leaders for

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/280-5
00:01:10.158 --> 00:01:13.267
updates related to budget, human
capital and more.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/293-0
00:01:14.147 --> 00:01:17.937
The USGS and the Center for
Accelerator Mass Spectrometer.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/299-0
00:01:17.987 --> 00:01:19.407
Mass spectrometry.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/308-0
00:01:19.547 --> 00:01:22.047
Ohh alright, let's try one more
time, please.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/331-0
00:01:22.057 --> 00:01:25.407
Spectrometry at Lawrence
Livermore National Lab are

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/331-1
00:01:25.407 --> 00:01:29.272
planning a joint workshop to
discuss scientific areas right

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/331-2
00:01:29.272 --> 00:01:30.947
for future collaborations.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/356-0
00:01:31.147 --> 00:01:34.843
The workshop is tentatively
scheduled for late January, and

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/356-1
00:01:34.843 --> 00:01:38.477
those interested in attending are 
 asked to RSVP by Wednesday,

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/356-2
00:01:38.477 --> 00:01:39.277
November 8th.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/358-0
00:01:39.507 --> 00:01:39.837
See 

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/399-0
00:01:39.847 --> 00:01:43.976
Christine Goulet's email from last
Wednesday for further details

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/399-1
00:01:43.976 --> 00:01:48.240
and finally the not ready for
prime Time Colloquium will occur

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/399-2
00:01:48.240 --> 00:01:52.707
tomorrow and it will be somewhat
of an open format and the timing

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/399-3
00:01:52.707 --> 00:01:54.737
is, tentatively at 3 p.m.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/417-0
00:01:54.747 --> 00:01:56.397
It sounds like so for Moffett Field
folks 

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/421-0
00:01:56.407 --> 00:01:59.177
that is relevant for you.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/424-0
00:01:57.507 --> 00:01:58.187
[noise]

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/438-0
00:01:59.267 --> 00:02:02.121
With that, I'm gonna turn it
over to Kate for an introduction

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/438-1
00:02:02.121 --> 00:02:03.087
of our speaker today.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/441-0
00:02:04.547 --> 00:02:04.997
Hi.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/444-0
00:02:05.807 --> 00:02:07.217
So welcome everybody.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/454-0
00:02:07.227 --> 00:02:09.177
I'm excited to introduce Zhiang 
Chen.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/482-0
00:02:09.667 --> 00:02:14.499
He comes to us with a Bachelor's
degree in Mechatronics, which

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/482-1
00:02:14.499 --> 00:02:18.378
is a word that was new to me. He's 
from Changsha in Hunan

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/482-2
00:02:18.378 --> 00:02:19.037
province.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/561-0
00:02:19.047 --> 00:02:23.027
He has a Master's in robotics
from Case Western in the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/561-1
00:02:23.027 --> 00:02:27.301
midwestern province of the US
and recently got his PhD at

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/561-2
00:02:27.301 --> 00:02:31.797
Arizona State University in
exploration system design and he

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/561-3
00:02:31.797 --> 00:02:36.145
currently is a post doc that's
joint funded by Caltech and

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/561-4
00:02:36.145 --> 00:02:40.788
USGS, and he has started working;
we just came back

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/561-5
00:02:40.788 --> 00:02:45.579
Devin and Zhiang and I just came
back from a couple days of field

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/561-6
00:02:45.579 --> 00:02:45.947
work.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/575-0
00:02:45.957 --> 00:02:48.867
So I can tell you that it's
really exciting.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/599-0
00:02:48.877 --> 00:02:51.586
The ideas he has in the ways
he's going to push us in new

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/599-1
00:02:51.586 --> 00:02:54.107
directions, to think about
fragile geologic features.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/614-0
00:02:54.117 --> 00:02:57.501
So he'll be talking about his
past work and some future

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/614-1
00:02:57.501 --> 00:02:58.407
concepts today.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/616-0
00:02:58.757 --> 00:02:59.207
Thanks.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/618-0
00:02:59.657 --> 00:03:00.247
Yeah.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/623-0
00:03:00.317 --> 00:03:00.757
Thank you 

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/637-0
00:03:00.967 --> 00:03:03.821
Kate for the introduction and
thank you everyone for having

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/637-1
00:03:03.821 --> 00:03:04.677
this presentation.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/669-0
00:03:04.947 --> 00:03:08.724
So I'm gonna talk about my
research of using robotics and

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/669-1
00:03:08.724 --> 00:03:12.697
machine learning for geoscience,
and I'll focus on fault zone 

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/669-2
00:03:12.697 --> 00:03:15.757
mapping and fragile geological
feature analysis.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/699-0
00:03:16.067 --> 00:03:20.668
So those are the two research
directions for my research and

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/699-1
00:03:20.668 --> 00:03:25.117
most of the presentation is from
my previous work at ASU 

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/709-0
00:03:25.207 --> 00:03:28.427
and also I'll talk about my
research plan here at Caltech.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/727-0
00:03:30.087 --> 00:03:34.657
So automated geoscience is very
important conceptive format 

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/776-0
00:03:34.667 --> 00:03:38.150
in my research. Automated 
geoscience is a practice that

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/776-1
00:03:38.150 --> 00:03:41.820
leverages robotics to automated
data collection and machine

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/776-2
00:03:41.820 --> 00:03:45.490
learning to automated data
processing such that researchers

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/776-3
00:03:45.490 --> 00:03:48.973
can redirect their focus towards
high level calculative

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/776-4
00:03:48.973 --> 00:03:49.657
activities.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/806-0
00:03:49.927 --> 00:03:54.674
If you can see these scientific
method of geoscience research,

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/806-1
00:03:54.674 --> 00:03:59.346
usually we have dual 
geoscientific observations and

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/806-2
00:03:59.346 --> 00:04:00.777
we raise questions and

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/878-0
00:04:00.787 --> 00:04:05.723
we form [indiscernible] that we
have experimented design data

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/878-1
00:04:05.723 --> 00:04:10.740
collection and data processing
and if we're lucky and we will

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/878-2
00:04:10.740 --> 00:04:15.433
form or discover general
theories and we can apply the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/878-3
00:04:15.433 --> 00:04:20.126
theory to other senses
and where our use the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/878-4
00:04:20.126 --> 00:04:25.143
theory to predict the future. 
The time consuming parts are

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/878-5
00:04:25.143 --> 00:04:28.137
data collection and data
processing.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/904-0
00:04:28.247 --> 00:04:33.033
If we can use robots and machine
learning to automate these two

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/904-1
00:04:33.033 --> 00:04:37.669
time consuming parts and 
geoscientists can focus on other

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/904-2
00:04:37.669 --> 00:04:38.117
parts.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/918-0
00:04:38.267 --> 00:04:42.317
So using robots for data
collection is not new. Either 

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/946-0
00:04:42.327 --> 00:04:46.401
is using machine learning for
data processing, but automated

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/946-1
00:04:46.401 --> 00:04:50.022
users really highlight the
relationship between

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/946-2
00:04:50.022 --> 00:04:52.737
automation, geoscience and
geoscientists.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1011-0
00:04:54.817 --> 00:04:59.497
AutoZone mapping so far, most of
research focus on large-scale

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1011-1
00:04:59.497 --> 00:05:04.103
features such as fault trace,
large fractures, and seismic

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1011-2
00:05:04.103 --> 00:05:08.857
landslides and usually people
use satellite imagery, InSAR and

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1011-3
00:05:08.857 --> 00:05:13.612
airborne lidar that can provide
submeter/meter or even greater

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1011-4
00:05:13.612 --> 00:05:15.617
ground sampling resolution.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1022-0
00:05:15.707 --> 00:05:17.777
But what about the small-scale
features?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1038-0
00:05:17.887 --> 00:05:21.428
And we have secondary faults
triggered creep, small

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1038-1
00:05:21.428 --> 00:05:23.607
fractures, rock cracks, offsets.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1056-0
00:05:24.707 --> 00:05:27.407
flipped rocks, fault- 
related vegetation.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1092-0
00:05:27.417 --> 00:05:31.747
Then it sends seismic rock force and
if we want to map them, if you

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1092-1
00:05:31.747 --> 00:05:36.009
want to study them, we need a
very high resolution centimeter

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1092-2
00:05:36.009 --> 00:05:37.727
or even millimeter ground

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1117-0
00:05:37.737 --> 00:05:41.853
sampling resolution. 
Those features are usually

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1117-1
00:05:41.853 --> 00:05:46.194
searched and mapped by field
work in regional, small and

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1117-2
00:05:46.194 --> 00:05:47.167
local region.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1139-0
00:05:47.497 --> 00:05:50.941
So our understanding of small
scale geomorphic evidence

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1139-1
00:05:50.941 --> 00:05:53.893
remains limited in a
comprehensive and big-data

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1139-2
00:05:53.893 --> 00:05:54.877
context.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1162-0
00:05:56.327 --> 00:06:01.371
You figure out a way to
automatically search for and map

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1162-1
00:06:01.371 --> 00:06:03.567
those small-scale features.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1169-0
00:06:03.577 --> 00:06:05.627
We can ask the following
questions.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1179-0
00:06:05.857 --> 00:06:08.707
How are the small scale-features
spatially distributed?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1202-0
00:06:08.817 --> 00:06:12.032
How are they correlated with the
each other? And how are

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1202-1
00:06:12.032 --> 00:06:14.127
they correlated with large-scale
features?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1211-0
00:06:14.517 --> 00:06:16.807
How do they reflect the ground
motions?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1219-0
00:06:16.817 --> 00:06:20.087
Can they probabilistically inform
future ruptures?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1228-0
00:06:20.297 --> 00:06:21.827
And how can they improve 

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1233-0
00:06:21.837 --> 00:06:22.657
hazard analysis?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1273-0
00:06:24.087 --> 00:06:27.981
So in my previous work I had
opportunity to look at the rocky

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1273-1
00:06:27.981 --> 00:06:31.435
fault scarp and 
the formation and the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1273-2
00:06:31.435 --> 00:06:35.517
development of rocky fault scarp
involved three major processes.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1354-0
00:06:35.847 --> 00:06:40.514
Initially we have volcanic
cooling which forms those

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1354-1
00:06:40.514 --> 00:06:46.237
columnar joints which are those
large boulders, and then we have

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1354-2
00:06:46.237 --> 00:06:51.255
to collect 14 which create
displacements and also expose

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1354-3
00:06:51.255 --> 00:06:56.626
those large boulders to the
surface and the border is 

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1354-4
00:06:56.626 --> 00:07:02.085
on the top, maybe top hold and
fracture into small pieces and

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1354-5
00:07:02.085 --> 00:07:06.927
the last process is referred as
geomorphic fracturing.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1380-0
00:07:07.117 --> 00:07:11.821
It's probably very obvious that
there are many large rocks on

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1380-1
00:07:11.821 --> 00:07:13.187
rocky fault scarp.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1398-0
00:07:13.397 --> 00:07:18.375
Our hypothesis is
rock distributions reflected the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1398-1
00:07:18.375 --> 00:07:21.827
processes of rocky fault scarp
development.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1418-0
00:07:23.817 --> 00:07:29.557
Of course, while engineering
challenges, how can we map and

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1418-1
00:07:29.557 --> 00:07:31.757
detect all these rocks?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1496-0
00:07:31.767 --> 00:07:36.656
Because the number is too large,
so I came up with a data

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1496-1
00:07:36.656 --> 00:07:41.714
processing pipeline which
start from piloted aircraft

