النص 🔗
WEBVTT
00:00:02.008 --> 00:00:07.087
Well, hello and welcome to Visions, A Machine Vision and Automation Solutions Podcast.
00:00:07.621 --> 00:00:11.307
I'm your host, Jim Tatum, Senior Editor of Vision Systems Design,
00:00:11.580 --> 00:00:16.607
and Visions is an Endeavor Business Media production from your friends at Vision Systems Design.
00:00:17.205 --> 00:00:21.637
Here you'll find the latest on everything from end-user machine vision solutions
00:00:21.637 --> 00:00:26.708
to trends, developments, and perspectives on all things machine vision and imaging.
00:00:27.573 --> 00:00:30.667
Whether you've been working in the industry for a while or you're just starting
00:00:30.667 --> 00:00:34.867
to take a closer look at it, this podcast is designed to grow your knowledge
00:00:34.867 --> 00:00:38.956
and bring greater focus to your understanding of the imaging and machine vision industry.
00:00:39.630 --> 00:00:41.278
And now, on to our show.
00:00:42.079 --> 00:00:46.626
Well, hi, everyone, and welcome to Visions. You know, in this age of high-tech
00:00:46.626 --> 00:00:52.006
navigation, it seems second nature to just assume something like GPS is going
00:00:52.006 --> 00:00:56.834
to guide us wherever, whenever we need to go, and whatever we need to do.
00:00:57.491 --> 00:01:00.312
And for the most part, for most of us, that's very true.
00:01:01.014 --> 00:01:04.822
But have you ever wondered what would happen if the GPS signal disappeared?
00:01:05.426 --> 00:01:09.436
Because it's a very real issue. Indeed, for military operators,
00:01:09.436 --> 00:01:14.940
first responders, and autonomous systems working in contested or confined environments,
00:01:15.515 --> 00:01:20.745
GPS denial is becoming a real-world challenge across both air and ground domains.
00:01:21.535 --> 00:01:24.780
So today, we are pleased to welcome Dr. Stefan Hrabar,
00:01:25.308 --> 00:01:29.986
co-founder and chief strategy officer for Emesent, an Australian robotics and
00:01:29.986 --> 00:01:34.282
mapping technology company that specializes in autonomous navigation,
00:01:34.526 --> 00:01:36.831
simultaneous localization and mapping,
00:01:37.318 --> 00:01:40.997
LIDAR mapping, and data analytics for challenging environments.
00:01:41.705 --> 00:01:45.536
Welcome, Dr. Hrabar, and thanks very much for joining us on Visions.
00:01:46.000 --> 00:01:50.540
First, if you would, please give us a little background about yourself and what you've been doing.
00:01:51.186 --> 00:01:54.286
Sure. Thanks, James. Great to be here. So, a bit of background about myself.
00:01:54.706 --> 00:01:59.586
I did my undergrad in mechanical engineering in South Africa and then a master's
00:01:59.586 --> 00:02:00.750
in mechanical engineering.
00:02:01.150 --> 00:02:05.506
I started getting interested in the connection between computers and the physical
00:02:05.506 --> 00:02:07.582
world, so I wanted to get into robotics.
00:02:08.047 --> 00:02:11.797
So I moved over to the US
Load full transcript
and did a PhD in robotics at USC.
00:02:12.302 --> 00:02:16.866
And my thesis was vision-based navigation in urban environments for UAVs.
00:02:16.866 --> 00:02:21.178
I started working on drones back in around 2001.
00:02:22.066 --> 00:02:26.256
You've been out there for awhile? Yeah, no, this is sort of in the days before you could
00:02:26.256 --> 00:02:29.066
just go and buy a drone off the shelf. we had to hand-build,
00:02:30.019 --> 00:02:32.364
convert gas-powered helicopters and,
00:02:33.259 --> 00:02:37.453
Yeah, so I started working with putting sort of cameras on drones back then to do,
00:02:37.933 --> 00:02:42.393
stereo vision and optic flow based navigation and then ended up moving to Australia
00:02:42.393 --> 00:02:47.133
to join the CSIRO, which is a national research lab in Australia and continued
00:02:47.133 --> 00:02:48.933
working on drone autonomy,
00:02:49.433 --> 00:02:54.609
eventually adding LiDAR and sometimes radar to drones for doing perception and navigation.
