Abstract: In this paper we present a virtual force guidance system for motion planning .This method is developed to work with real-time least mean square error estimation algorithm (LMSE) and can effectivelynavigate the robot and avoid the collision with obstacle.We adjust the weight dynamically according to the prediction error and different navigation sessons,so that the prediction is as accurate as possible.The basic idea is to acquire the motion information of the robot and to enable virtual force guidance system to determine the future motion according to the estimation.The VFG system assures that the goal attracts the robot and obstacles repulse it, the resultant force determines the motion.Simulation results are presented to show fine real-time property,avoid the obstacle effectively and arrive at the goal.