Abstract: Obstacles avoiding is the difficult problem of autonomous robot in dynamic environment.Firstly，the model of environment including robot，obstacles and target are established and given.Then humanoid strategies of walking are analyzed and formulated into Virtual Force Field（VFF） algorithm.In order to deal with moving obstacles，the adjacent smoothing method is used in accumulation computing of Confirmation Factors（CF） which mean the degree of grid being obstacle.The radius of adjacent area is designed to be relevant to the obstacle moving speed.Uncertain reasoning is adopted for computing complex CF with uncertain of environment and sensors signal，and synthesize method is designed for multi-sensors detecting obstacles.When target is out robot sensor workspace，the attraction force computing formula using direction parameters is given.The weight varying with angle from robot moving direction to obstacles is designed，and used to make robot more sensitivity to obstacles on moving way like human.In order to walk out local minima point，the virtual target is used，which is obtained by temporal changing target direction.Indoor and outdoor virtual environment are established on MobotSim，and the algorithm proposed is programmed using Basic.The simulation tests are finished and analyzed.The given results demonstrate the feasibility and effectiveness of proposed method.The route of robot is better by VFF with humanoid strategy than common VFF.

摘要： 动态环境下避障是机器人实现自主运动的关键。首先建立了适合虚拟力场算法的机器人工作环境数学描述。将人避障行走策略引入虚拟力场中，具体包括：设计了单元格障碍物可信度的邻域平滑累积值计算方法，模拟人对移动障碍物的躲避策略；建立可信度的不确定推理计算方法，处理信号和环境存在干扰问题；设计了基于目标点方位角的吸引力计算公式来解决目标点超出感知空间问题；设计了变权重加权排斥力计算方法，使机器人对前进方向的障碍更敏感；借鉴人绕开障碍物策略，采用临时旋转目标点方向得到的虚拟目标点来使机器人沿障碍物运动直到绕开。针对房间和街面环境，在MobotSim平台上进行仿真实验，给出了实验结果和分析。在合理设置参数下，机器人能避开障碍物到达目标点，且避障路径优于传统的虚拟力场方法。结果验证了该方法的有效性。