To satisfy the requirements of efficiency, feasibility, and optimality of unmanned aerial vehicle path planning in dynamic environment, an anytime repairing sparse A* search (AR-SAS) algorithm is proposed, by incorporating the sparse A* search (SAS) into anytime repairing framework and introducing double criteria ordering, memory bounded and adaptivestep expanding strategies into the process of path optimization. Monte-Carlo simulations in static environment demonstrate that AR-SAS takes less time to generate the feasible path and optimal path compared with standard SAS and hierarchical SAS. Simulation results in dynamic environment show that AR-SAS can satisfy the requirements of dynamic planning to rapidly produce a feasible path and gradually improve the path quality in given time.