I am designing a simple multi-agent system in which the agents' action space is represented as a Markov chain, defining how likely an agent is to perform an action and to switch between those actions.

My concern is that agents would switch between actions/tasks too frequently instead of sticking to a selected action for some time.

I guess this is a common problem in multi-agent systems. Could someone please suggest where I could learn more about this and how to resolve it?

thx, Manuel

  • $\begingroup$ The fear before a too frequent changes of actions make sense, because if the agent takes the wrong oscillating frequency, the overall statespace will explode which makes the policy unlearnable. The problem to determine the right actions at the right time has to do with reducing the state space. This can be realized with a more elaborated model. In most cases the answer to the problem is, to extend! the number of possible actions and create a high-abstraction layer in the action model. $\endgroup$ – Manuel Rodriguez Apr 18 '19 at 19:33

Your Answer

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

Browse other questions tagged or ask your own question.