Using Perturbation to Improve Robustness of Solutions Generated by Genetic Programming for Robot Learning

This paper proposes a method to improve robustness of the robot programs generated by genetic programming. The main idea is to inject perturbation into the simulation during the evolution of the solutions. The resulting robot programs are more robust because they have been evolved to tolerate the changes in their environment. We set out to test this idea using the problem of navigating a mobile robot from a starting point to a target in an unknown cluttered environment. The result of the experiments shows the e ectiveness of this scheme. The analysis of the result shows that the robustness depends on the `experience' that a robot program acquired during evolution. To improve robustness, the size of the set of `experience' should be increased and/or the amount of reusing the `experience' should be increased.