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1496-2
00:07:41.714 --> 00:07:46.519
system and using structure
promotion we can get

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1496-3
00:07:46.519 --> 00:07:51.746
also map and DEM nothing new
here and then we have used deep

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1496-4
00:07:51.746 --> 00:07:56.382
learning which is one branch of
machine learning. Machine

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1496-5
00:07:56.382 --> 00:08:01.440
learning models and we have
posted data processing and then

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1496-6
00:08:01.440 --> 00:08:03.547
we created a semantic map.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1554-0
00:08:03.617 --> 00:08:08.220
If you are familiar with GIS as
you can consider a semantic map

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1554-1
00:08:08.220 --> 00:08:12.464
as a shapefile which composed of
a large number of polygon

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1554-2
00:08:12.464 --> 00:08:16.923
vectors and each polygon vector
describes outline of rock and

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1554-3
00:08:16.923 --> 00:08:21.383
with a semantic map we can
do geomorphic analysis and 

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1554-4
00:08:21.383 --> 00:08:24.547
trying to answer some scientific
questions.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1587-0
00:08:26.647 --> 00:08:32.532
And my research book, one of my
research books is applying

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1587-1
00:08:32.532 --> 00:08:38.417
those cutting edge technologies
for geoscience.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1602-0
00:08:38.707 --> 00:08:42.971
And when I'm applying some
machine learning model to this

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1602-1
00:08:42.971 --> 00:08:45.837
application, there are two
challenges.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1615-0
00:08:46.017 --> 00:08:50.077
One is the first one is
annotation map splitting.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1660-0
00:08:50.227 --> 00:08:53.918
That means when we have, when we
use deep learning model, we

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1660-1
00:08:53.918 --> 00:08:57.850
cannot train deep learning model
with entire map because the map

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1660-2
00:08:57.850 --> 00:09:01.601
is too large and then you create
a lot that requires a lot of

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1660-3
00:09:01.601 --> 00:09:04.747
computation and memory for the
deep learning model.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1695-0
00:09:04.897 --> 00:09:08.926
What we usually do is we split a
large map, a large high

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1695-1
00:09:08.926 --> 00:09:13.238
resolution map to image tiles
and the same time it will also

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1695-2
00:09:13.238 --> 00:09:16.207
split annotation map to
annotation tiles.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1711-0
00:09:16.337 --> 00:09:21.207
So annotation map is something
we draw those polygons on [indiscernible]

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1716-0
00:09:21.577 --> 00:09:21.877
[noise]

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1770-0
00:09:21.887 --> 00:09:26.151
[indiscernible] and it provides a subset
of training data set to the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1770-1
00:09:26.151 --> 00:09:30.272
[indiscernible] model to train it, and
once we have the [indiscernible]

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1770-2
00:09:30.272 --> 00:09:34.180
learning model, once we train
the [indiscernible] 

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1770-3
00:09:34.180 --> 00:09:38.657
learning model, we can use it to
to predict other sites.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1780-0
00:09:39.047 --> 00:09:42.097
And one challenge is we have a
large annotation map.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1796-0
00:09:42.107 --> 00:09:45.512
How can we split the large
annotation map to annotation

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1796-1
00:09:45.512 --> 00:09:45.877
tiles?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1911-0
00:09:46.167 --> 00:09:50.074
The other one is once we have
the well trained [indiscernible]

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1911-1
00:09:50.074 --> 00:09:54.400
learning model and we get the
prediction tiles and how can we

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1911-2
00:09:54.400 --> 00:09:58.656
merge them to get a prediction
map and those are the two
 
3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1911-3
00:09:58.656 --> 00:10:02.563
challenges that hasn't been
solved before and there are

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1911-4
00:10:02.563 --> 00:10:07.098
probably other ways you can you
can kind of solve this these two

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1911-5
00:10:07.098 --> 00:10:11.214
challenges and my focus my
research focus is to come

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1911-6
00:10:11.214 --> 00:10:15.052
up, come up algorithms and data
structure and then can

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1911-7
00:10:15.052 --> 00:10:19.029
efficiently solve these two
problems because when we are

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1911-8
00:10:19.029 --> 00:10:23.494
dealing with large scale maps in
remote sensing and computation

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/1911-9
00:10:23.494 --> 00:10:25.517
efficiency is very important.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2018-0
00:10:25.567 --> 00:10:31.108
And so some results here from
these 200 meter by 500 meter

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2018-1
00:10:31.108 --> 00:10:36.836
study area, we detected more
than 200,000 thousand rocks and

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2018-2
00:10:36.836 --> 00:10:42.940
on this major fault scarp, there
are more than 30,000 rocks from

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2018-3
00:10:42.940 --> 00:10:49.044
this semantic map, we can create
a heat maps by showing rocks

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2018-4
00:10:49.044 --> 00:10:54.773
or green size in meters and
green size feet we can get today the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2018-5
00:10:54.773 --> 00:10:59.938
histogram of green size because
we approximate each

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2018-6
00:10:59.938 --> 00:11:05.948
individual rock with a ellipse
and then we use the eccentricity

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2018-7
00:11:05.948 --> 00:11:10.737
of the ellipse to represent the
narrowness of the rock.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2048-0
00:11:11.007 --> 00:11:14.801
So we can get eccentricity
distribution and shows the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2048-1
00:11:14.801 --> 00:11:18.594
narrow list distribution of
those rocks and also rose

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2048-2
00:11:18.594 --> 00:11:22.247
diagram showing the rock
orientation distributions.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2072-0
00:11:24.627 --> 00:11:30.218
And then we study the coalition
between rock size distribution

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2072-1
00:11:30.218 --> 00:11:33.057
and a fault scarp height.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2111-0
00:11:33.167 --> 00:11:37.812
What we found is on our higher
fault scarp we have smaller big

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2111-1
00:11:37.812 --> 00:11:42.383
in green size and those are the
rocks we present in the rocks

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2111-2
00:11:42.383 --> 00:11:43.857
here and the bottom.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2125-0
00:11:44.007 --> 00:11:46.627
That means a higher rock, higher
fault scarp,

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2153-0
00:11:46.807 --> 00:11:51.922
There are smaller those
fractured rocks and those 

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2153-1
00:11:51.922 --> 00:11:57.501
smaller rocks are kind of 
caused by a

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2153-2
00:11:57.501 --> 00:11:59.547
geomorphic fracturing.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2179-0
00:11:59.737 --> 00:12:04.168
So that means that kind of
indicate higher for the scarp

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2179-1
00:12:04.168 --> 00:12:08.287
has probably accumulated more
geomorphic fracturing.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2208-0
00:12:08.397 --> 00:12:12.880
fault scarp and we found that

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2208-1
00:12:12.880 --> 00:12:15.767
there are larger boulders on the
top.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2221-0
00:12:16.217 --> 00:12:20.589
And those larger boulders are
are displaced by tectonic

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2221-1
00:12:20.589 --> 00:12:21.107
faulting.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2237-0
00:12:21.557 --> 00:12:25.027
So which also kind of indicate
our higher fault scarp 

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2241-0
00:12:25.037 --> 00:12:27.727
has accumulated more
tectonic faulting.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2282-0
00:12:29.827 --> 00:12:34.395
This is probably very obvious,
but this is kind of the 

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2282-1
00:12:34.395 --> 00:12:39.118
studies about the rock evidence
of large scale

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2282-2
00:12:39.118 --> 00:12:42.757
geomorphic features and
processes development.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2294-0
00:12:42.767 --> 00:12:45.607
In the process, we also  

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2329-0
00:12:45.617 --> 00:12:49.835
examined or inspect uh
particle transportation model,

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2329-1
00:12:49.835 --> 00:12:53.985
so the medium size of the rock
of the rocks are similar to a

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2329-2
00:12:53.985 --> 00:12:55.277
size of a football.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2361-0
00:12:55.447 --> 00:12:59.081
If you think you can imagine you
stand on the top of the Rocky

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2361-1
00:12:59.081 --> 00:13:01.157
for Skype and you throw a
football.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2397-0
00:13:02.647 --> 00:13:06.581
And ideally with owl, ideally
without some considering any

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2397-1
00:13:06.581 --> 00:13:10.514
friction and football will end
up with orientation that is

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2397-2
00:13:10.514 --> 00:13:13.047
parallel to the photo scope of
trees.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2425-0
00:13:13.607 --> 00:13:18.695
And we studied rock orientation
distribution and its correlation

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2425-1
00:13:18.695 --> 00:13:23.704
with the the full scope height
to support such kind of particle

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2425-2
00:13:23.704 --> 00:13:25.347
transportation model.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2437-0
00:13:28.117 --> 00:13:30.487
And a rock is just a rock.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2459-0
00:13:30.497 --> 00:13:34.717
Distribution is just one
geomorphic evidence of footing,

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2459-1
00:13:34.717 --> 00:13:38.641
and there are, as we just
mentioned, there are other

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2459-2
00:13:38.641 --> 00:13:39.307
features.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2466-0
00:13:39.457 --> 00:13:42.887
For example, rocket cracks and
displays the rocks.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2487-0
00:13:43.117 --> 00:13:47.830
This is from our Turkey
earthquake earlier this year and

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2487-1
00:13:47.830 --> 00:13:50.227
the rock was was here before.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2512-0
00:13:50.297 --> 00:13:55.104
It's about 2 meter height walk
and from the earthquake the walk

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2512-1
00:13:55.104 --> 00:13:57.657
was displaced by 1.5 to 2
meters.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2526-0
00:13:58.527 --> 00:14:03.337
We should have triggered a creep
and seismic rockfall.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2544-0
00:14:03.547 --> 00:14:09.448
And then you may expect on on
for the trees, the vegetation

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2544-1
00:14:09.448 --> 00:14:12.497
distributions can be different.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2565-0
00:14:15.807 --> 00:14:21.177
So for my research here at
Caltech, I want to kind of

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2565-1
00:14:21.177 --> 00:14:22.867
generalize those.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2581-0
00:14:22.927 --> 00:14:28.471
The rock, the feature detection
from rock detection to a more

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2581-1
00:14:28.471 --> 00:14:32.137
general photo related feature
detection.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2599-0
00:14:32.347 --> 00:14:37.844
For example, rock cracks at both
the fractures that trigger the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2599-1
00:14:37.844 --> 00:14:39.647
creep and vegetation.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2644-0
00:14:42.917 --> 00:14:48.675
And we are thinking about, I
mean, Kate, Devin and me, we are

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2644-1
00:14:48.675 --> 00:14:54.711
thinking about those study area
candidates around Salton Sea, as

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2644-2
00:14:54.711 --> 00:14:59.447
you can see the photo structures
on quite complex.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2675-0
00:14:59.517 --> 00:15:04.831
And this is also an area that is
with less well understood and

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2675-1
00:15:04.831 --> 00:15:07.867
there are many funders as
medicity.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2707-0
00:15:07.877 --> 00:15:12.644
There are swarm of earthquakes,
so if we we will see if we have

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2707-1
00:15:12.644 --> 00:15:17.038
opportunity to so to have new
earthquake and see how those

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2707-2
00:15:17.038 --> 00:15:20.687
features might, how those
fractures may develop.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2736-0
00:15:23.767 --> 00:15:29.076
And they're constantly saying
such a Jong, uh, light show on

