Control of Cellular Automata by Moving Agents with Reinforcement Learning
Exploratory study for multi-agent reinforcement learning in discrete dynamical systems, but limited by the impossibility result for active environments.
This paper introduces a problem where cognitive agents learn to modify a cellular automaton environment via local sensing to achieve a global goal. Results show agents can approximate the goal with a passive environment but fail when the environment follows active dynamics.
In this exploratory paper we introduce the problem of cognitive agents that learn how to modify their environment according to local sensing to reach a global goal. We concentrate on discrete dynamics (cellular automata) on a two-dimensional system. We show that agents may learn how to approximate their goal when the environment is passive, while this task becomes impossible if the environment follows an active dynamics.