The "last kilometer" logistics distribution is the last link of the logistics terminal facing the customers. The road limitations of the vehicle are overcome through the vehicle-mounted drone distribution mode, but due to the short endurance time because of the limited capacity of the battery of the drone, each battery of the drone need to be replaced after delivery mission is completed, which makes the labor cost as well as the increase of the distribution time. Based on this cooperative distribution mode of unmanned aerial vehicle and vehicle, the battery power consumption of UAV (unmanned aerial vehicle) and charging rate limitation on vehicle were considered, a distribution path planning model for charging on vehicle when UAV had no delivery task was established. An improved GRASP(greedy randomized adaptive search procedure) algorithm combining genetic algorithm and neighborhood search algorithm was designed to solve this model. Test experiments are conducted based on the designed test set, and sensitivity analysis is conducted on the main factors affecting the path planning to make full use of the UAV's distribution capability.