00:02:55.259 --> 00:03:00.663
And then eventually, yeah, just kind of combined the use of LiDAR with SLAM
00:03:00.663 --> 00:03:05.833
on a drone and saw commercial applications that decided to co-found Emesent
00:03:05.833 --> 00:03:08.513
and spin out the technology and, yeah, kind of,
00:03:09.321 --> 00:03:12.393
gone from there eight years ago. Okay, great.
00:03:13.455 --> 00:03:21.483
Well, it looks like what we'll be talking today is working with your work in GPS-denied areas.
00:03:22.104 --> 00:03:27.913
And this, I'm sure, sounds basic, but just to start off, what exactly is GPS
00:03:27.913 --> 00:03:30.757
denial and just how widespread of a problem is it?
00:03:31.035 --> 00:03:35.452
And how far reaching is it? And why is it so important to get around that?
00:03:36.398 --> 00:03:41.725
Yeah, I mean, GPS denied or, I guess, deprived as well as any area where you
00:03:41.725 --> 00:03:43.095
don't get a good GPS signal.
00:03:43.415 --> 00:03:48.602
So the obvious ones are underground in a mine or inside in a building.
00:03:49.118 --> 00:03:53.445
But then even outdoors, if you're underneath, you know, in a forest under canopy,
00:03:53.445 --> 00:03:55.785
you know, trees can impact the GPS signal.
00:03:55.785 --> 00:03:59.305
And in downtown areas, you can get multipaths, so you can get the GPS signal
00:03:59.305 --> 00:04:04.315
kind of bouncing off between buildings. So you don't get a good GPS signal.
00:04:04.815 --> 00:04:09.985
It doesn't resolve well, so you don't get a very accurate position estimation from your GPS.
00:04:09.985 --> 00:04:16.515
And these days it's becoming more and more, I guess, widespread and common in conflict zones as well.
00:04:16.515 --> 00:04:20.405
So it's not just about things that interfere with GPS naturally,
00:04:20.405 --> 00:04:23.885
like trees and mountains, it can be actively jammed or spoofed.
00:04:23.885 --> 00:04:28.635
So that's very common in different conflicts these days to actively interfere with GPS.
00:04:28.955 --> 00:04:34.115
So then again, you can't rely on GPS to figure out where you are and how you
00:04:34.115 --> 00:04:37.875
are moving, because GPS kind of provides both position and velocity.
00:04:38.215 --> 00:04:41.625
And if you don't know how you're moving in terms of velocity,
00:04:41.625 --> 00:04:43.275
you can't control yourself.
00:04:43.615 --> 00:04:46.855
If you're a robot, you can't control your velocity if you don't know how you're
00:04:46.855 --> 00:04:51.945
moving. So yeah, it becomes very important for robotics and autonomy to understand
00:04:51.945 --> 00:04:53.336
where you are and how you're moving.
00:04:54.020 --> 00:04:58.795
Okay. Well, with that in mind, things like machine vision systems,
00:04:58.795 --> 00:05:03.245
for example, have become very good at identifying objects, which I assume extrapolates
00:05:03.245 --> 00:05:05.189
to vision for autonomous vehicles.
00:05:06.087 --> 00:05:11.619
What's required to move from recognizing things to truly understanding the 3D world around them?
00:05:13.001 --> 00:05:16.209
Yeah, I mean, being able to just recognize something in the world.
00:05:16.209 --> 00:05:21.619
And obviously, when I'm talking about perception, what we do and what I'm mostly
00:05:21.619 --> 00:05:25.649
involved in these days is LIDAR for real-time perception and navigation.
00:05:26.131 --> 00:05:31.209
At Emerson, we do capture imagery as well, but it's usually not used in the real-time control.
00:05:31.209 --> 00:05:36.869
We capture imagery, which is then used in the process to add sort of RGB data
00:05:36.869 --> 00:05:41.789
to the LIDAR point cloud or to detect features in it if you're trying to find
00:05:41.789 --> 00:05:42.700
things in the environment.
00:05:43.397 --> 00:05:48.309
But yeah, so to understand the scene, I think, and sort of really have an understanding
00:05:48.309 --> 00:05:52.619
of what's going on around you for an autonomous system, it's obviously semantic.
