GA-based path planning for mobile robots: An empirical evaluation of seven techniques.

Previous research suggests that genetic algorithms (GAs) offer a promising solution to path planning for mobile robots. We examine six simple GAs used in prior studies, comparing them to a new node sequence approach that includes a two-step fitness function. Through a series of repeated trials using a simple 16x16 grid, a 100x100 grid, a 600x600 Mars landscape, and a complex maze-like environment, we compare the chromosome structures and fitness functions of these seven methods. The results of our empirical testing indicate that the proposed dual goal approach, which uses a fixed length chromosome structure, outperformed both monotonic and other node sequence approaches, consistently finding a feasible path in even the most challenging of these environments.