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2736-1
00:15:29.076 --> 00:15:31.077
Twitter, which is cool.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2738-0
00:15:31.087 --> 00:15:31.387
OK.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2750-0
00:15:31.397 --> 00:15:35.317
We have Joe phone containment,
but what about for signs?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2763-0
00:15:35.907 --> 00:15:38.477
So it's a swarm UAV mission
planning.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2777-0
00:15:38.647 --> 00:15:42.663
This is quite important because
after an earthquake it's

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2777-1
00:15:42.663 --> 00:15:44.777
important to scale up quickly.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2784-0
00:15:44.787 --> 00:15:46.217
Scale up your mapping.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2819-0
00:15:46.687 --> 00:15:51.097
For example, if we can find a
new fractures and how can we use

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2819-1
00:15:51.097 --> 00:15:55.157
this new fractures to guide
which is in the field so they

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2819-2
00:15:55.157 --> 00:15:58.657
can have a better idea of the
earthquake effects?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2829-0
00:16:03.107 --> 00:16:06.757
Next is a technology
development.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2838-0
00:16:07.367 --> 00:16:11.397
Next, technology development is
automate augmented mapping.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2875-0
00:16:11.407 --> 00:16:15.444
So previous work I didn't assume
we have any prior knowledge of

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2875-1
00:16:15.444 --> 00:16:19.166
the of the study area, but
that's not true because in most

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2875-2
00:16:19.166 --> 00:16:21.437
cases we have some prior
knowledge.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2912-0
00:16:21.707 --> 00:16:25.282
If we have those prior
knowledge, for example, if we

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2912-1
00:16:25.282 --> 00:16:29.464
know the photo map and then we
can use this photo map to have

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2912-2
00:16:29.464 --> 00:16:31.757
information guided path
planning.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2920-0
00:16:31.927 --> 00:16:35.887
To which we can improve the Joan
mapping efficiency.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2939-0
00:16:39.027 --> 00:16:43.517
Also, we want to support this
research photo zone mapping.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2968-0
00:16:43.527 --> 00:16:47.606
We we want to look at the the
coalitions between features and

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2968-1
00:16:47.606 --> 00:16:50.697
features and features and the
ground emotions.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2995-0
00:16:50.927 --> 00:16:55.779
And if we like, we do find those
correlations, these strong

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/2995-1
00:16:55.779 --> 00:17:00.227
correlations, and we can do
active search and mapping.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3041-0
00:17:00.237 --> 00:17:04.232
That means if we can find the
feature A and also feature A and

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3041-1
00:17:04.232 --> 00:17:07.719
feature B have strong
correlations, and very likely we

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3041-2
00:17:07.719 --> 00:17:11.650
will find if we find feature a
from the drone and very likely

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3041-3
00:17:11.650 --> 00:17:14.567
the drone can also find feature
being nearby.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3054-0
00:17:15.077 --> 00:17:16.587
Similar idea.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3084-0
00:17:16.727 --> 00:17:21.038
If we find earthquake and follow
the ground, motion has some

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3084-1
00:17:21.038 --> 00:17:24.642
correlation with the some
features and we have new

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3084-2
00:17:24.642 --> 00:17:29.093
earthquake that we can probably
use this information to search

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3084-3
00:17:29.093 --> 00:17:30.577
new new new raptures.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3108-0
00:17:32.517 --> 00:17:36.149
OK, moving from Photozone
mapping to fragile Judge Koh

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3108-1
00:17:36.149 --> 00:17:40.507
features, I specifically look at
the precariously balanced rocks.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3127-0
00:17:40.577 --> 00:17:44.870
So those are the rocks that are
balanced down but not attached

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3127-1
00:17:44.870 --> 00:17:45.687
to pedestal.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3171-0
00:17:45.907 --> 00:17:50.412
So if you go to field, you see a
PR and that might indicate for a

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3171-1
00:17:50.412 --> 00:17:54.507
while there hasn't been large
earthquakes and also not just

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3171-2
00:17:54.507 --> 00:17:55.667
the PR's on ours.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3194-0
00:17:55.677 --> 00:18:00.845
Ours, we also found PPRS on
Mars, so those are the photos

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3194-1
00:18:00.845 --> 00:18:02.627
from Mars 2020 over.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3208-0
00:18:04.037 --> 00:18:08.655
And there are many ways you can
study the the fragility of PBR

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3208-1
00:18:08.655 --> 00:18:10.487
and one simple parameter.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3241-0
00:18:10.497 --> 00:18:14.621
One intuitive parameter is
minimum contact angle, so you

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3241-1
00:18:14.621 --> 00:18:18.890
can see after an earthquake if
PBR is toppled and then the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3241-2
00:18:18.890 --> 00:18:21.567
minimum contact angle will
increase.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3252-0
00:18:21.577 --> 00:18:25.107
So this minimum contact angle is
1 parameter.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3259-0
00:18:25.117 --> 00:18:28.137
You can describe how fragile a
PBR is.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3276-0
00:18:30.437 --> 00:18:33.967
And a reason a reason to
research by two men.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3296-0
00:18:34.257 --> 00:18:38.346
And they are looking at this
study area and between the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3296-1
00:18:38.346 --> 00:18:41.047
border of California and the
Nevada.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3312-0
00:18:41.117 --> 00:18:47.265
So even before 2000 and the
James James Booty founded those

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3312-1
00:18:47.265 --> 00:18:47.777
PR's.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3329-0
00:18:48.357 --> 00:18:50.747
Uh, it's quite pretty.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3351-0
00:18:50.757 --> 00:18:55.416
And there's just balanced here,
but after a recent earthquake, a

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3351-1
00:18:55.416 --> 00:18:59.717
man be magnitude 6 earthquake
and the PR's are still there.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3358-0
00:19:00.877 --> 00:19:02.777
But they also observed the rock
falls.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3400-0
00:19:03.857 --> 00:19:09.259
So what they found is OK, the
PGA is PGA of point, 3G is

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3400-1
00:19:09.259 --> 00:19:14.945
probably large enough to table
PBR, but they also found the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3400-2
00:19:14.945 --> 00:19:18.167
tuition is too small to table
it.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3423-0
00:19:18.247 --> 00:19:24.510
So to tumble PMR to topple PR,
you need both large PGA and

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3423-1
00:19:24.510 --> 00:19:28.437
duration, so that bring us bring
us.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3441-0
00:19:28.447 --> 00:19:32.268
That brings us a questions or
what constitutes fragile

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3441-1
00:19:32.268 --> 00:19:34.907
judgable features in a broader
sense.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3492-0
00:19:35.007 --> 00:19:39.994
So we know PR's and Rock Peters
are typically considered as

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3492-1
00:19:39.994 --> 00:19:45.480
fragile geological features, but
in this case we have block falls

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3492-2
00:19:45.480 --> 00:19:50.716
and and but we didn't observe
the table PBR, so our how can we

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3492-3
00:19:50.716 --> 00:19:53.957
quantify the the fragility of an
area?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3520-0
00:19:53.967 --> 00:19:58.467
Can we also come up with some
idea to to quantify the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3520-1
00:19:58.467 --> 00:20:01.967
fragility of lab, landslide and
rockfall?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3590-0
00:20:02.217 --> 00:20:07.069
And also previously I mean
existing researchers only look

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3590-1
00:20:07.069 --> 00:20:12.255
at no one or several individual
specific PBR's, but there are

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3590-2
00:20:12.255 --> 00:20:17.358
many PR's and if we can map and
do, if we can map and detect

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3590-3
00:20:17.358 --> 00:20:22.795
those large number of PR's, how
can a large number of FGFS offer

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3590-4
00:20:22.795 --> 00:20:25.137
insights into going motions?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3615-0
00:20:25.527 --> 00:20:29.439
Because if we have a few samples
and we the samples might be

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3615-1
00:20:29.439 --> 00:20:32.517
biased by free, look at the
large distribution.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3628-0
00:20:32.597 --> 00:20:35.327
And might be a better or more
accurate representation.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3646-0
00:20:37.907 --> 00:20:41.857
So in my previous work I have
this data processing pipeline.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3718-0
00:20:42.067 --> 00:20:45.768
I have US survey and soft
structure from motion and we

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3718-1
00:20:45.768 --> 00:20:49.805
have also mosaic and use the
same 2 drug detection model to

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3718-2
00:20:49.805 --> 00:20:54.043
detect the PBR's and also map
and then we use the bounding box

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3718-3
00:20:54.043 --> 00:20:58.080
to crop the PR's on point cloud
and we have another machine

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3718-4
00:20:58.080 --> 00:21:01.983
learning deep learning module to
do the segmentation rock

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3718-5
00:21:01.983 --> 00:21:02.857
segmentation.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3721-0
00:21:03.407 --> 00:21:04.417
Rock segmentation.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3729-0
00:21:04.427 --> 00:21:07.317
So this method is offboard
method.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3737-0
00:21:07.407 --> 00:21:08.417
What does it that means?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3760-0
00:21:08.977 --> 00:21:12.856
So outward method is separates
the data collection and data

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3760-1
00:21:12.856 --> 00:21:13.567
processing.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3807-0
00:21:13.637 --> 00:21:18.371
So in this case we have to pre
plan a flight trajectory for the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3807-1
00:21:18.371 --> 00:21:23.031
drone and the drone collect the
data and then we can come back

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3807-2
00:21:23.031 --> 00:21:27.248
to office and then we can
process the data on the laptop

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3807-3
00:21:27.248 --> 00:21:28.727
or survey or server.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3851-0
00:21:30.727 --> 00:21:35.193
And the problem is if we we plan
a flight trajectory for the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3851-1
00:21:35.193 --> 00:21:39.367
drone and in some case the
minimum contact angle is very

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3851-2
00:21:39.367 --> 00:21:43.833
small and the drone may miss
this, this very important basal

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3851-3
00:21:43.833 --> 00:21:45.297
contact information.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3864-0
00:21:46.087 --> 00:21:52.357
So to solve this problem I come
up with a onboard method.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3873-0
00:21:52.467 --> 00:21:57.817
So we build these customized UV
from scratch.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3880-0
00:21:57.907 --> 00:22:00.117
This one has two companion
computers.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3893-0
00:22:00.127 --> 00:22:02.177
One is in charge of trajectory
planning.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3934-0
00:22:02.607 --> 00:22:07.356
Another one has an age GPU that
enables onboard machine learning

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3934-1
00:22:07.356 --> 00:22:11.593
in real time and RTK GPS and two
stereo cameras 1/4 joint

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3934-2
00:22:11.593 --> 00:22:15.027
localization, the other one for
drone mapping.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3985-0
00:22:15.997 --> 00:22:20.321
So again, the off board method
is kind of separate the data

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3985-1
00:22:20.321 --> 00:22:24.717
processing and the data, data
collection and data processing

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3985-2
00:22:24.717 --> 00:22:29.040
and on board method because we
we we we equipped those very

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3985-3
00:22:29.040 --> 00:22:33.364
powerful computers and we can
integrate data collection and

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3985-4
00:22:33.364 --> 00:22:34.517
data processing.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/3995-0
00:22:37.577 --> 00:22:39.667
So the idea is very
straightforward.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4010-0
00:22:39.677 --> 00:22:42.107
So we haven't drawn do this.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4032-0
00:22:42.117 --> 00:22:45.869
Do this survey from a high lawn
mower survey from a high

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4032-1
00:22:45.869 --> 00:22:46.527
elevation.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4069-0
00:22:46.917 --> 00:22:51.627
Once it can detect the PBR from
this camera and the real time

