I do not understand why ISAAC came to be when there is ROS. The time and resources that have been placed into ISAAC could of been put towards ROS development like other tech leaders such as Amazon and Microsoft are doing. Instead we have ISAAC which is a closed and proprietary source that reinvents the numerous wheels provided by ROS and Gazebo.
What am I missing here?
It helps if you can simulate photo-realistic visual inputs using a simulator for training agents, using deep reinforcement learning techniques.
The Gazebo simulator that comes with ROS distro’s isn’t photorealistic enough.
The Quixel ReBirth video is a good example of the level of realism that can be achieved using UE4. [url]Rebirth: Introducing photorealism in UE4 - YouTube
NVIDIA have made advances in using techniques such as Structured Domain Randomization [url]https://research.nvidia.com/publication/2018-10_Structured-Domain-Randomization[/url], to use pure simulator data to train neural networks using photorealistic simulator visual inputs and perform well on standard datasets like KITTI.
If you need photorealistic visual input, then your choices are more or less limited to UnrealEngine 4 or Unity. Others have developed photorealistic simulators around UE4, such as Microsoft with AirSim [url]https://github.com/microsoft/AirSim[/url] and DeepDrive.io [url]https://deepdrive.io/[/url].
While the UE4 license is proprietary, you do have access to the full source code, once you sign up. There is a 5% royalty if you commercialize your product based on the UE4 engine. Epic Games put in a phenomenal amount of support behind the engine, in terms of constant updates to the engine sources, tutorials, roadshows and evangelization.
There is a Isaac SDK ROS Bridge example here:[url]https://docs.nvidia.com/isaac/isaac/apps/samples/navigation_rosbridge/doc/ros.html[/url]
Microsoft AirSim similarily has some ROS integration examples here
[url]https://github.com/microsoft/AirSim/tree/master/ros/src/airsim_tutorial_pkgs[/url]
So, you can still continue to use ROS, but drop using the Gazebo as simulator at some point and use UE4 instead for training and simulation.
Facebook just open sourced that the other day as well for training ai agents
Hi folks, what’s the conclusion then?
Why do we have the Isaac simulator instead of adding Structured Domain Randomization and the missing features to Microsoft’s Airsim?
Hello All
Isaac SDK and Isaac Sim is built to help developers use the capabilities of NVIDIA’s graphics/computing/AI platforms (with RTX/PhysX/CUDA/CuDNN/TensorRT) for robotics application development / work flows (both cloud and edge). It is complementary to all that you have mentioned above (such as UE4, Unity, ROS, Gazebo etc.).
Regarding the question “Why do we have the Isaac simulator instead of adding Structured Domain Randomization and the missing features to Microsoft’s Airsim?”
As mentioned above, Isaac Sim will enable multiple workflows for robot development. We already support ML workflows @ https://docs.nvidia.com/isaac/isaac/packages/ml/ml.html; stay tuned for more updates in this area.
We will be sharing details of many more workflows in the future.
Check some of them out @ https://developer.nvidia.com/isaac-sim
