Developing reactive controllers for mobile robots navigation in unknown environments using infrared range sensors

Autonomous vehicles have undergone tremendous development for an increasing number of applications in recent years. The autonomous navigation in unknown environments remains a challenging task. This thesis focuses on reactive navigation strategies for off road mobile robots based on infrared range sensors. The proposed controllers aimed at detecting the obstacles in the unstructured local environment, and going around or reaching the target along a reasonable path. The local minima in conventional approaches should be avoided. Simplicity in architecture and capability of self-learning are also considered.; A behaviour-based mobile robot was developed for indoor navigation experiments. The hardware and software design is detailed in this thesis. Based on the infrared sensor combination on board, a neuro-fuzzy control system was developed to steer a mobile robot to wander in an enclosed room with unknown obstacles. To investigate the complete navigation strategy, a trajectory-based controller combining the global target location and local sensory information is proposed for a new robot model. The output of the trajectory-based controller is one of the nine preset trajectories other than the robot velocity or acceleration in most existing approaches. By using the proposed controller, the trapping situation like "symmetric indecision" and "dead cycle" can be avoided. The proposed controller can be tuned automatically instead of "trial and error" work. A new co-evolution approach to sensor placement and controller parameter setting is presented to improve the tuning process of the proposed trajectory-based controller. The co-evolution approach considers that both hardware and control software are mutable. The simulation studies indicate that the co-evolution approach can lead to effective robot design.