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4069-1
00:22:51.627 --> 00:22:55.957
machine learning and then it
will move around to collect

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4069-2
00:22:55.957 --> 00:22:59.908
images from multiple
perspectives to verify if it's

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4069-3
00:22:59.908 --> 00:23:00.667
PR or not.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4085-0
00:23:00.797 --> 00:23:03.298
Because if we use machine
learning, we may have false

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4085-1
00:23:03.298 --> 00:23:03.807
detections.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4121-0
00:23:03.957 --> 00:23:08.398
This this process to just verify
the PBR candidates and once the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4121-1
00:23:08.398 --> 00:23:12.565
PR is verified they will fly
closer to the PBR and conduct a

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4121-2
00:23:12.565 --> 00:23:15.297
circular motion to map it in
real time.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4126-0
00:23:18.817 --> 00:23:19.247
Cool.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4147-0
00:23:19.257 --> 00:23:23.836
Once we can map once, we can map
PR's and then we need to study

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4147-1
00:23:23.836 --> 00:23:25.267
the dynamics of PBS.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4161-0
00:23:25.507 --> 00:23:31.447
And so I build a virtual shake
robot and a mini shake robot.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4174-0
00:23:31.857 --> 00:23:34.467
It's essentially the those are
shake tables.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4216-0
00:23:34.477 --> 00:23:38.764
I I call it robots because I use
the technologies and concepts in

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4216-1
00:23:38.764 --> 00:23:42.531
robots like the control and
perception model and also use

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4216-2
00:23:42.531 --> 00:23:46.558
the tools in robotics to build
those two virtual and physical

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4216-3
00:23:46.558 --> 00:23:47.207
shippable.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4232-0
00:23:48.297 --> 00:23:50.887
And you can see and it's not
just the the the.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4254-0
00:23:50.947 --> 00:23:54.245
The model is pretty
straightforward and base near

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4254-1
00:23:54.245 --> 00:23:58.267
linear rail and a pedestal and
PVR which very training here.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4305-0
00:23:58.277 --> 00:24:03.325
It's actually 22.5 meter height
and also we can replace the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4305-1
00:24:03.325 --> 00:24:08.793
pedestal with realistic terrains
and mapped by UV and do uh from

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4305-2
00:24:08.793 --> 00:24:14.008
the meaning from the mini shake
table and we can use it to to

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4305-3
00:24:14.008 --> 00:24:17.457
study the overturning of 3D
printed PBR.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4321-0
00:24:17.547 --> 00:24:22.917
So this is a small downsized 3D
printed PBR.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4331-0
00:24:23.007 --> 00:24:25.637
Just just the purpose is to
verify the simulation.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4336-0
00:24:28.027 --> 00:24:28.697
OK.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4361-0
00:24:28.827 --> 00:24:32.411
I particularly look at it two
processes, one is overturning,

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4361-1
00:24:32.411 --> 00:24:34.937
so this is a short term reaction
from PBR.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4396-0
00:24:34.947 --> 00:24:39.840
So with the ground motion, if a
PBR will topple, be not be not

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4396-1
00:24:39.840 --> 00:24:44.577
and and the other one is a long
term long term PBR response.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4404-0
00:24:44.587 --> 00:24:46.297
Recall large displacement.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4449-0
00:24:46.307 --> 00:24:50.640
So once the PR is toppled and
the what the trajectory it can

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4449-1
00:24:50.640 --> 00:24:54.546
possibly have and we have 12
parallel simulations here

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4449-2
00:24:54.546 --> 00:24:58.382
because people are not
symmetric, I mean the geometry

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4449-3
00:24:58.382 --> 00:25:00.157
people are not symmetric.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4465-0
00:25:00.167 --> 00:25:03.887
So we want to study the
overturning response with with

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4465-1
00:25:03.887 --> 00:25:06.457
respect to different ground
emotions.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4474-0
00:25:06.547 --> 00:25:08.297
So for those 12 similar.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4488-0
00:25:08.307 --> 00:25:12.758
So those for those 12 parallel
simulations, they have different

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4488-1
00:25:12.758 --> 00:25:14.427
initial PR orientations.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4497-0
00:25:17.137 --> 00:25:18.367
So why don't we learn this?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4501-0
00:25:20.317 --> 00:25:20.787
Uh.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4519-0
00:25:20.797 --> 00:25:25.170
From overturning experiment we
can get upper bound ground

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4519-1
00:25:25.170 --> 00:25:26.527
motion constraint.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4525-0
00:25:26.657 --> 00:25:27.767
So the red dots.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4536-0
00:25:27.777 --> 00:25:30.587
So this is the overturning
response diagram.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4563-0
00:25:30.657 --> 00:25:34.623
So the reader dots represent the
status that PR is coupled from

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4563-1
00:25:34.623 --> 00:25:38.217
the ground motion, and the
Buddhas represent the balanced

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4563-2
00:25:38.217 --> 00:25:38.527
ctas.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4590-0
00:25:39.017 --> 00:25:42.573
So if you go to field, if you
find the PR, that means if you

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4590-1
00:25:42.573 --> 00:25:45.137
find the PR, that means the
quantum motion.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4615-0
00:25:47.157 --> 00:25:53.774
Is is not hasn't he hasn't
exceeded this this this upper

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4615-1
00:25:53.774 --> 00:25:55.747
bound constraint.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4665-0
00:25:55.757 --> 00:26:00.301
This boundary and from a large
displacement experiment and for

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4665-1
00:26:00.301 --> 00:26:04.989
example, if you go to field and
you feel you, you can see if you

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4665-2
00:26:04.989 --> 00:26:09.677
find a tuple PBR and if you also
happen to know it's trajectory.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4706-0
00:26:09.687 --> 00:26:14.052
For example, if we say if you
know the trajectory is 30 meters

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4706-1
00:26:14.052 --> 00:26:18.002
and then from this large
displacement experiment we know

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4706-2
00:26:18.002 --> 00:26:22.298
the ground motion for example
this one has to be greater than

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4706-3
00:26:22.298 --> 00:26:23.337
this threshold.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4725-0
00:26:23.587 --> 00:26:27.690
Otherwise, if it's smaller, it
cannot really result trajectory

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4725-1
00:26:27.690 --> 00:26:28.797
that is 30 meter.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4745-0
00:26:28.867 --> 00:26:32.987
So from the large displacement
experiment, we can we can get a

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4745-1
00:26:32.987 --> 00:26:35.407
lower bounded one motion
constraint.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4776-0
00:26:35.817 --> 00:26:39.808
So with so we can combine
overturning and large

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4776-1
00:26:39.808 --> 00:26:44.546
displacement, we can have a
better upper bound and lower

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4776-2
00:26:44.546 --> 00:26:49.617
bound ground motion constraints
for for the hazard analysis.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4791-0
00:26:52.407 --> 00:26:54.797
And there are some limitations.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4832-0
00:26:54.807 --> 00:26:59.460
In my previous previous work,
the off word method off word off

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4832-1
00:26:59.460 --> 00:27:03.743
word mapping method because I
used deep learning model to

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4832-2
00:27:03.743 --> 00:27:08.100
segment point clouds and it is
supervised machine learning

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4832-3
00:27:08.100 --> 00:27:08.617
models.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4883-0
00:27:08.627 --> 00:27:12.880
That means I need annotations
and annotating a point clouds is

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4883-1
00:27:12.880 --> 00:27:17.066
very different annotating images
because annotating point has

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4883-2
00:27:17.066 --> 00:27:21.117
very time consuming and for the
onboard mapping method, the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4883-3
00:27:21.117 --> 00:27:23.817
flight time is quite quite a
challenge.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4925-0
00:27:24.107 --> 00:27:27.877
And also I didn't consider
obstacles because in reality and

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4925-1
00:27:27.877 --> 00:27:31.520
there are there might be
obstacles around PR, for example

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4925-2
00:27:31.520 --> 00:27:34.347
vegetations or PR that is very
close to all.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4950-0
00:27:35.287 --> 00:27:40.610
And the virtual Shaker World War
Two only has one one degree, one

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4950-1
00:27:40.610 --> 00:27:44.077
emotion, and this is one major
limitation.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4979-0
00:27:44.187 --> 00:27:50.166
So for for the PBR study, we are
looking at the same study area

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4979-1
00:27:50.166 --> 00:27:54.837
and that's Jim Broom and Trugman
Dukan to before.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/4998-0
00:27:57.247 --> 00:28:01.407
Now we just went there like uh,
from from Friday to no.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5007-0
00:28:01.417 --> 00:28:05.067
From Sunday to and we come back,
we came back yesterday night.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5036-0
00:28:06.707 --> 00:28:10.499
And so to improve the off
boarding method, I have this

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5036-1
00:28:10.499 --> 00:28:13.257
idea of semantic structure from
motion.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5057-0
00:28:13.267 --> 00:28:17.434
So structure from motion is not
new, but how can we use create

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5057-1
00:28:17.434 --> 00:28:20.277
semantics for structure motions,
products?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5083-0
00:28:20.387 --> 00:28:23.759
So if we can combine deep
learning and structure from

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5083-1
00:28:23.759 --> 00:28:27.568
motion, we can get the point
cloud that is, that is for each

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5083-2
00:28:27.568 --> 00:28:28.817
point is classified.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5098-0
00:28:31.357 --> 00:28:35.507
And also off word method has
some limitations.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5119-0
00:28:35.517 --> 00:28:38.875
It couldn't really see the basal
contact information, but it

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5119-1
00:28:38.875 --> 00:28:41.682
still can provide us some
information about the PR

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5119-2
00:28:41.682 --> 00:28:42.287
candidates.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5180-0
00:28:42.397 --> 00:28:46.196
So some of them, even them, are
not that fragile, but it can

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5180-1
00:28:46.196 --> 00:28:49.373
provide the candidates
information for the onboard

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5180-2
00:28:49.373 --> 00:28:52.923
method and then onboarding
method can save the five time

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5180-3
00:28:52.923 --> 00:28:56.785
and skip skip searching, skip
the time for searching and only

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5180-4
00:28:56.785 --> 00:28:59.587
focus on the mapping of those PR
candidates.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5192-0
00:29:02.057 --> 00:29:04.147
And what I'm particularly
interesting.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5195-0
00:29:04.407 --> 00:29:04.667
Uh.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5222-0
00:29:04.677 --> 00:29:09.337
Combining UAV, LIDAR and
structure for motion so from UB

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5222-1
00:29:09.337 --> 00:29:11.707
Lidar it can, it can provide.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5322-0
00:29:12.057 --> 00:29:16.904
It can provide a good accuracy
about the about the point cloud,

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5322-1
00:29:16.904 --> 00:29:21.525
but in the resolution that are
very good and I also couldn't

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5322-2
00:29:21.525 --> 00:29:25.614
provide a good texture and
precision is not very good

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5322-3
00:29:25.614 --> 00:29:30.083
because the precision of the
points might uh slow floating

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5322-4
00:29:30.083 --> 00:29:34.703
around some centimeter and UAB
structure from motion mapping

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5322-5
00:29:34.703 --> 00:29:39.475
system and can provide a good
position and also the texture is

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5322-6
00:29:39.475 --> 00:29:43.943
also good but the quality is not
very good and also it has

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5322-7
00:29:43.943 --> 00:29:47.957
problem of global deformation
and local deformation.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5373-0
00:29:48.117 --> 00:29:51.990
The global deformation can be
corrected by ground control

