Multi-agent decentralized scheduling for dynamic client requirements in logistics

For the dynamic client of pickup in logistic, the existing model is that distribution center redeploys vehicles to service for client after accumulation of a certain amount of orders. This approach can't respond to client in time. Re-deployment of vehicles will also increase distribution costs. For the present situation, the paper proposes a multi-agent decentralized control dispatching model which integrates pickup and delivery process. All in-transit vehicles are an independent agent. When a dynamic client has a request of pickup, the nearby in-transit agents can process it in time. If an agent satisfies the condition, it can optimize the distribution route through the insertion algorithm and the self-learning algorithm. Finally, according to the difference of the objective functions before and after inserting dynamic client, setting different priority, the agent best suited the order has the highest priority, when more than one agents meet the conditions for the order.