Humanoid strategies of avoiding obstacles were analyzed and formulated into virtual force field (VFF) algorithm. In view of obstacles moving, smoothing accumulation computing method was designed for updating credibility factors (CF) of grid held by obstacle, and two examples were given. Uncertainty reasoning was adopted for computing complex CF with uncertainty of sensors signal, and synthesize method was designed. The weight of repulsive force was designed to be varying with angle from robot moving direction to obstacles, in order to make robot more sensitivity to obstacles on moving way. Indoor and outdoor virtual environment were established and simulations were carried out. The results demonstrate that the path is better using proposed than common VFF.