Inproceedings,

Using non-determinism to improve robustness of robot programs generated by genetic programming

, and .
Proc. of 4th National Computer Science and Engineering Conference (NCSEC 2000), page 81--85. Queen Sirikit National Convention Center, Bangkok, Thailand, (16-17 November 2000)

Abstract

Robustness is essential for programs generated by Genetic Programming (GP). This paper presents a method to improve the robustness. The method employs non-determinism in two ways: one is to evolve robot programs in noisy environments and another is to use probabilistic branch in the function set. The experiment is carried out on robot navigation problems. The result of the experiment shows that the robustness of robot programs has been improved. The analysis shows that the robustness is caused by the acquired experience and the amount of reuse of this experience while performing the task.

Tags

Users

  • @brazovayeye

Comments and Reviews