This paper introduces a new path planning in dynamic nondeterministic environments. We combine POMDP and APF into the new path planning which takes full account of the uncertainty of the information in real world. Based on the APF's advantage of avoiding the expensive computation, it guides the setting of POMDP's rewards value to improve the efficiency of decision making. The result of the simulation shows that the proposed algorithm has higher search efficiency and can make the robot reach the target faster.