00:05:52.619 --> 00:05:56.719
So understanding what things are in the environment. Is that a chair?
00:05:56.719 --> 00:05:58.834
Is it a table? Is it a person?
00:05:59.508 --> 00:06:04.739
And then it kind of helps to understand relationships between objects and how
00:06:04.739 --> 00:06:06.999
they typically are related.
00:06:06.999 --> 00:06:11.529
So, you know, usually you would see a chair near a table or something would
00:06:11.529 --> 00:06:16.668
be placed on a table. So those relationships and how they interact becomes important.
00:06:17.192 --> 00:06:20.999
And then the context of what's happening around you helps as well.
00:06:20.999 --> 00:06:26.159
And ultimately, even like the intent, if you detect something in an environment
00:06:26.159 --> 00:06:30.329
like a person, you know, is that person intending to cross the street in front
00:06:30.329 --> 00:06:35.669
of me or are they on the sidewalk so I don't need to worry about them being in my way?
00:06:35.669 --> 00:06:39.949
So, yeah, there's a lot of things that I think go into full understanding rather
00:06:39.949 --> 00:06:42.909
than just being able to identify a person in a scene that's really about,
00:06:43.249 --> 00:06:47.289
you know, what's their intent, what's the context, how they relate to other
00:06:47.289 --> 00:06:49.380
objects around in the environment as well.
00:06:50.178 --> 00:06:56.627
Okay. Well, with, for example, Hovermap, that's more LiDAR than SLAM or it's a combination thereof.
00:06:57.739 --> 00:07:02.631
Yeah, it's SLAM-based LiDAR. So we use the LiDAR data in real time.
00:07:02.631 --> 00:07:07.271
We run the SLAM algorithm on the LiDAR data, and that gives us position and
00:07:07.271 --> 00:07:13.841
orientation without GPS. So we know exactly how our Hovermap device is positioned,
00:07:13.841 --> 00:07:15.929
oriented, and moving through the environment.
00:07:16.637 --> 00:07:20.291
Once you know that, you can take all the LiDAR range measurements that come
00:07:20.291 --> 00:07:25.671
from the LiDAR device, reproject them into the known coordinate or common coordinate
00:07:25.671 --> 00:07:27.858
frame and generate a 3D point cloud.
00:07:28.316 --> 00:07:33.471
So that's when you have your perception. You've got your 3D point cloud being
00:07:33.471 --> 00:07:36.391
created in real time. You know your position and orientation.
00:07:36.908 --> 00:07:43.038
And then, yeah, you can start detecting features in the point cloud to understand the scene.
00:07:43.835 --> 00:07:48.955
That's not something we do yet. We're not doing semantic labeling in real time of the scene.
00:07:49.680 --> 00:07:55.729
For our kind of navigation, we simply need to know is something an obstacle or not.
00:07:56.002 --> 00:07:59.611
Can I move into the space or do I need to go around it? That's kind of the level
00:07:59.611 --> 00:08:04.421
of understanding our autonomy requires to operate in the environments we're
00:08:04.421 --> 00:08:07.331
operating today. Okay. Okay.
00:08:08.003 --> 00:08:12.361
Well, then what are some common environmental challenges you're running into with this?
00:08:12.361 --> 00:08:17.891
So for SLAM in particular, I mean, the way SLAM works is by recognizing features
00:08:17.891 --> 00:08:24.111
in the environment again and again over time. So imagine you see a wall in front
00:08:24.111 --> 00:08:25.658
of you three meters away.
00:08:26.152 --> 00:08:29.271
A second later, you recognize the same wall, and it's two meters away.
00:08:29.271 --> 00:08:33.311
So you know you've moved one meter in that direction. So what it's doing is,
00:08:33.977 --> 00:08:40.001
identifying thousands of recognizable 3D patches or features in the environment
00:08:40.001 --> 00:08:42.150
and tracking those over time.
00:08:42.771 --> 00:08:47.241
So if there's an environment where everything looks the same or there's nothing
00:08:47.241 --> 00:08:50.317
to track, then SLAM is going to have challenges.
