Animats adaptation to complex environments as learning guided by evolution

Abstract

In this paper, a new approach to adapt animats to complex environments is proposed. It is based on the advantages and drawbacks of two known strategies: learning and evolution. The proposed approach uses a new learning by reinforcement mechanism guided by an innate and general knowledge -obtained by an evolutionary mechanism- that triggers off a developmental process of learning general behavior. This developmental process increases the animats set of behaviors, characterized by sequences of actions, and facilitates the solving of more complex tasks. This approach is studied, described and finally illustrated with a set of experiments in a complex environment.