Robot Path Planning in Unstructured Environments Using a Knowledge-based Genetic Algorithm
Abstract
In this paper, a knowledge-based genetic algorithm is proposed for dynamic path planning of a mobile robot in unstructured complex environments. A unique representation method is proposed for the 2-dimensional complex robot environment, where the obstacles could be of arbitrary shape. According to the problem representation, an effective evaluation method is developed specially for the proposed genetic algorithm. The evaluation method is capable of accurately detecting collisions between a robot path and an arbitrarily-shaped obstacle, and assigns a cost function that is effective for the proposed genetic algorithm. The proposed approach uses a problem-specific GA for robot path planning instead of the standard one. It incorporates the domain knowledge of robot path planning into its specialised genetic operators, some of which also involve a local search technique. The effectiveness and efficiency of the proposed genetic algorithm are demonstrated by simulation and comparison studies.
