1
00:00:00,760 --> 00:00:03,940
Congratulations on making it through to the end of the section.

2
00:00:04,300 --> 00:00:10,450
In this section, we took a look at ignition gazebo as a possible tool for simulating your robotic systems.

3
00:00:10,720 --> 00:00:14,080
That's because Gazebo has been historically paired with ROS.

4
00:00:14,080 --> 00:00:19,990
And most common when first getting into robotic simulations which use Ros Ignition gazebo is a relatively

5
00:00:19,990 --> 00:00:25,570
new framework, only really being released in 2022 and missing some features which Gazebo Classic had.

6
00:00:25,690 --> 00:00:30,520
Now there are a number of reasons for this, and I only expect ignition gazebo features and documentation

7
00:00:30,520 --> 00:00:32,020
to improve over time.

8
00:00:32,170 --> 00:00:37,660
But keep in mind there are additional simulation tools to be used depending on your use case.

9
00:00:38,170 --> 00:00:42,370
A lot of my students who take this course are interested in robotic arms.

10
00:00:42,670 --> 00:00:48,580
These robots are common around many manufacturing workplaces and all involve similar types of problem

11
00:00:48,580 --> 00:00:52,750
solving, such as path planning, manipulator control and task planning.

12
00:00:53,050 --> 00:00:58,960
This type of specialty robotics can benefit from specialty frameworks such as Move It, a common framework

13
00:00:58,960 --> 00:01:05,050
tool used with ROS to easily combine features such as motion planning, inverse kinematics, 3D perception

14
00:01:05,050 --> 00:01:06,640
and various controllers.

15
00:01:06,730 --> 00:01:11,650
They recently released their ROS to humble release, so feel free to check that out to see what they

16
00:01:11,650 --> 00:01:14,020
have to offer if that seems interesting to you.

17
00:01:14,920 --> 00:01:16,600
Besides specialty frameworks like Move It.

18
00:01:16,600 --> 00:01:19,570
There are complete simulation alternatives out there.

19
00:01:19,870 --> 00:01:24,640
One of which seems to be catching traction in the Ross community is Reebok's.

20
00:01:24,940 --> 00:01:31,390
One of its biggest selling points is being a simplistic simulation tool, not strictly tied to a framework

21
00:01:31,390 --> 00:01:37,360
such as ROS, which can be used for general use simulations while also having a decent amount of example

22
00:01:37,360 --> 00:01:39,310
simulations to leapfrog off of.

23
00:01:39,910 --> 00:01:45,010
We bought does have a Ross driver, so you can both use the simulation to create robotic models and

24
00:01:45,010 --> 00:01:51,220
control them over Ross all without having to mess with any SDF files and seemingly have very good cross

25
00:01:51,220 --> 00:01:52,450
operating system support.

26
00:01:53,280 --> 00:01:58,050
Another popular simulation platform in the field of robotics and academia, as well as research and

27
00:01:58,050 --> 00:02:03,540
development, is Moukoko, which stands for Multi Joint Dynamics with Contact.

28
00:02:03,780 --> 00:02:07,230
This platform was acquired and made open source by DeepMind.

29
00:02:07,380 --> 00:02:13,140
Again, this is a good pick for robotics applications which need accurate multi body contact dynamics.

30
00:02:13,260 --> 00:02:18,300
But hands down one of the major simulations I am eyeing down as the potential winner takes all candidate

31
00:02:18,300 --> 00:02:22,440
for the future of robotics simulations is Nvidia's Isaac SIM.

32
00:02:22,920 --> 00:02:28,410
The future of robotics is the advancements of its software capabilities through AI and machine learning,

33
00:02:28,410 --> 00:02:30,600
and Isaacson was built with that in mind.

34
00:02:30,900 --> 00:02:36,870
Previous GTC conferences hosted by NVIDIA showed a promising advances in truly bridging the gap from

35
00:02:36,870 --> 00:02:43,230
simulation to reality through the work of several research institutions operating on the animal platform.

36
00:02:43,320 --> 00:02:48,660
A quadruped robot similar to Boston Dynamics spot robot but with wheels for feet.

37
00:02:49,200 --> 00:02:54,510
Leveraging their graphics card supremacy, Nvidia was able to create simulation environments capable

38
00:02:54,510 --> 00:03:00,780
of generating realistic synthetic data for robot model training as well as their Isaac Jim framework,

39
00:03:00,780 --> 00:03:07,380
which allows for reinforcement learning training in extremely high capacities within simulation in parallel.

40
00:03:08,100 --> 00:03:13,050
Nvidia omniverse, which is the base platform Isaac runs off of, also has connectors to various popular

41
00:03:13,050 --> 00:03:18,480
CAD softwares, which will really help bridge the gap between updating mechanical designs in a way which

42
00:03:18,480 --> 00:03:22,320
update directly to your simulations and to top the cake.

43
00:03:22,350 --> 00:03:28,140
You can also use this tool leveraging Nvidia's online data centers to quickly train your robot models

44
00:03:28,140 --> 00:03:33,570
in the cloud on super computer level systems, taking training time down significantly.

45
00:03:33,780 --> 00:03:39,300
Now, of course this comes quite the monetary cost, but nonetheless, being able to easily expand compute

46
00:03:39,300 --> 00:03:45,420
capabilities to anyone is a great way to expedite robotic innovations to those trying out new things.

47
00:03:45,780 --> 00:03:50,130
There's so much happening in the world of robotics and there are many tools and simulation frameworks

48
00:03:50,130 --> 00:03:51,270
that I fail to cover.

49
00:03:51,450 --> 00:03:58,320
But hopefully you can see why robotics is growing at a rapid pace and the tools we are using are evolving

50
00:03:58,320 --> 00:03:59,120
every day.

51
00:03:59,130 --> 00:04:04,470
So Cognition Gazebo is a great place to start your journey in general purpose robotics simulation.

52
00:04:04,470 --> 00:04:09,570
But based on whatever your project objectives are, there may be other tools and frameworks out there

53
00:04:09,570 --> 00:04:12,870
for you to use which can expedite your progress.

54
00:04:12,990 --> 00:04:18,450
So hopefully I covered at least enough of them in this video as a reference point to help you understand

55
00:04:18,450 --> 00:04:19,500
what's out there.