I understand that you are experiencing an issue with the reward for the first agent in your multi-agent PPO setup.
Here are a few things you can check to resolve the issue:
- Reward Function: Inspect your environment's step function. Ensure that the reward vector (or structure) includes a non-zero value for the first agent (“rlPPOAgent”).
- Agent Configuration: Make sure “rlPPOAgent” is correctly associated with its environment and policy.
- Environment Setup: You can double-check the environment setup to make sure all agents are interacting with it as intended.
- Training Parameters: Review the training parameters specific to the first agent, like the learning rate and discount factor.
These are some of the ways that might help you to fix the problem. If not, please provide the code that you are working with so that I can take a deeper look.
Feel free to refer this documentation on “Agents”:
