Lamarckian Learning in Multi-Agent Environments.
John J. Grefenstette · 1991
Genetic algorithms gain much of their power from mechanisms derived from the field of population genetics. However, it is possible, and in some cases desirable, to augment the standard mechanisms with additional features not available in biological systems. In this paper, we examine the use of Lamarckian learning operators in the SAMUEL architecture. The use of the operators is illustrated on three tasks in multi-agent environments. 1 INTRODUCTION The goal of this work is to explore the application of machine learning techniques to reactive control problems arising in competitive, multi-agent domains. In such domains, traditional AI planning approaches are usually infeasible, because of the complexity of the multi-agent interactions and the inherent uncertainty about the future actions of other agents. On the other hand, genetic algorithms [11] appear to be a promising approach to developing high performance control strategies. SAMUEL is our platform for exploring the use of genetic...