We consider a multi-company, less-than-truckload, dynamic VRP based on the concept of multi-agent systems. We focus on the intelligence of one vehicle agent and especially on its bidding strategy. We address the problem how to price loads that are offered in real-time such that available capacity is used in the most profitable way taking into account possible future revenues. We develop methods to price loads dynamically based on revenue management concepts.
We consider a one leg problem, i.e., a vehicle travels from i to j and can wait at most τ time units in which it can get additional loads from i to j. We develop a DP to price loads given a certain amount of remaining capacity and an expected number of auctions in the time-to-go. Because a DP might be impractical if parameters change frequently and bids has to be determined in real-time, we derived two approximations to speed up calculations. The performance of these approximations are compared with the performance of the DP. Besides we introduce a new measure to calculate the average vehicle utilisation in consolidated shipments. This measure can be calculated based on a limited amount of data and gives an indication of the efficiency of schedules and the performance of vehicles.
|Name||BETA Working Paper|
|Publisher||Beta Research School for Operations Management and Logistics, University of Twente|