In three dimensional environments feasible paths for UAVs should be smooth, flyable and safe. To solve the problem of multi-UAVs cooperative path planning with kinds of constraints, a multi-objective optimization approach based on differential evolution is proposed, which uses a group of UAV paths as evolution individuals. In this paper, we adopt a novel combined strategy to enhance the efficiency of difference. To fit UAV constraints and cooperative constraints, a cooperative fitness function is given to measure the pareto optimal solution. And then the control vertexes of a B-spline curve are calculated reversely to get a group of flyable smoothing paths for UAVs. The simulation results show that the proposed approach effectively improves the cooperative accuracy and works well in smoothing paths for UAVs.