00:08:50.759 --> 00:08:52.332
So if you're, for example,
00:08:53.034 --> 00:08:58.319
perfectly smooth steel or concrete lined pipe everything looks the same.
00:08:59.119 --> 00:09:04.019
Around you in that direction so very hard to estimate how far down the the pipe
00:09:04.019 --> 00:09:08.929
you've moved or if you're flying over very flat terrain like a i guess a,
00:09:09.679 --> 00:09:13.929
runway or some big large parking structure where there's no features everything
00:09:13.929 --> 00:09:18.409
looks the same then you won't be able to estimate your sort of your your xy
00:09:18.409 --> 00:09:19.900
orientation or position.
00:09:20.470 --> 00:09:23.999
Or if you don't see anything at all, if you're flying too high,
00:09:24.659 --> 00:09:28.959
on a drone and all the terrain is beyond the range of the LIDAR and you don't
00:09:28.959 --> 00:09:33.579
see anything, then you've got no way of estimating your motion or if you've
00:09:33.579 --> 00:09:35.489
got something that's blocking the LIDAR signal.
00:09:35.489 --> 00:09:43.169
So if you fly into, I guess, very thick dust or smoke or very thick or dense
00:09:43.169 --> 00:09:45.120
water vapors, of water absorbs LIDAR.
00:09:45.648 --> 00:09:50.679
So, if you fly into a very, I guess, misty or cloudy environment where the LIDAR
00:09:50.679 --> 00:09:54.519
data is just being absorbed, then again, you've got another way of perceiving
00:09:54.919 --> 00:09:57.012
the world around you and running the SLAM algorithm.
00:09:57.558 --> 00:10:03.670
That was my next question. How do you get around those challenges like that? Yeah.
00:10:04.579 --> 00:10:07.652
I mean, that can be built into the autonomy. So, our autonomy.
00:10:08.337 --> 00:10:11.642
Adapts to some of those conditions. For example, if we're flying in an underground
00:10:11.642 --> 00:10:16.679
mine, you do get conditions which create kind of an inversion layer or basically a cloud underground.
00:10:17.190 --> 00:10:20.992
As you fly into a void, there can be a lot of moisture that's trapped underneath
00:10:20.992 --> 00:10:23.782
it gets hot and humid so that condenses.
00:10:23.782 --> 00:10:28.572
Imagine the drone's flying into a void and starts flying up into basically a cloud.
00:10:28.572 --> 00:10:34.912
So we can sense whether the LIDAR data is returning and if we start seeing those
00:10:34.912 --> 00:10:38.212
points drop off and we're not getting many returns, we can detect that this
00:10:38.212 --> 00:10:40.742
is in a situation like I described.
00:10:40.742 --> 00:10:45.332
So basically avoid that area, port the mission and return to home before you
00:10:45.332 --> 00:10:47.664
get into a situation where you're losing all the points.
00:10:48.111 --> 00:10:55.632
Similar for dust, we've got fail-safes designed to handle very sort of thick dust environments.
00:10:55.632 --> 00:10:59.652
If you've flown down a mine tunnel, kicking up dust from,
00:11:00.332 --> 00:11:04.062
the drone's
props and then need a return to home and suddenly everything's very
00:11:04.062 --> 00:11:09.212
dusty and you get blinded, we eventually go into a mode which basically backtracks
00:11:09.212 --> 00:11:13.402
the way we've come, assuming that there aren't any obstacles because we didn't encounter.
00:11:14.012 --> 00:11:18.392
As long as we can retrace our path, if there wasn't an obstacle on the way in,
00:11:18.392 --> 00:11:22.252
we assume there's not an obstacle on the way out, so we basically fly blind
00:11:22.252 --> 00:11:26.194
to get out of that dusty area and back into the clear space.
00:11:26.824 --> 00:11:33.872
Okay, interesting. Sounds like it kind of begets more potential challenges to
00:11:33.872 --> 00:11:37.194
solve, but that's very interesting how that works.
00:11:38.166 --> 00:11:43.079
What was the most surprising thing you came across when it comes to moving SLAM
00:11:43.079 --> 00:11:47.104
technology from research prototypes into commercial challenges?
