ROBIL: Robot Path Planning Based on PBIL Algorithm

Abstract

Genetic algorithm (GAs) have attracted considerable interest for their usefulness in solving complex robot path planning problems. Specifically, researchers have combined conventional GAs with problem-specific operators and initialization techniques to find the shortest paths in a variety of robotic environments. Unfortunately, these approaches have exhibited inherently unstable performance, and they have tended to make other aspects of the problem-solving process (e.g., adjusting parameter sensitivities and creating high-quality initial populations) unmanageable. As an alternative to conventional GAs, we propose a new population-based incremental learning (PBIL) algorithm for robot path planning, a probabilistic model of nodes, and an edge bank for generating promising paths. Experimental results demonstrate the computational superiority of the proposed method over conventional GA approaches.