Evolutionary Architecture for Lifelong Learning and Real‐Time Operation in Autonomous Robots

Abstract

This chapter deals with evolution in the creation of a cognitive architecture for robots to be able to learn and adapt throughout their lifetime. The emphasis of the architecture is placed on modularity, progressive construction, and real-time adaptation to changes in the environment, in the robot itself or in its motivations. With this combination, an evolutionary architecture is proposed for autonomous robots to create complex behavior controllers that provide quick real-time response and that can be adjusted or even developed during the robot's lifetime. Several examples of the use of these types of architectures and their elements are presented and discussed. The multilevel Darwinist brain (MDB) is a general cognitive architecture that permits automatic acquisition of knowledge in a real agent through interaction with its environment, so that it can autonomously adapt its behavior to achieve its design objectives. Controlled Vocabulary Terms evolutionary computation; intelligent robots; learning (artificial intelligence)