Multiple RL Agent Simulink Model Keeps Compiling Before Training Starts

Hi,

I'm trying to implement a multiple RL agent model similar to the examples shown in the MATLAB documentation links:

• https://uk.mathworks.com/help/reinforcement-learning/ug/train-2-agents-to-collaborate.html
• https://uk.mathworks.com/help/reinforcement-learning/ug/train-agents-for-path-following.html

I'm trying to control a dynamic system with two agents that use two control signals to control the environment (4 signals total). I've implemented an RL algorithm that uses a single agent and outputs 4 control signals simulataneously (using MATLAB 2020b).

The problem I'm running into is when I run the training algorithm to build the agents' actor and critic networks (just like in the above examples) it repeatedly starts compiling and resetting and cannot begin training. I've tried investigating the reset function using some displays to see if it is running and it is repeatedly running the reset function instead of begining an episode for training. I never encountered this issue of repeatedly compiling for the single RL agent Simulink model.

Any help or advice with how to resolve this issue would help me greatly.

Thanks,

Julian

[Shahriar]

Hi! How did you solve this problem? I have a similar situation.