Line patterns formed by cellular automata agents

Abstract

Considered is a 2D cellular automaton with moving agents. Each cell contains a particle with a certain spin/color that can be turned by an agent. Four colors are used. The objective is to align the spins in parallel along horizontal and vertical lines, in order to form long orthogonal "line patterns". The quality of a line pattern is measured by a degree of order computed by counting matching 3 x 3 patterns. Additional markers are used and signals between agents are introduced in order to improve the task efficiency. The agents’ behavior is controlled by a finite state machine (FSM). An agent can perform 128 actions altogether as combinations of moving, turning, color changing, marker setting and signaling. It reacts on its own state and on the sensed colors, markers and signals. For a given set of n x n fields, near optimal FSM were evolved by a genetic algorithm. The evolved agents are capable of forming line patterns with a limited degree of order. The scalability of two FSM against a varying number of agents is studied as well as the efficiency gain through the newly introduced signals.