May i ask has anyone know how to save the final agents for multi-anegt reforcement learning ?

For a multi-agent RL with three agents , after the agents are well trained, i use the follow command

trainingOpts.SaveAgentDirectory = pwd + "\Agents";
save(trainingOpts.SaveAgentDirectory + "/finalAgent1.0.mat",'agent')

to save the trained agents, but the commond window shows

Variable 'agent' not found

I have used this command to save the single agent, and it worked.

Has anyone know how to save the final agents for multi-anegt reforcement learning ?

thank you

Answer 1

[Yash Sharma]

Hi 张 冠宇 ,

I understand that you have multi-agent reinforcement learning and want to save multiple agents, here is how you can do the same.

• Create a cell array or structure to store the agents. Each element of the cell array or structure represents an agent.
• Assign the trained agents to the corresponding elements of the cell array or structure.
• Use the save command to save the entire cell array or structure containing the agents.

Here's an example of how you can modify your code to save multiple agents:

% crate cell array to store the agents  
agents = cell(1, 3); 
agents(1) = agentA; 
agents(2) = agentB; 
agents(3) = agentC; 
%save agents 
save("Agents/finalAgents.mat", 'agents'); 

Make sure the directory "Agents" exists before saving the agents, or adjust the path accordingly to a valid directory where you want to save the agents.

Please find links to below documentation which I believe will help you for further reference.

cell: https://www.mathworks.com/help/matlab/ref/cell.html