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5373-1
00:29:51.990 --> 00:29:56.263
points, but if we want to study
small skill features like those

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5373-2
00:29:56.263 --> 00:30:00.336
triggered creep that only have
the development of three to 7

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5373-3
00:30:00.336 --> 00:30:04.543
millimeters a year, we do need
to to fix the local deformation

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5373-4
00:30:04.543 --> 00:30:05.077
problem.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5400-0
00:30:05.827 --> 00:30:11.354
So if we can combine LIDAR and
structure from motion, and to do

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5400-1
00:30:11.354 --> 00:30:16.277
that can provide us a, maybe a
new novel mapping system.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5426-0
00:30:18.617 --> 00:30:22.473
And also virtual Shaker robot
that kind of want to expand the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5426-1
00:30:22.473 --> 00:30:26.142
from one one degree granting
motion to three degree ground

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5426-2
00:30:26.142 --> 00:30:26.577
motion.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5486-0
00:30:28.587 --> 00:30:32.697
And in you may see that can
there are many projects, many

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5486-1
00:30:32.697 --> 00:30:36.522
development, technology
development and but projects,

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5486-2
00:30:36.522 --> 00:30:40.348
but there are quite a large
overlap between these two

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5486-3
00:30:40.348 --> 00:30:44.529
studies for for example, the
semantic structure for motion

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5486-4
00:30:44.529 --> 00:30:48.142
can be used for photozone
mapping and also fragile

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5486-5
00:30:48.142 --> 00:30:49.417
tragical features.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5507-0
00:30:50.027 --> 00:30:54.122
And this is one advantage of
this of of such kind of

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5507-1
00:30:54.122 --> 00:30:54.817
research.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5523-0
00:30:54.927 --> 00:30:58.592
Once the technology is
developed, we can we can easily

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5523-1
00:30:58.592 --> 00:31:01.657
transfer that the technology to
other domain.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5530-0
00:31:04.057 --> 00:31:05.737
And project management.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5610-0
00:31:05.957 --> 00:31:10.412
So we usually have off board,
off board or offline method that

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5610-1
00:31:10.412 --> 00:31:14.796
use commercial drones and we
just use a drone to collect data

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5610-2
00:31:14.796 --> 00:31:18.685
and then we focus on data
processing also have onboard

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5610-3
00:31:18.685 --> 00:31:22.786
method and onboard requires
hardware development and this

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5610-4
00:31:22.786 --> 00:31:26.675
sounds easy, but it's quite
there are many components,

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5610-5
00:31:26.675 --> 00:31:30.847
mechanical design, electrical
design and you need to think

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5610-6
00:31:30.847 --> 00:31:32.827
about what sensors you need.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5621-0
00:31:32.837 --> 00:31:35.774
You need to use and in the
battery management

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5621-1
00:31:35.774 --> 00:31:36.667
communication.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5659-0
00:31:37.067 --> 00:31:41.043
The communication between the UV
and the operator and if you want

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5659-1
00:31:41.043 --> 00:31:44.718
to have multiple UV mapping
system you also need to consider

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5659-2
00:31:44.718 --> 00:31:48.393
communication among Jones and
simulation tests, because once

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5659-3
00:31:48.393 --> 00:31:49.417
you develop this.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5677-0
00:31:50.737 --> 00:31:54.007
Hardware and you just go to feel
the very likely you will.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5686-0
00:31:54.017 --> 00:31:55.567
You will crash it so.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5716-0
00:31:55.677 --> 00:31:58.999
So usually we have we will test
the Jones and test the software

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5716-1
00:31:58.999 --> 00:32:01.957
in a simulation and we are ready
and we have field test.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5766-0
00:32:03.047 --> 00:32:06.632
So the off board method is the
offline method is more

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5766-1
00:32:06.632 --> 00:32:10.217
straightforward and the online
method is require more

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5766-2
00:32:10.217 --> 00:32:14.466
technology development, but the
online method is also promising

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5766-3
00:32:14.466 --> 00:32:18.648
and can do many things that the
the off boarding method cannot

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5766-4
00:32:18.648 --> 00:32:18.847
do.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5776-0
00:32:21.357 --> 00:32:22.697
Let's go back to these.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5779-0
00:32:22.707 --> 00:32:23.287
Uh.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5795-0
00:32:23.327 --> 00:32:27.012
Ultimate issues science because
I'm very interested in

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5795-1
00:32:27.012 --> 00:32:29.357
philosophy of some of
methodology.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5800-0
00:32:30.087 --> 00:32:31.997
So all this is start from.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5809-0
00:32:32.827 --> 00:32:36.237
Umm geoscientific observations.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5820-0
00:32:36.247 --> 00:32:39.457
But what if geoscientific
observations are not accessible

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5820-1
00:32:39.457 --> 00:32:40.397
or interpretable?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5837-0
00:32:40.607 --> 00:32:43.117
This sounds are abstract
question, but it's quite common.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5864-0
00:32:43.127 --> 00:32:47.412
So before Mars Rover, close
range marks observations are not

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5864-1
00:32:47.412 --> 00:32:51.345
accessible and volcanoes and
deep oceans are not easily

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5864-2
00:32:51.345 --> 00:32:52.117
accessible.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5871-0
00:32:52.207 --> 00:32:54.477
Nowadays we have too much data.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5878-0
00:32:54.927 --> 00:32:55.857
There's too much data.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5895-0
00:32:55.867 --> 00:33:00.581
We cannot really interpolate
them by person, so in this case,

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5895-1
00:33:00.581 --> 00:33:01.417
how can we?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5902-0
00:33:01.507 --> 00:33:03.807
How can we handle those
situations?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5941-0
00:33:04.317 --> 00:33:08.192
So what I'm what's the automated
jewel size is proposing is to

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5941-1
00:33:08.192 --> 00:33:11.576
use robots to make the sides
more accessible and using

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5941-2
00:33:11.576 --> 00:33:14.897
machine learning to make the
data more interpretable.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5978-0
00:33:16.287 --> 00:33:21.109
And then we kind of expanded the
idea of automatic Jose as to a

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5978-1
00:33:21.109 --> 00:33:25.555
general automated Jews as so we
can use robots and machine

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/5978-2
00:33:25.555 --> 00:33:29.397
learning to to assist Geo
scientific observations.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6016-0
00:33:29.467 --> 00:33:34.515
So I I for a short time short
term, I don't think we can like

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6016-1
00:33:34.515 --> 00:33:39.319
really use robots and the
machine running to replace human

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6016-2
00:33:39.319 --> 00:33:40.377
observations.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6032-0
00:33:40.527 --> 00:33:44.007
But now we can do this very high
likelihood.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6061-0
00:33:44.017 --> 00:33:48.991
We can use these tools to assist
observations, so this idea is we

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6061-1
00:33:48.991 --> 00:33:52.307
will have more exploration
driven research.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6089-0
00:33:52.617 --> 00:33:56.048
So sometimes we just have to
explore more, collect more data

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6089-1
00:33:56.048 --> 00:33:59.705
and honest data, understand the
data better, and then we can ask

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6089-2
00:33:59.705 --> 00:34:01.167
more meaningful questions.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6107-0
00:34:03.667 --> 00:34:07.179
And then particularly interested
in one application, it's

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6107-1
00:34:07.179 --> 00:34:09.237
disaster decision support
system.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6135-0
00:34:09.357 --> 00:34:13.917
This is different from early
warning, so this is we we have

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6135-1
00:34:13.917 --> 00:34:17.717
data from centralized robots and
sensor networks.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6155-0
00:34:17.807 --> 00:34:22.241
How can we use this data and
combine them to build a disaster

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6155-1
00:34:22.241 --> 00:34:23.957
decision support system?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6163-0
00:34:24.007 --> 00:34:26.877
So once earthquake were tornado?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6196-0
00:34:26.987 --> 00:34:32.580
Hurricane Arthur and and then we
can integrate this data and use

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6196-1
00:34:32.580 --> 00:34:37.657
that to assist a search and
rescue and also maybe observe.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6208-0
00:34:38.147 --> 00:34:43.667
Uh recover weight to your long
term and that's my presentation.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6212-0
00:34:43.677 --> 00:34:44.877
Thank you every.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6221-0
00:34:45.007 --> 00:34:48.157
Thank you, everyone and I'm
happy to take questions.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6228-0
00:34:54.747 --> 00:34:55.647
Thank you so much.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6243-0
00:34:56.477 --> 00:35:02.924
Umm, so there's there's people
clapping here, I guess well of

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6243-1
00:35:02.924 --> 00:35:04.067
my Moffett.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6249-0
00:35:04.077 --> 00:35:05.237
Menlo gets warmed up.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6257-0
00:35:05.247 --> 00:35:06.587
If there's any questions in the
room.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6259-0
00:35:07.967 --> 00:35:08.227
Yeah.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6261-0
00:35:08.287 --> 00:35:08.937
Yeah.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6270-0
00:35:08.987 --> 00:35:10.967
So thanks for really excited
presentation.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6307-0
00:35:11.047 --> 00:35:16.447
I was wondering on the so the
figure you showed where you had

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6307-1
00:35:16.447 --> 00:35:21.759
the PGA versus ratio PGA PGB
toppling versus not there was a

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6307-2
00:35:21.759 --> 00:35:22.717
clear line.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6415-0
00:35:22.727 --> 00:35:27.244
But to me, whether rock was
gonna toggle or not, if a given

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6415-1
00:35:27.244 --> 00:35:31.686
acceleration, for example do
obviously some uncertainty in

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6415-2
00:35:31.686 --> 00:35:36.279
there so you don't know the
contact between the rock and its

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6415-3
00:35:36.279 --> 00:35:38.537
pedestal and strong it is ohh.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6419-0
00:35:38.547 --> 00:35:42.726
Thinking about ways to address
or include that kind of

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6419-1
00:35:42.726 --> 00:35:47.285
uncertainty in uh in these in
these types of plots, because

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6419-2
00:35:47.285 --> 00:35:52.072
part of the question goes to to
understand what ground motions

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6419-3
00:35:52.072 --> 00:35:56.783
have happened in the past after
we think about whether or not

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6419-4
00:35:56.783 --> 00:36:01.494
brown motion has been exceeded
about the and based on whether

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6419-5
00:36:01.494 --> 00:36:04.077
or not a fragile too one
feature.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6446-0
00:36:04.767 --> 00:36:09.563
So but if you have a bunch of
fragile job geologic features,

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6446-1
00:36:09.563 --> 00:36:14.594
maybe there's probability that
some fraction of them believe in

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6446-2
00:36:14.594 --> 00:36:14.987
that.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6456-0
00:36:16.267 --> 00:36:16.567
Sure.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6460-0
00:36:16.577 --> 00:36:17.847
Of my concert?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6462-0
00:36:18.437 --> 00:36:18.747
Yeah.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6472-0
00:36:18.757 --> 00:36:21.777
So the first part is. So this is
a.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6487-0
00:36:21.897 --> 00:36:27.085
This is a overturning response
from experiment from one ground

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6487-1
00:36:27.085 --> 00:36:28.567
emotion direction.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6570-0
00:36:28.737 --> 00:36:32.864
So in simulation it's
deterministic and we can we can

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6570-1
00:36:32.864 --> 00:36:37.067
eventually we do need a
probabilistic model if we want

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6570-2
00:36:37.067 --> 00:36:41.651
to incorporate it for hazard
analysis and what we can do is

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6570-3
00:36:41.651 --> 00:36:46.466
we can do Monte Carlo and for
example we can change the we can

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6570-4
00:36:46.466 --> 00:36:51.051
kind of like slightly disturb
the orientation a little bit,

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6570-5
00:36:51.051 --> 00:36:55.483
not very large distribution
distribution change not large

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6570-6
00:36:55.483 --> 00:36:58.157
orientation change but very
small.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6593-0
00:36:58.167 --> 00:37:02.054
And we do a Monte Carlo
simulations and we can get, uh,

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6593-1
00:37:02.054 --> 00:37:06.287
we can build a probabilistic
model for overturning response.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6601-0
00:37:07.447 --> 00:37:08.697
And I didn't capture.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6613-0
00:37:08.707 --> 00:37:12.617
The later part is there also
that the question not really.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6648-0
00:37:12.627 --> 00:37:17.578
So I guess the other piece is
just the sort of segmentation

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6648-1
00:37:17.578 --> 00:37:22.364
that between off the pedestal
potentially and how how you

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6648-2
00:37:22.364 --> 00:37:24.427
understands the range of.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6652-0
00:37:26.177 --> 00:37:27.127
Possibilities.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6662-0
00:37:27.197 --> 00:37:30.587
Yeah, it's very it's very
difficult the.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6689-0
00:37:31.587 --> 00:37:36.144
So the dynamics of PR is very
complex, is definitely

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6689-1
00:37:36.144 --> 00:37:40.357
nonlinear, and even small change
in the contact.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6706-0
00:37:40.367 --> 00:37:47.127
My hugely affected the fragility
and there are hard research by

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6706-1
00:37:47.127 --> 00:37:48.817
Christine Wedge.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6721-0
00:37:48.967 --> 00:37:52.037
They did some work about the
very sophisticated,

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6721-1
00:37:52.037 --> 00:37:54.167
sophisticated models of
studying.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6738-0
00:37:55.567 --> 00:37:58.929
Follow the contact how most
small change of contact affect

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6738-1
00:37:58.929 --> 00:37:59.727
the fragility.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6761-0
00:38:00.017 --> 00:38:05.681
So I'm just thinking, well, we
can build, we can build a very

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6761-1
00:38:05.681 --> 00:38:10.887
accurate world, very complex
model for individual PBR's.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6799-0
00:38:11.197 --> 00:38:15.833
But if we really nap a large
number of PR to see if we can

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6799-1
00:38:15.833 --> 00:38:20.547
map 101,000 PBR, do we still
need a very accurate model for

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6799-2
00:38:20.547 --> 00:38:21.647
individual PR?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6813-0
00:38:22.637 --> 00:38:26.687
Because we have a large number
of people's, maybe we will see.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6835-0
00:38:26.697 --> 00:38:30.539
I don't know, but there's a very
good opportunity that we can

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6835-1
00:38:30.539 --> 00:38:33.947
that can provide a different
perspective, right? Yeah.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6861-0
00:38:38.277 --> 00:38:43.582
So the potential of repeat
surveys to potentially look for

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6861-1
00:38:43.582 --> 00:38:48.347
creep along faults scribes
Ridgecrest comes to mind.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6913-0
00:38:48.827 --> 00:38:52.914
Do you think you can get you the
resolution to maybe tease out

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6913-1
00:38:52.914 --> 00:38:56.157
whether there is slow
deformation happening along

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6913-2
00:38:56.157 --> 00:39:00.179
those uh surface exposures of
rupture is and whatnot, is that

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6913-3
00:39:00.179 --> 00:39:03.487
do you think that's possible
with this technology?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6920-0
00:39:03.557 --> 00:39:05.587
That's that's what I'm trying to
do.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6923-0
00:39:05.737 --> 00:39:05.957
We.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6953-0
00:39:05.967 --> 00:39:11.373
I'll I'll join the fob a
campaign and we will look at

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6953-1
00:39:11.373 --> 00:39:16.077
some fractures and then from a
high elevation.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6988-0
00:39:16.087 --> 00:39:22.410
For example, if the drone if the
above ground level for the drone

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6988-1
00:39:22.410 --> 00:39:28.637
is 13 meter 30 meters, they can
provide 1 centimeter resolution,

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/6988-2
00:39:28.637 --> 00:39:33.427
which is not enough to capture a
triggered creep.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7031-0
00:39:33.677 --> 00:39:41.444
But if we can map the map the
the existing fractures and and

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7031-1
00:39:41.444 --> 00:39:49.083
we can build a a model certain
of what 3D model of that and

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7031-2
00:39:49.083 --> 00:39:56.467
then we can use that to to
improve the the drone mapping.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7085-0
00:39:56.557 --> 00:40:02.093
So in that case, maybe the next
time the drone can fly, fly very

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7085-1
00:40:02.093 --> 00:40:06.947
close to the the mapped
fractures and see getting higher

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7085-2
00:40:06.947 --> 00:40:12.227
resolution image and then maybe
we can detect make millimeter

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7085-3
00:40:12.227 --> 00:40:14.697
level of resolution features.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7095-0
00:40:16.507 --> 00:40:19.897
So is the idea of like automated
augmented mapping.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7140-0
00:40:19.907 --> 00:40:24.656
So based on a mapped a mapped
model and then we can have a

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7140-1
00:40:24.656 --> 00:40:29.647
smarter way for the joint to
plan more efficient pass so they

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7140-2
00:40:29.647 --> 00:40:34.235
can focus on those damaged
features and then we will see

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7140-3
00:40:34.235 --> 00:40:36.247
how how we can push that.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7147-0
00:40:38.797 --> 00:40:38.997
Yeah.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7150-0
00:40:39.007 --> 00:40:40.157
We're gonna go to salty.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7165-0
00:40:40.167 --> 00:40:42.897
We'll be there in the weekend
and Monday.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7169-0
00:40:42.907 --> 00:40:47.247
This is first time there so
little bit informative.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7187-0
00:40:49.267 --> 00:40:53.033
So are there any questions out
there in the audience, the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7187-1
00:40:53.033 --> 00:40:53.617
Internet?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7201-0
00:40:53.667 --> 00:40:53.797
What?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7203-0
00:40:54.717 --> 00:40:57.927
Yeah, this is Austin Elliott at
Moffett Field.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7210-0
00:40:58.817 --> 00:41:00.867
Thanks for the really great
talk.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7224-0
00:41:00.877 --> 00:41:03.277
It's pretty exciting to see what
you're doing here.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7234-0
00:41:03.667 --> 00:41:08.287
Umm I I'm curious about the you
showed.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7270-0
00:41:10.007 --> 00:41:13.752
Obviously this sort of like
iterative dynamic surveying

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7270-1
00:41:13.752 --> 00:41:18.033
where the robot the drone goes
out and identifies features that

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7270-2
00:41:18.033 --> 00:41:21.577
are of further interest and then
focuses in on them.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7306-0
00:41:21.947 --> 00:41:26.033
I'm curious what sort of where
that stands currently

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7306-1
00:41:26.033 --> 00:41:30.426
technologically, and what you
imagine to be the how much

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7306-2
00:41:30.426 --> 00:41:35.282
adjustment and further training
and in sort of manual input is

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7306-3
00:41:35.282 --> 00:41:36.207
required to.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7323-0
00:41:37.397 --> 00:41:41.499
Adapt it to the pretty wide
variety of scenarios you'd

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7323-1
00:41:41.499 --> 00:41:42.767
encounter, right?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7346-0
00:41:42.777 --> 00:41:47.065
Different geometries of these
things, different things that

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7346-1
00:41:47.065 --> 00:41:51.567
you different aspects of it that
you'd rather focus on, right?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7358-0
00:41:51.577 --> 00:41:53.507
Maybe you need to get really
right under the rock.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7369-0
00:41:53.517 --> 00:41:55.877
Or maybe you need it from a
different vantage point.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7372-0
00:41:56.957 --> 00:41:57.607
How?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7383-0
00:41:58.017 --> 00:41:59.447
I don't know what do you what do
you see as the?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7404-0
00:42:01.997 --> 00:42:07.267
How much work and how much
variety exists in that sort of

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7404-1
00:42:07.267 --> 00:42:09.357
honing of the dynamics?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7426-0
00:42:10.017 --> 00:42:14.619
Yeah, I mean the the robotic
sending machine learning

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7426-1
00:42:14.619 --> 00:42:19.477
technologies are developing very
rapidly like last year.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7434-0
00:42:19.487 --> 00:42:20.837
We have challenge GPT.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7454-0
00:42:20.847 --> 00:42:25.686
That's almost Lily changed and
personally changed my life, at

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7454-1
00:42:25.686 --> 00:42:27.247
least of working in.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7473-0
00:42:27.397 --> 00:42:30.651
Also, sometimes when I send out
the email, I'll also ask chat,

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7473-1
00:42:30.651 --> 00:42:30.857
GPT.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7503-0
00:42:32.467 --> 00:42:37.748
So the idea is here, so most of
previous computer vision and

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7503-1
00:42:37.748 --> 00:42:42.941
perception models are based on
supervised learning and they

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7503-2
00:42:42.941 --> 00:42:46.317
require large amount of training
data.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7534-0
00:42:46.807 --> 00:42:51.285
But here are new some models
like Facebook where meta they

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7534-1
00:42:51.285 --> 00:42:55.764
come up with a machine learning
model that requires a very

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7534-2
00:42:55.764 --> 00:42:57.737
minimum training data set.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7540-0
00:42:58.107 --> 00:42:58.757
So.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7567-0
00:42:58.967 --> 00:43:04.707
So at that and that means we in
the future, we don't really, we

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7567-1
00:43:04.707 --> 00:43:09.999
don't really need to collect
some some damage to train the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7567-2
00:43:09.999 --> 00:43:10.537
model.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7572-0
00:43:10.547 --> 00:43:11.257
So it can know.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7578-0
00:43:11.267 --> 00:43:12.097
Oh, that's a PBR.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7605-0
00:43:12.107 --> 00:43:15.881
Maybe it can know based on the
the segmented image you know OK

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7605-1
00:43:15.881 --> 00:43:18.277
and also some other geometric
features.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7615-0
00:43:18.287 --> 00:43:19.637
I know it's that's a PBR.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7652-0
00:43:20.757 --> 00:43:25.125
And also in terms of the, that's
the perception and in terms of

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7652-1
00:43:25.125 --> 00:43:29.629
the the path planning and we are
just talking about actually it's

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7652-2
00:43:29.629 --> 00:43:30.857
not a new problem.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7665-0
00:43:30.867 --> 00:43:34.598
So in robotics, people have
studied view planning for a

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7665-1
00:43:34.598 --> 00:43:34.997
while.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7695-0
00:43:35.067 --> 00:43:40.235
So it's like, what are the best
the camera perspective the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7695-1
00:43:40.235 --> 00:43:45.665
robotic can cover can cover the
to can cover the environment,

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7695-2
00:43:45.665 --> 00:43:47.767
cover the mapping space.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7698-0
00:43:47.897 --> 00:43:48.797
So what?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7713-0
00:43:48.847 --> 00:43:53.113
What are the the best
configurations of the person of

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7713-1
00:43:53.113 --> 00:43:55.087
the mapping perspectives?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7760-0
00:43:55.197 --> 00:43:59.465
So, I mean those technology are
already there, but for

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7760-1
00:43:59.465 --> 00:44:04.275
geoscience studies work for,
for, for the survey and we don't

