I am confused about 'single' vs 'multiple' agent reinforcement learning.

Let's say that I have 1 hunter who I am training to hunt 1 static prey, so that only the hunter is moving around. This is definitely single-agent RL.

If there are two hunters and two preys (moving or static), then this becomes multiple agent RL. It would be 'easier' to study two independent hunters, perhaps, but that is a different problem.

My question is this: If there is 1 hunter only, but there are two preys. There is 'some form of interaction' (vaguely defined) between the hunter and prey. Is this still a single-agent RL? or a multiple one?

Is it possible that my agent is the hunter, and all preys be just part of the environment?


1 Answer 1


A key difference between RL and MARL arises when you consider that other agents are strategic and their behaviour is adaptive. Game theoretical concepts are very important within the scope of MARL and bring in a new range of issues. From a single agent perspective, the environment is non-stationary as every strategic agent is changing their policies to best respond to the each other.

A good discussion on the challenges of MARL can be found here: http://busoniu.net/files/papers/marl_chapter_springer.pdf

Regarding your question: Yes, you can use a single agent RL as long as the preys are not learning. If you plan to apply RL to every agent (hunter and preys), you will find new issues such as convergence problems, etc. There is a lot of research going on around this. To start I would suggest you have a look at some basic and classic concepts such as Fictitious Play and WoLF (Win or Learn Fast).


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.