A new fusion algorithm is proposed for the trajectory planning problem of patrol robots avoiding obstacles to detect land in neighborhood environment. Probabilistic search algorithm（ PRM） is selected by simulation comparison experiments of various algorithms in a complex static block environment. In dynamic obstacle-intensive environment

the introduction of"random reachable set"improves the artificial potential field（ APF） method. Simulation experiments show that the fusion algorithm reduces the trajectory planning time of the robot and improves the obstacle avoidance success rate of the robot.