Genetic programming for the behavior acquisition of the perception-based robotics

This paper deals with a behavior acquisition of the percept ion-based robotics. A robot that interacts with an unknown environment needs evaluation criteria to acquire its behavioral rules. Evaluation criteria or functions influence the strategies for the behavior acquisition. Furthermore, they determine the searching directions of the behavior acquisition. Behavioral rules can be acquired if the robot has evaluation functions. Consequently, the accumulation of evaluation function is more important than that of behavioral rules. In general, evaluation functions are given by human operators. But the robot should generate its own evaluation criteria suitable to the facing environment, given tasks, and the robot itself. Therefore, we apply genetic programming (GP) for the generation of evaluation criteria. The teaching information or evaluation criteria from human operator can be grounded to the evaluation functions by using the GP, It is shown that GP can generate evaluation functions for the mobile robot through computer simulation.