Control of Cellular Automata by Moving Agents with Reinforcement Learning

TL;DR

This paper explores how reinforcement learning agents can modify cellular automata environments to achieve global goals, highlighting the challenges posed by passive versus active dynamics.

Contribution

It introduces the problem of agents learning to control cellular automata environments and analyzes the impact of environment dynamics on their success.

Findings

Agents can learn to reach goals in passive automata environments.

Active automata environments prevent agents from achieving their goals.

The study highlights the importance of environment dynamics in control tasks.

Abstract

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.