A hybrid GA-ANN approach for autonomous robots topological navigation

Abstract

This paper proposes a hybrid approach using genetic algorithm and artificial neural networks for autonomous path planning and motion control for mobile robots. A topological navigation approach is adopted, using the environment mapped as a graph. A genetic algorithm is used to generate and evolve a set of feasible actions, aiming to lead the robot to the goal considering the shortest path. Each action is a different reactive behavior designed for a specific environment feature such as corridors, turns or intersections. Then, an artificial neural network is trained to recognize the different environment features, and the next behavior is activated every time the ANN detects a transition. Experiments were performed in Player/Stage robotics simulator and obtained results showed this approach as a promising way to plan and execute a path.