This paper considers the problem of multiple automated guided vehicles (multi-AGVs) outbound scheduling in warehousing and logistics. The considered problem combines two sub problems, i.e., the task allocation problem for multiple AGVs and the path planning problem for each AGV. A reinforcement learning (RL) method is presented to address this problem. The presented RL is a kind of Q-learning-based hyper-heuristic (QHH), which utilizes Q-learning as the upper-level strategy and dynamically selects heuristic operations from pre-designed lower-level Operators to execute search in solution space. The RL employs a hyper-heuristic algorithm based on Q-learning (QHH) algorithm. The QHH framework uses Q-learning as the upper-level strategy and selects heuristic operations from pre-designed lower-level Operators. Execute on the solution space to obtain better solutions. Considering task allocation and path planning for multi AGV outbound in warehousing, the dual layer encoding scheme for task allocation problem was designed, and the corresponding decoding strategies were designed in the path planning stage. In addition, the Taguchi method is adopted to adjust the key parameters of the presented RL. Experimental results on the simulation instances demonstrate that the RL algorithm is effective for dealing with the considered problem.