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7760-2
00:44:04.275 --> 00:44:09.008
have those applications and
that's why I'm interested in the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7760-3
00:44:09.008 --> 00:44:10.327
application part.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7773-0
00:44:10.437 --> 00:44:15.042
So we we how can we adapt the
existing technology to those

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7773-1
00:44:15.042 --> 00:44:16.057
applications?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7778-0
00:44:18.277 --> 00:44:19.017
Yeah, really cool.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7782-0
00:44:19.027 --> 00:44:19.357
Thank you.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7783-0
00:44:19.827 --> 00:44:20.067
Yep.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7792-0
00:44:23.567 --> 00:44:24.437
No one else out there.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7794-0
00:44:25.537 --> 00:44:26.057
Oh yeah.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7802-0
00:44:28.057 --> 00:44:29.647
So Steve Delong says.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7809-0
00:44:29.657 --> 00:44:30.927
Thanks for the excellent time.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7830-0
00:44:30.937 --> 00:44:35.090
Most of what you propose is of
limited spatial extent, as I'm

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7830-1
00:44:35.090 --> 00:44:37.367
sure makes sense for
development.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7851-0
00:44:37.477 --> 00:44:40.452
I wonder if the next logical
step is to automate these ideas

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7851-1
00:44:40.452 --> 00:44:42.597
for long endurance, higher
altitude drones?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7868-0
00:44:43.977 --> 00:44:46.517
No, I lost track or other
platforms that can provide

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7868-1
00:44:46.517 --> 00:44:47.907
monitoring during post event.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7878-0
00:44:47.917 --> 00:44:51.351
Hazard cascades, so you know,
like an earthquake causes a

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7878-1
00:44:51.351 --> 00:44:51.647
fire.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7890-0
00:44:53.297 --> 00:44:55.037
So that's probably what it
means.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7900-0
00:44:55.047 --> 00:44:58.287
Or what are your thoughts about
larger or a tornado causes?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7907-0
00:44:58.717 --> 00:44:59.407
I don't know fire.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7912-0
00:44:59.417 --> 00:45:00.307
Could they cause fires?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7936-0
00:45:00.817 --> 00:45:03.667
What are your thoughts about
larger spatial and temporal

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7936-1
00:45:03.667 --> 00:45:06.767
scale adaptive monitoring that
could include real time change

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7936-2
00:45:06.767 --> 00:45:07.267
detection?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7944-0
00:45:07.577 --> 00:45:10.147
Yeah, I mean not just scale
application.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7950-0
00:45:10.157 --> 00:45:11.467
Definitely interesting in that.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/7969-0
00:45:11.637 --> 00:45:15.997
So that's why I'm thinking about
on multiple UV mapping system.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8033-0
00:45:16.007 --> 00:45:21.180
So instead of just one UV, if we
can have a group of UV, how can

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8033-1
00:45:21.180 --> 00:45:26.194
we use them to to to increase
the scale of the mapping and for

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8033-2
00:45:26.194 --> 00:45:31.128
other applications actually this
one because I I talked about

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8033-3
00:45:31.128 --> 00:45:35.982
photos on mapping and fragile
judicial features because they

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8033-4
00:45:35.982 --> 00:45:40.677
are closely related to my the
story of my my dissertation.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8048-0
00:45:40.847 --> 00:45:45.608
But during my PhD I also did
some work of tornado damage

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8048-1
00:45:45.608 --> 00:45:46.527
estimation.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8092-0
00:45:46.537 --> 00:45:51.322
So still using the same 2 drug
detection data processing

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8092-1
00:45:51.322 --> 00:45:56.359
pipeline, but we use it to
detect a tornado damage features

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8092-2
00:45:56.359 --> 00:46:00.808
like tornado damage roofs and
I'm still working with

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8092-3
00:46:00.808 --> 00:46:02.487
researcher and know.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8114-0
00:46:02.587 --> 00:46:07.100
So we are looking at tornado
damage vegetations like fallen

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8114-1
00:46:07.100 --> 00:46:09.807
trees, and they come up with
ideas.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8150-0
00:46:09.917 --> 00:46:13.382
If we can detect the fallen
trees and we can also estimate

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8150-1
00:46:13.382 --> 00:46:16.964
the the fallen tree do action
and then we can build a vector

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8150-2
00:46:16.964 --> 00:46:18.667
because there are many trees.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8188-0
00:46:18.677 --> 00:46:22.670
OK, then we can build a vector
of those fallen trees and that

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8188-1
00:46:22.670 --> 00:46:26.019
can indicate the tornado
direction, and that's also

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8188-2
00:46:26.019 --> 00:46:29.367
important input model for
tornado speed estimation.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8216-0
00:46:29.537 --> 00:46:33.578
So I do see because as I said,
like once the technologies

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8216-1
00:46:33.578 --> 00:46:37.967
developed need has can be easily
transferred to other domains.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8227-0
00:46:38.237 --> 00:46:42.784
I do see there could be many
other applications potentially,

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8227-1
00:46:42.784 --> 00:46:43.157
yeah.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8253-0
00:46:45.357 --> 00:46:49.756
And and and also why do you
mentioned this known animation

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8253-1
00:46:49.756 --> 00:46:52.067
where the the drone capability.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8315-0
00:46:52.137 --> 00:46:56.954
So I'm also there are so most of
the time I use the multi voter

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8315-1
00:46:56.954 --> 00:47:01.695
multi water but there is also a
vital vertical takeoff so that

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8315-2
00:47:01.695 --> 00:47:05.835
it combines multi voter and
fixed wing and then it can

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8315-3
00:47:05.835 --> 00:47:10.426
provide good and it can improve
like largely improve the the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8315-4
00:47:10.426 --> 00:47:12.307
range and flight time of.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8334-0
00:47:12.577 --> 00:47:16.599
So that one is probably more
suitable for large scale and

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8334-1
00:47:16.599 --> 00:47:17.847
long range survey.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8356-0
00:47:17.917 --> 00:47:21.619
But of course the cost is
higher, so I'll see if I can

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8356-1
00:47:21.619 --> 00:47:25.051
find a good funding
opportunities to support those

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8356-2
00:47:25.051 --> 00:47:25.657
research.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8371-0
00:47:28.707 --> 00:47:30.057
But I have a last question.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8383-0
00:47:30.067 --> 00:47:33.437
If there's anyone, or if there's
someone, speak up.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8401-0
00:47:33.447 --> 00:47:36.687
If you're in Moffett. Ohh.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8402-0
00:47:35.237 --> 00:47:38.427
There's one set in China in go
ahead.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8406-0
00:47:38.437 --> 00:47:39.467
Go for my uh.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8418-0
00:47:39.797 --> 00:47:41.047
Went that your buffet.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8426-0
00:47:42.067 --> 00:47:44.227
First of all, this is way
cooler.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8436-0
00:47:44.237 --> 00:47:46.707
I was just blown away by by your
presentation.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8442-0
00:47:46.717 --> 00:47:47.997
I think we're all in shock here.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8475-0
00:47:50.657 --> 00:47:55.596
I think you mentioned in passing
some of one of the limitations

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8475-1
00:47:55.596 --> 00:48:00.689
of the methodology, particularly
for detecting and characterizing

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8475-2
00:48:00.689 --> 00:48:04.547
small scale features is
basically georeferencing.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8499-0
00:48:04.737 --> 00:48:08.421
And you mentioned again in
passing that I guess

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8499-1
00:48:08.421 --> 00:48:13.180
georeferencing it, you do now is
with RTK real time kinematic

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8499-2
00:48:13.180 --> 00:48:13.487
GPS.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8503-0
00:48:13.957 --> 00:48:14.767
Is that correct?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8510-0
00:48:16.307 --> 00:48:19.017
Yeah, RTK is is something.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8536-0
00:48:19.187 --> 00:48:24.495
Yeah, I used and and and we are
also gonna use it for our our

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8536-1
00:48:24.495 --> 00:48:26.207
small scale feature.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8537-0
00:48:27.217 --> 00:48:27.687
OK.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8575-0
00:48:27.737 --> 00:48:31.802
So my follow up question, which
is really follows up on I think

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8552-0
00:48:29.627 --> 00:48:29.847
Umm.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8575-1
00:48:31.802 --> 00:48:35.994
something you mentioned briefly,
if you had more ground GPS sites

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8575-2
00:48:35.994 --> 00:48:39.677
than the GEOREFERENCING would
would be would be approved.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8607-0
00:48:40.637 --> 00:48:45.460
I have you any thoughts about
that and also you know could you

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8607-1
00:48:45.460 --> 00:48:50.205
automate this so that your your
drone parachutes GPS sites to

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8607-2
00:48:50.205 --> 00:48:51.047
the ground.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8610-0
00:48:52.587 --> 00:48:53.437
Space stations.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8612-0
00:48:54.707 --> 00:48:55.407
Ohh.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8625-0
00:48:57.367 --> 00:48:59.347
I'm I didn't really capture your
first.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8631-0
00:48:59.357 --> 00:49:01.997
Your first question, can you
repeat that?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8651-0
00:49:02.957 --> 00:49:07.936
Uh, if you would improve the
georeferencing with a more

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8645-0
00:49:07.147 --> 00:49:07.397
Mm-hmm.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8651-1
00:49:07.936 --> 00:49:09.447
ground GPS sites.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8656-0
00:49:10.337 --> 00:49:11.107
Base stations.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8660-0
00:49:11.357 --> 00:49:11.767
Yeah.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8663-0
00:49:11.527 --> 00:49:12.197
Base station.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8667-0
00:49:11.827 --> 00:49:13.107
Yeah. Mm-hmm.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8674-0
00:49:12.947 --> 00:49:16.517
Uh, and and my my what's the
right word?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8679-0
00:49:18.667 --> 00:49:19.817
Provocative slash.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8686-0
00:49:19.827 --> 00:49:22.787
Funny question was, could you
parachute?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8692-0
00:49:22.987 --> 00:49:23.367
Yeah.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8709-0
00:49:24.037 --> 00:49:26.657
You base stations down from from
the Supergirl.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8702-0
00:49:24.907 --> 00:49:25.727
Uh.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8707-0
00:49:25.737 --> 00:49:26.247
Parachute.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8710-0
00:49:26.427 --> 00:49:27.027
Yes, yes.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8712-0
00:49:27.037 --> 00:49:28.217
Yeah, I see.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8714-0
00:49:28.787 --> 00:49:30.077
Yeah.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8719-0
00:49:30.367 --> 00:49:32.277
Yeah, for RTK GPS.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8734-0
00:49:34.057 --> 00:49:38.367
Conventionally, we need a base
station and a mobile station.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8742-0
00:49:38.377 --> 00:49:40.307
Mobile Base station is like kind
of you.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8766-0
00:49:40.317 --> 00:49:44.889
You put on the rod, you fix it
and you wait maybe 410 to 30

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8766-1
00:49:44.889 --> 00:49:48.547
minutes, and then the GPS
signals can converge.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8791-0
00:49:48.637 --> 00:49:53.130
And then you have a mobile stage
mobile station that you mount on

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8791-1
00:49:53.130 --> 00:49:55.717
the drone and then combine these
two.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8803-0
00:49:55.727 --> 00:50:00.153
You can provide sending me one
centimeter to 2 centimeter

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8803-1
00:50:00.153 --> 00:50:01.297
resolution and.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8847-0
00:50:03.327 --> 00:50:07.625
And but now there just found
like a few days ago, there are

