shr1ftyy / atom-issues Goto Github PK

Most of the works in the field are homogeneous MARL, meaning that all the agents in the scenarios explored are identical in terms of ability and skill. However, in real-world applications, we likely face a multi-agent problem where agents have different skills and abilities. Therefore, an additional problem here would be how different agents should utilize the other agents’ abilities to learn a more efficient policy.

Read more papers.

I need to spend some time thinking about how I'm going to spend my time going forwards - starting with the following week or two. (see Discord message for project plan)

• Harry mentioned that the suggestions for future work in Multi-Agent Actor-Critic for... regarding trying to decrease the complexity of the inputs of the $Q$ function would be a good direction to head towards. Should look more into this.