00:11:47.572 --> 00:11:50.659
I think, yeah, one of the things is just the variety of applications or potential
00:11:50.659 --> 00:11:55.309
applications there was with SLAM. I mean, we came to SLAM from a robotics perspective.
00:11:55.309 --> 00:11:57.867
You know, it was developed for autonomous capabilities.
00:11:58.268 --> 00:12:01.169
If you're sending a robot into an environment it's never seen before,
00:12:01.509 --> 00:12:04.589
it needs to build a map. I mean, it needs to know where it is in that map.
00:12:05.029 --> 00:12:07.474
And it's kind of a chicken and egg problem. If it doesn't have a map,
00:12:07.892 --> 00:12:11.699
it doesn't know where it is and vice versa. So that's why SLAM was developed,
00:12:11.699 --> 00:12:13.581
is to solve that robotics problem.
00:12:14.127 --> 00:12:18.069
But then getting into other commercial applications, not just on robots,
00:12:18.069 --> 00:12:19.829
but just for people to map environments.
00:12:19.829 --> 00:12:23.899
Yeah, the variety of applications is immense. I mean, there's so many environments
00:12:23.899 --> 00:12:28.709
where you might not have GPS or good GPS signal, but still want to map an area
00:12:29.049 --> 00:12:30.729
or operate autonomously.
00:12:31.327 --> 00:12:35.968
So, yeah, I think that was one of the pleasant surprises. And then on the flip side.
00:12:36.989 --> 00:12:42.859
If you're using it as a surveying tool, surveyors are used to existing techniques
00:12:42.859 --> 00:12:46.429
and technologies that they've known and trusted and they can understand.
00:12:46.429 --> 00:12:49.749
I mean, if you're using a laser measurement tool or a tape measure or something
00:12:49.749 --> 00:12:52.709
that you really understand, and then somebody comes along with a SLAM-based
00:12:52.709 --> 00:12:56.819
system, which kind of looks like black magic, and it's not always easy to understand.
00:12:56.819 --> 00:13:01.406
So yeah, definitely some skepticism in the early days, but that's changed dramatically
00:13:01.772 --> 00:13:05.951
over the last number of years as SLAM has become pretty much mainstream and,
00:13:06.445 --> 00:13:09.202
surveyors are relying on SLAM-based systems every day now.
00:13:10.009 --> 00:13:14.399
It's now something that is understood. And obviously, the quality of SLAM-based
00:13:14.399 --> 00:13:18.399
solutions has improved. So, yeah, but early on, yes, people didn't know what
00:13:18.399 --> 00:13:22.008
SLAM was. So, you know, a lot of education from our side.
00:13:22.642 --> 00:13:26.789
Surveyors have certainly come a long way from laying chains and shooting lines, I guess.
00:13:27.789 --> 00:13:33.449
Yeah. Wow, interesting. How much of modern autonomy consists of perception challenges
00:13:33.449 --> 00:13:35.732
versus planning and control challenges?
00:13:36.202 --> 00:13:39.291
You were just talking about flying backwards. Yeah. That kind of thing.
00:13:40.591 --> 00:13:43.659
I think, I mean, there's still definitely our challenges in both areas.
00:13:44.785 --> 00:13:47.921
My background is more on the perception side of things, so I still think there
00:13:47.921 --> 00:13:50.027
are more challenges to overcome then.
00:13:50.851 --> 00:13:54.211
I'm not a controls expert, but if I see what robotic systems are doing these
00:13:54.211 --> 00:13:58.502
days, you have humanoids doing backflips and gymnastics, and so it seems like,
00:13:59.176 --> 00:14:03.601
control is obviously very well advanced, and so is perception.
00:14:03.601 --> 00:14:08.031
But I mean, there still will be environments where it's difficult to perceive.
00:14:08.031 --> 00:14:10.644
I mean, I mentioned some SLAM challenges.
00:14:11.108 --> 00:14:15.657
If you're going into very dark or dusty environments, there's not always one,
00:14:16.290 --> 00:14:20.671
way of perceiving or a sensor that's going to solve all challenges.
00:14:20.671 --> 00:14:25.091
So SLAMM is great because it's an active sensor, or LIDAR is great because it's
00:14:25.091 --> 00:14:31.135
an active sensor, and it illuminates the scene versus cameras which require external lighting.
00:14:31.582 --> 00:14:35.281
But then both of those can be disrupted by thick smoke.
00:14:35.281 --> 00:14:38.142
So then you might want to use radar, but radar has its own challenges. So,
00:14:39.081 --> 00:14:43.291
I think, yeah, from my understanding, the perception side of things still has
00:14:43.631 --> 00:14:46.821
challenges, and if you can't perceive the world, then it doesn't matter how
00:14:46.821 --> 00:14:50.631
good your control is, you know, it doesn't help having great control if you
00:14:50.631 --> 00:14:52.728
don't know what's around you. Sure.
00:14:54.036 --> 00:14:59.170
Okay. If you were designing a perception stack for an autonomous robot from
00:14:59.170 --> 00:15:02.980
scratch today, what would you think it would look like?
00:15:02.980 --> 00:15:06.050
Well, yeah, I guess I alluded to some of the points
before.
00:15:06.050 --> 00:15:11.120
There's not one sensor that is suitable for every kind of environment.
00:15:11.120 --> 00:15:17.300
So definitely some multimodal sensing capability, having a combination of cameras,
00:15:17.300 --> 00:15:19.710
LiDAR, radar, acoustics.
00:15:19.710 --> 00:15:25.080
I mean, obviously, it really depends on whether you can afford to carry those
00:15:25.080 --> 00:15:28.270
sensors weight-wise and compute-wise. If you're operating on a drone,
00:15:28.270 --> 00:15:32.085
then obviously, you know, weight is very important, or so is power consumption.
00:15:32.694 --> 00:15:36.320
But if you're on an autonomous vehicle where those are not such an issue,
00:15:36.320 --> 00:15:40.480
then definitely makes sense to use a combination of different sensors.
00:15:40.480 --> 00:15:45.209
So I think designing that perception stack, you'd first look at the intended
00:15:45.435 --> 00:15:46.996
platform and use case and then,
00:15:47.548 --> 00:15:52.150
where possible, try to use as many different types of sensors as possible to
00:15:52.150 --> 00:15:56.789
get redundancy and overcome some of the challenges that one particular sensor would have by itself.
00:15:56.987 --> 00:16:03.520
Right. Well, as LiDAR-equipped robots, such as Hovermap, move beyond mapping
00:16:03.520 --> 00:16:08.196
into autonomous inspection and decision-making, where do you see the boundary between,
00:16:08.835 --> 00:16:12.736
the imaging, the robotics, and spatial AI disappear.
00:16:13.497 --> 00:16:19.264
Or is there? Disappear might not be the right word. Maybe meld? meld and come together.
00:16:20.216 --> 00:16:23.740
Yeah it's interesting question i mean because sometimes you'd have a sensor
00:16:23.740 --> 00:16:28.198
or capability that's just being used for the navigation or autonomy,
00:16:28.663 --> 00:16:32.372
and sometimes you'd have a sensor that's just been used for the inspection task
00:16:32.700 --> 00:16:38.030
i mean often on a if you're trying to inspect a bridge with a drone then,
00:16:38.951 --> 00:16:44.830
you know a small lightweight vision autonomy collision avoidance sensor might
00:16:44.830 --> 00:16:49.090
not give you the resolution that you need to detect very small cracks or corrosion
00:16:49.090 --> 00:16:53.152
so you can't just use one camera system for both applications.
00:16:54.225 --> 00:16:59.262
Likewise with LiDAR, sometimes LiDAR, something like Hovermap is great for providing
00:16:59.262 --> 00:17:03.862
the autonomy and it gives you a very detailed 3D point cloud,
00:17:03.862 --> 00:17:08.124
but you might not get the resolution or precision in the point cloud to find,
00:17:09.212 --> 00:17:11.304
small cracks or corrosion.
00:17:12.349 --> 00:17:17.942
So, where there are applications where the same sensor can be used for both
00:17:17.942 --> 00:17:22.922
autonomy and the mapping or the inspection task, obviously, it's a great outcome
00:17:22.922 --> 00:17:25.501
because then you
don't have to double up on sensors.
00:17:25.861 --> 00:17:29.952
So, yeah, we have many customers that, you know, where Hovermap makes sense
00:17:29.952 --> 00:17:36.392
because the data that's generated is good enough for the inspection or measurement
00:17:36.392 --> 00:17:40.554
side of things, and it's providing the autonomy, you don't have to double up on sensors.
