Multiple Ego Vehicles with Multi GPU Setup

[repa1030]

Hi together,

I am currently stucked on a problem. What I wanted to do at the end:

• I want to have multiple Ego-Vehicles in the same simulation environment (so they should be visible for each other)
• For each Ego-Vehicle I have a GPU for Rendering (so I want to use the Multi-GPU Subserver System of CARLA)
• All of that is running on one machine (of course with multiple GPUs)

I am trying to do as follow (here, only with 2 Subservers):
In Docker 1 (main server, no rendering): docker run --net=host carlasim/carla:0.9.15 /bin/bash ./CarlaUE4.sh -nullrhi
In Docker 2 (Subserver 1): docker run --gpus '"device=0"' --net=host carlasim/carla:0.9.15 /bin/bash ./CarlaUE4.sh -carla-rpc-port=3000 -RenderOffScreen
In Docker 3 (Subserver 2): docker run --gpus '"device=1"' -e CUDA_VISIBLE_DEVICES=1 --net=host carlasim/carla:0.9.15 /bin/bash ./CarlaUE4.sh -carla-rpc-port=4000 -RenderOffScreen

and on the Host itself, I am running in two terminals (the manual control from examples):
python manual_control.py -p 3000 and python manual_control.py -p 4000

Here, i get an error of the secondary started python script:
RuntimeError: trying to create rpc server for traffic manager; but the system failed to create because of bind error.

Where did I went wrong in this scenario? I also tried to change the TM port for the different clients, resulting in the same Error but with starting the pygame window before it crashes.

Thanks for your help in advance!

Patrick

[Blyron]

Please fulfill the template to receive feedback