Abstract: For the path planning and autonomous obstacle avoidance problem of robot in complex environment， a real-time obstacle avoidance algorithm is proposed based on the prediction of the reachable area of dynamic obstacles. The static obstacles are described and modelled， and the state update prediction equations of dynamic obstacles are established to realize the prediction of the reachable area of dynamic obstacle mass center. For dynamic and static obstacles， a multi-step elliptic envelope potential field based on reachable area prediction and a potential field based on the novel Sigmoid function are proposed， respectively， to modify the target’s logarithmic Lyapunov gravitational field. A real-time obstacle avoidance algorithm is given for robot in the environment of multi-type obstacles. The numerical simulation and experimental results show that compared with traditional methods， the real-time obstacle avoidance algorithm can realize shorter path length， higher security and smaller driving angle during obstacle avoidance process.

Key words: robot; reachable area prediction; multi-step elliptic enveloping potential field; Sigmoid function potential field; real-time obstacle avoidance