Performance of Incremental Genetic Programming on Adaptability of Snake-like Robot

Abstract

Abstract In this paper we discuss the performance of the proposed Incremental Genetic Programming on the efficiency of evolution of locomotion gaits of simulated snake-like robot (Snakebot). Compared to the wheeled and legged robots, the Snakebot features better robustness and adaptability characteristics. However, as a modular robot featuring several degrees of freedom, Snakebots is difficult to control, especially when situated in challenging environments. Moreover, evolving the Snakebot in such environments from scratch via canonical genetic programming (GP) is rather inefficient. In our work we propose a two-staged, incremental genetic programming (IncGP) and apply it for the evolution of Snakebot. In order to verify the performance of IncGP, we conducted experiments on evolution of locomotion gaits of Snakebot via both GP and IncGP. The experimental results show that two-staged incremental evolution via IncGP features much improved efficiency over the canonical GP.