Delivery vehicle dynamic scheduling optimization method based on DDQN algorithm

The invention discloses a distribution vehicle dynamic scheduling optimization method based on a DDQN algorithm, and belongs to the technical field of fresh food distribution vehicle scheduling based on deep reinforcement learning. According to the method, a fresh food delivery dynamic vehicle scheduling problem is regarded as a continuous time process, modeling is carried out based on an SMDP (Semi-Markov Decision Process) framework, double Agents are trained by adopting a DDQN (Double Deep Q-Learning) algorithm, and scheduling allocation is carried out when a new order event and a vehicle event are processed. According to the method, the combination complexity of the distribution space is remarkably reduced, and better average distribution time is shown while a plurality of distribution limiting factors are considered. By improving the system resource utilization rate and the scheduling efficiency, the problem that the timeliness of fresh products is reduced due to fresh product delivery delay is solved. 本发