00:17:41.716 --> 00:17:49.042
So, yeah, again, if you're designing a solution that can have one sensor that
00:17:49.286 --> 00:17:53.547
gives you everything you need for autonomy, for scene understanding and perception,
00:17:53.680 --> 00:17:54.989
that's the best outcome.
00:17:54.989 --> 00:17:57.869
What's the next step for something like Hovermap, for example?
00:17:57.869 --> 00:18:00.071
What's the next great thing you want to accomplish with it?
00:18:00.506 --> 00:18:05.899
Yeah, I mean, we're always improving our capabilities from the autonomy side.
00:18:05.899 --> 00:18:10.949
So being able to handle more challenging types of environments and situations,
00:18:10.949 --> 00:18:16.065
getting into smaller spaces or handling some of the environments that I mentioned before.
00:18:17.035 --> 00:18:21.449
And then, yeah, I mean, it would definitely help to be able to do more real-time
00:18:21.449 --> 00:18:25.356
scene understanding and semantics coupled with the autonomy.
00:18:25.607 --> 00:18:28.779
For example, I mentioned, you know, when you're using SLAM, you need to be able
00:18:28.779 --> 00:18:31.187
to see the environment around you as you're navigating.
00:18:31.727 --> 00:18:36.599
If you fly over a large lake or over the ocean, you're not going to get a result.
00:18:36.639 --> 00:18:40.279
So it would make sense if you're flying autonomously from A to B and you come
00:18:40.279 --> 00:18:45.349
across a big body of water to then hug the shoreline so that you're not flying out over the middle.
00:18:45.349 --> 00:18:49.909
So those kind of real-time decisions in the autonomy to make sure that the SLAM
00:18:49.909 --> 00:18:56.009
solution remains stable and you still achieve the goal of things we build into our autonomy stack,
00:18:56.514 --> 00:19:01.999
as well as integrating additional kinds of sensing and sensing capability either
00:19:01.999 --> 00:19:05.941
for the real-time perception or for post-processing overlay.
00:19:06.173 --> 00:19:10.509
Sense of fusion, if you care about measuring gas levels in an environment or
00:19:10.509 --> 00:19:15.839
radiation levels, we can bring in data from gas or radiation sensors,
00:19:16.784 --> 00:19:21.579
display that data real-time overlaid on our point cloud, but now we can start
00:19:21.579 --> 00:19:26.013
doing autonomies based on that data. So, if you're trying to find the source of.
00:19:28.143 --> 00:19:33.859
A radiation source, then you'd be able to react in real time and create,
00:19:34.279 --> 00:19:38.993
I guess, a digital radiation sniffer, which goes and locates the strongest source.
00:19:40.569 --> 00:19:45.109
For us, there's always infinite possibilities that we can do in terms of improvements
00:19:45.429 --> 00:19:51.269
and new use cases, but we generally listen to our customers, understand,
00:19:52.029 --> 00:19:55.759
what their needs are, and then try to work with them to solve problems.
00:19:55.759 --> 00:19:58.739
We're never going to run out of problems to solve, but that's the great thing
00:19:58.739 --> 00:20:00.369
about working in this space.
00:20:01.469 --> 00:20:02.717
Something new every day, yeah.
00:20:03.977 --> 00:20:08.049
Well, that's a wrap for this episode of Visions, produced by Endeavor Business
00:20:08.049 --> 00:20:10.740
Media, a division of Endeavor B2B.
00:20:11.082 --> 00:20:14.079
Thanks very much for tuning in. If you enjoyed today's show,
00:20:14.079 --> 00:20:17.879
be sure to subscribe to the podcast and share this episode with a colleague
00:20:17.879 --> 00:20:18.838
who would find it helpful.
00:20:19.482 --> 00:20:25.369
Until our next episode, you can find us at vision-systems.com or on LinkedIn,
00:20:25.369 --> 00:20:30.024
Facebook, or X for more insights, updates, and breaking news to keep you in the know.
00:20:30.790 --> 00:20:34.488
Thanks for tuning in. Until next time, stay focused on your visions.