Genetic decision mechanism for reasoning and behaviour generation in adaptive intelligent agents

Abstract

The intelligent adaptive agent designed in this paper focuses on finding the best configuration of a robot that needs to make a decision in order to fulfil a task, using a genetic algorithm for reasoning and behaviour generation. This intelligent control system can run locally, on the robot's platform, or globally, if the agent is a supervisor that assigns tasks to robots under its control. The robot's configuration contains parameters that define its internal model, first in the decision making process, and then to construct its behaviour. The novelty of this paper consist in the heterogeneous encoding including fuzzy model inside the chromosome and specialized mechanisms for selection, mutation and crossover. An implementation of the algorithm is provided for download.