Mobile robot obstacle avoidance through behavior fusion using fuzzy logic methods

“Obstacle avoidance for mobile robots is an area of great interest in mobile robotics. Accomplishing tasks while not colliding with the environment is a requirement if robots are to be used in varied environments. Avoidance can be approached as a global optimization problem. In this setting, the environment is fully mapped and path trajectories are usually planned off-line before starting a task. Avoidance can be approached as a local optimization problem as well. For local avoidance, the environment is not fully modeled. If the robot encounters an obstacle, it must avoid that obstacle. Since the final path of the robot cannot be realized until the robot encounters the obstacle, local avoidance is usually an on- line procedure performed in parallel with other procedures. This thesis deals with mobile robot motion in an unknown environment, necessitating the use of local avoidance methods. A control system is designed to avoid static obstacles while attempting to reach a desired location by using behavior decomposition. The system was divided into two task-oriented controllers, a fuzzy-neural obstacle avoidance controller and a drive controller for navigation to a location. The results from both systems were then fused for final control velocities using fuzzy logic methods. This system was successfully tested both in simulation with Matlab and experimentally. Experimental results using Khepera robots and an overhead vision system confirmed the simulation results and demonstrated robustness with regard to obstacle configuration and motion --Abstract, page iii