Developing control table for multiple agents using GA-Based Q-learning with neighboring crossover

Abstract

In this paper, we show the effectiveness of a GA-based Q-learning method to develop a control table for multiple agents. As a GA-based Q-learning method, we employ a method called “Q-learning with Dynamic Structuring of Exploration Space Based on Genetic Algorithm (QDSEGA)”. In QDSEGA, Q-table for Q-learning is dynamically restructured by a genetic algorithm. QDSEGA combines Q-learning and genetic algorithm effectively, however, it has just employed simple genetic operations in their QDSEGA. We have proposed a crossover for QDSEGA to accelerate the convergence speed to develop a control table for multi-legged robot. In this paper, we show the effectiveness of the proposed neighboring crossover to develop a compact control table for multiple agents.