Abstract: Obstacle avoidance is the key of autonomous navigation process for mobile robots all along. The current methods are almost the optimizing process for some one designated destination. Some deadlocks or collision phenomena often occur for the optimal solution or path. Therefore, the theory of multiple objective optimizations is introduced and an approach of path planning is proposed based on dynamic field weighting. The algorithm divides the obstacle avoidance into the process of three behaviors fusion. It adjusts the ratio of behaviors output by setting different weighting for different sub behavior function. The solution or path obtained at current moment is not optimal, but the most efficient or most satisfying. The availability and robustness of the obstacle avoidance system are analyzed and approved in theory. From the real and simulation experiments of autonomous navigation, the algorithm not only retains the real-time and robustness of obstacle avoidance, but also improves its availability and safety.

Key words: Behavior fusion, Deadlock, Multiple objective optimization, Reliability of obstacle avoidance