Most of the papers on multi-agent RL (MARL) that I have encountered have multiple agents who have a common action space.
In my work, my scenario involves $m$ numbers of a particular agent (say type A) and $n$ numbers of another type of agent. Here, the type A agents deal with a similar problem due to which they have the same action space, and type B deal with another type of problem and they have the same action space.
The type A agents are involved in an intermediary task that doesn't reflect in the final reward, and the final reward comes from the actions of type B agents. But the actions of type B are dependent on type A agents.
Any idea on what kind of MARL algorithm is suitable for such a scenario?