Radio-Frequency Identification (RFID) inventory refers to a method of inventorying and managing stock using RFID technology. To conduct RFID inventory in indoor environments, it is now common practice to deploy RFID readers at all key locations to cover all items. This approach has the disadvantages of high cost and high redundancy. In this paper, we propose a RFID inventory method based on unmanned vehicles. We equip RFID devices on the unmanned vehicles and the unmanned vehicles to complete the inventory of all items. We first use the improved Grey Wolf Optimizor (GWO) to find the optimal RFID reader working locations, and then use these locations as nodes to be traversed by unmanned vehicles and plan the traversal path using a reinforcement learning algorithm. By traversing these nodes, the unmanned vehicles completes the inventory of all items. Our improved GWO can achieve 100% coverage, which is about 15% better compared to the coverage of the original GWO. The node traversal algorithm’s model has a shorter computation time than the optimization algorithm after it has been trained.