People and crowd simulation

[ShauryaJain03]

System specifications - Isaac Sim 5.1.0 in docker container, Ubuntu 22.04 on GCP VM

Hi, I aim to simulate a hospital environment with people and crowds in order to test social navigation for my robot. For this purpose, I identified that the hospital asset in Isaac Sim is wonderful for my use case and I can directly start using it, however I am stuck with the simulation of people and crowds. I looked at the documentation here and tried to understand it, however I felt it is too congested and couldn’t really narrow down on one method to script people walking behaviours and crowding.

I would appreciate if someone could provide me complete workflow examples which I can take a look at and understand how to get started with it, or just help me wrap my head around the documentation given at the link I shared earlier.

Many thanks.

[lhuang-svg]

Will this video help https://www.youtube.com/watch?v=STRu2IfKAfE? Or this chapter https://docs.isaacsim.omniverse.nvidia.com/5.1.0/action_and_event_data_generation/ext_replicator-agent/actor_control.html#actor-behaviors?

[ShauryaJain03]

Hey @lhuang-svg, the video you shared is using omni.anim.people and I've been seeing warnings all over the forum and in the docs as well that it is getting deprecated and replaced with a newer extension. So should I continue with it or use IRA instead?

[ShauryaJain03]

Also, the documentation is still puzzling to me and I am not able to narrow the implementation down to one method clearly, could you provide a more detailed workflow for IRA that I could use to simulate people walking, queuing within the hospital ?
I would really appreciate some help over this.

[lhuang-svg]

For how to use IRA, will below links help?
https://docs.isaacsim.omniverse.nvidia.com/5.1.0/action_and_event_data_generation/tutorial_replicator_agent.html#configuration-file
https://docs.isaacsim.omniverse.nvidia.com/5.1.0/action_and_event_data_generation/tutorial_replicator_agent.html#actor-response-properties
https://docs.isaacsim.omniverse.nvidia.com/5.1.0/action_and_event_data_generation/ext_replicator-agent/actor_control.html#ira-response-trigger

[ShauryaJain03]

@lhuang-svg I've been through those links and got an idea of how IRA works, but I am not able to understand/grasp the end to end workflow for how I will use these to setup people simulation in my environment and test my robot's software.

Could you please provide a workflow example of how I can load all the characters into my environment along with my robot and once I press play, the characters start executing their scripted behaviours ?
Is there a python file I can use for that purpose ?

[ShauryaJain03]

Hey @lhuang-svg, any updates?

[lhuang-svg]

https://docs.isaacsim.omniverse.nvidia.com/5.1.0/action_and_event_data_generation/tutorial_replicator_agent.html#getting-started is an example for the E2E workflow. And for your use case, you just need to swap out the base environment USD to the hospital scene you want

[lhuang-svg]

Closing for now. Please re-open / open a new thread if you still need any support.