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8847-1
00:50:07.625 --> 00:50:12.066
and then they can new generation
of RTK then do that does not

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8847-2
00:50:12.066 --> 00:50:15.217
require base station with quite
surprising.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8858-0
00:50:15.347 --> 00:50:17.619
I don't know how how they did
that, but there's some

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8858-1
00:50:17.619 --> 00:50:18.347
technology there.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8861-0
00:50:20.277 --> 00:50:21.987
Umm uh.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8869-0
00:50:24.117 --> 00:50:25.397
Yeah, let me think.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8877-0
00:50:26.867 --> 00:50:27.777
Yeah, I don't know.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8874-0
00:50:27.137 --> 00:50:27.487
OK.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8882-0
00:50:27.497 --> 00:50:27.847
Thank you.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8901-0
00:50:27.787 --> 00:50:33.670
Like a large RTQ is, is is one
challenge and also there are

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8901-1
00:50:33.670 --> 00:50:37.787
many other challenges like
communication.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8926-0
00:50:37.797 --> 00:50:42.201
I think that's some more, more
important in this case, like how

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8926-1
00:50:42.201 --> 00:50:45.847
for example like in the field of
work and yesterday.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8993-0
00:50:46.537 --> 00:50:52.262
So we are trying to map a very
the Centennial plov and it's

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8993-1
00:50:52.262 --> 00:50:58.369
very deep and and when the drone
is flying to the other side of

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8993-2
00:50:58.369 --> 00:51:04.285
the slope and the huge huge a
blocks of voters that block the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/8993-3
00:51:04.285 --> 00:51:04.857
radio.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9023-0
00:51:05.037 --> 00:51:08.507
And then when the radio is
blocked and they don't have to

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9023-1
00:51:08.507 --> 00:51:12.037
take off and take off and come
back to the home point, no.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9043-0
00:51:12.047 --> 00:51:15.295
But I want to continue the
mission and finish the mapping

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9043-1
00:51:15.295 --> 00:51:15.687
for me.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9063-0
00:51:15.757 --> 00:51:19.446
But somehow because the radio is
disconnected and then they don't

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9063-1
00:51:19.446 --> 00:51:20.787
have to have to go back.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9088-0
00:51:20.897 --> 00:51:24.825
So I think the communication,
for example, how can we use the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9088-1
00:51:24.825 --> 00:51:28.183
mobile communication
communication radio tower to do

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9088-2
00:51:28.183 --> 00:51:29.767
a relay of communication.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9101-0
00:51:29.777 --> 00:51:33.971
I think those are the more
challenging tasks for this

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9101-1
00:51:33.971 --> 00:51:35.757
technology development.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9106-0
00:51:37.727 --> 00:51:38.107
Thank you.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9108-0
00:51:39.247 --> 00:51:40.157
Yeah, sure.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9115-0
00:51:40.627 --> 00:51:41.657
So are there other?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9118-0
00:51:41.017 --> 00:51:41.347
Out.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9122-0
00:51:41.707 --> 00:51:42.007
Ohh yeah.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9130-0
00:51:42.777 --> 00:51:45.727
Yeah, this is 10 minute, Moffett
Austin talk.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9133-0
00:51:45.737 --> 00:51:46.167
Thank you.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9156-0
00:51:46.177 --> 00:51:49.477
So and so I'll say that we're
looking at urban things in the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9156-1
00:51:49.477 --> 00:51:50.937
Hayward fault this weekend.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9185-0
00:51:51.287 --> 00:51:53.990
And I'm wondering how
technically and maybe like in

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9185-1
00:51:53.990 --> 00:51:57.368
terms of regulation challenging
is to do this kind of mapping in

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9185-2
00:51:57.368 --> 00:51:59.967
urban environments as mostly
desert environments.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9194-0
00:52:01.337 --> 00:52:03.107
That's that's definitely quite a
challenging.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9230-0
00:52:03.157 --> 00:52:07.875
I don't know how I can
personally work with if AI to to

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9230-1
00:52:07.875 --> 00:52:13.098
to work on, to work around those
airspace regulations for for

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9230-2
00:52:13.098 --> 00:52:14.277
some airspace.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9253-0
00:52:14.287 --> 00:52:18.759
No, we cannot definitely fly,
for example, those prohibited,

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9253-1
00:52:18.759 --> 00:52:21.837
uh, spare airspace 4 restricted
airspace.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9281-0
00:52:21.847 --> 00:52:26.659
We can get permissions and also
if if a has a regulation that

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9281-1
00:52:26.659 --> 00:52:30.849
you cannot really operate
multiple Jones and the same

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9281-2
00:52:30.849 --> 00:52:31.237
time.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9317-0
00:52:31.427 --> 00:52:35.451
So we have to, we have to do
something I have because I have

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9317-1
00:52:35.451 --> 00:52:39.673
multiple UAV research is not a
new I, but I don't know how much

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9317-2
00:52:39.673 --> 00:52:42.377
they they have done with the
field work.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9332-0
00:52:42.387 --> 00:52:45.627
So I need to this is 1
professor.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9361-0
00:52:45.637 --> 00:52:50.158
I'm going to have a meeting
later today and he did a lot of

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9361-1
00:52:50.158 --> 00:52:55.132
work and Caltech, about Joan and
also where we are thinking about

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9361-2
00:52:55.132 --> 00:52:56.337
multi UV system.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9368-0
00:52:56.447 --> 00:52:57.667
So of course I will see.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9384-0
00:52:58.207 --> 00:53:02.567
Uh, what do you have done about
the multi UV system?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9401-0
00:53:02.577 --> 00:53:06.055
I mean how they work with if,
how they work with the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9401-1
00:53:06.055 --> 00:53:06.777
regulation.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9419-0
00:53:07.177 --> 00:53:12.167
Uh ohso ration and also how they
deal with these regulations.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9453-0
00:53:12.377 --> 00:53:15.124
I mean, logistics and
regulations are are very

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9453-1
00:53:15.124 --> 00:53:18.805
important and yeah, we have we
think we have to do, we have to

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9453-2
00:53:18.805 --> 00:53:21.727
do something, but it's it's a
very good question.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9476-0
00:53:23.857 --> 00:53:27.787
I and I've got one questions in
action, neither I've offered.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9484-0
00:53:28.057 --> 00:53:31.007
So things first is young for a
really impressive talk.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9530-0
00:53:31.717 --> 00:53:35.265
I was really intrigued by your
ideas about automated

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9530-1
00:53:35.265 --> 00:53:39.483
geoscience, and I I share the
idea that there's this challenge

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9530-2
00:53:39.483 --> 00:53:43.366
and having an excess of data
that might make it even more

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9530-3
00:53:43.366 --> 00:53:47.315
difficult to actually explore
that data to answer pressing

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9530-4
00:53:47.315 --> 00:53:48.587
research questions.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9555-0
00:53:49.077 --> 00:53:52.827
Can you talk about how you would
use machine learning for

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9555-1
00:53:52.827 --> 00:53:56.770
interpreting the massive amounts
of data that you're kind of

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9555-2
00:53:56.770 --> 00:53:57.287
working?

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9575-0
00:53:57.297 --> 00:54:01.569
Collect especially in light of
the biases that machine learning

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9575-1
00:54:01.569 --> 00:54:02.837
methods often have.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9578-0
00:54:04.167 --> 00:54:04.557
Yeah.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-0
00:54:04.567 --> 00:54:09.428
For example you know mass Rover
and there are many interesting

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-1
00:54:09.428 --> 00:54:13.904
rocks on Mars and and they they
are building a a data a a

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-2
00:54:13.904 --> 00:54:18.688
category or no inven inventory
of those rocks on Mars pattern

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-3
00:54:18.688 --> 00:54:23.472
was one when the when the when
the Rover is is already moving

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-4
00:54:23.472 --> 00:54:28.333
already collected those those
blocks but when we see they have

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-5
00:54:28.333 --> 00:54:33.040
to snow to you only need to do
you only need to register new

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-6
00:54:33.040 --> 00:54:37.978
rocks that is not existed that
does not exist in their existing

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-7
00:54:37.978 --> 00:54:42.376
database so they use So what
they do is they have a much

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-8
00:54:42.376 --> 00:54:47.160
Rover and then they will they
will have some more operator to

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-9
00:54:47.160 --> 00:54:51.867
see a lot of missing a lot of
images every day and then they

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-10
00:54:51.867 --> 00:54:56.883
need to say OK this is rocket is
new and I need to put that into

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-11
00:54:56.883 --> 00:55:01.667
my database but the problem is
they have a large database and

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-12
00:55:01.667 --> 00:55:06.605
the operators sometimes if you
confused like is this we already

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-13
00:55:06.605 --> 00:55:11.312
registered So what they do is
they have the machine learning

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-14
00:55:11.312 --> 00:55:16.019
is they have these called a
novel novelty detection so based

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-15
00:55:16.019 --> 00:55:20.571
on their existing database
between machine machine machine

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-16
00:55:20.571 --> 00:55:25.433
learning model and then when we
see a new new rock that is not

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-17
00:55:25.433 --> 00:55:30.062
in the existing database and
then they will be destroyed so

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-18
00:55:30.062 --> 00:55:35.001
in this case we have to you have
to use machine learning and to

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-19
00:55:35.001 --> 00:55:39.862
learn and then we can OK that's
an interesting lesson you rock

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-20
00:55:39.862 --> 00:55:44.183
and also like distributed
acoustic system and they also

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-21
00:55:44.183 --> 00:55:48.890
use machine learning to to pick
to pick the the face I don't

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-22
00:55:48.890 --> 00:55:53.442
know exactly how they did that
but they say I can acoustic

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-23
00:55:53.442 --> 00:55:58.149
signals they are there are is
complicated and in most of the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-24
00:55:58.149 --> 00:56:03.088
time it's not an interesting and
the when the earthquake occurs

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-25
00:56:03.088 --> 00:56:07.949
and the data can be different
but we cannot really monitor the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-26
00:56:07.949 --> 00:56:12.964
data we we all the time so they
have a machine learning model to

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-27
00:56:12.964 --> 00:56:18.057
monitor the data and and know OK
that's a that's a novelty that's

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-28
00:56:18.057 --> 00:56:22.764
a novel thing that we haven't
seen before and then we can we

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-29
00:56:22.764 --> 00:56:27.316
can we can record it or we can
inspect it I mean this this

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-30
00:56:27.316 --> 00:56:32.023
methodology is new and I'm still
thinking about what are the

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-31
00:56:32.023 --> 00:56:36.730
challenges and what are the
opportunities there because that

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-32
00:56:36.730 --> 00:56:41.437
can really change our way of
thinking of the problems we can

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/9999-33
00:56:41.437 --> 00:56:42.517
have maybe if 

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/10005-0
00:56:49.127 --> 00:56:49.867
Awesome.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/10020-0
00:56:50.027 --> 00:56:53.877
I think on that very positive,
inspiring note, we can wrap this

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/10020-1
00:56:53.877 --> 00:56:54.057
up.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/10035-0
00:56:55.067 --> 00:56:58.617
So thanks everybody for joining
us and have a great Wednesday.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/10038-0
00:56:59.367 --> 00:57:00.017
Thank everyone.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/10051-0
00:57:06.057 --> 00:57:06.667
Wow.

3a125a5c-cc4a-477f-8c1c-61d5bdb4bf00/10058-0
00:57:07.017 --> 00:57:08.477
That's like, that's pretty cool,
actually.