A Reactive Rule-based System for Trajectory Planning of a Mobile Robot with Few Sensory Resources

This work reports on the application of artificial intelligence and neural networks techniques to the problem of trajectory planning and control of a mobile robot with low sensory power. The aim is to determine a spatial sequence through which the robot must navigate forming a trajectory between the starting and goal points. In the first test, the environment has no obstacles and a search tree is generated as the robot begins to navigate. At each level, the network selects a point that minimizes a given heuristic function. Later, obstacles are randomly inserted in the robot workspace. Despite the presence of obstacles the robot is still able to execute a planned trajectory. A successful trajectory is stored for posterior retrieval. The path planning and decision-taking are implemented online. The relation of the method implemented with traditional search methods is discussed.