Opportunity costs calculation in agent-based vehicle routing and scheduling

In this paper we consider a real-time, dynamic pickup and delivery problem with timewindows where orders should be assigned to one of a set of competing transportation companies. Our approach decomposes the problem into a multi-agent structure where vehicle agents are responsible for the routing and scheduling decisions and the assignment of orders to vehicles is done by using a second-price auction. Therefore the system performance will be heavily dependent on the pricing strategy of the vehicle agents. We propose a pricing strategy for vehicle agents based on dynamic programming where not only the direct cost of a job insertion is taken into account, but also its impact on future opportunities. We also propose a waiting strategy based on the same opportunity valuation. Simulation is used to evaluate the benefit of pricing opportunities compared to simple pricing strategies in different market settings. Numerical results show that the proposed approach provides high quality solutions, in terms of profits, capacity utilization and delivery reliability.