Bundle selection approaches for collaborative practical-oriented Pickup and Delivery Problems

Due to the increasing price pressure in the less-than-truckload (LTL) market, horizontal cooperation is an effective and efficient way for small- and medium-sized LTL carriers to enhance their profits by exchanging requests. For this purpose, a decentralized auction-based collaboration framework has proven to provide a good approach. In this paper, such a collaboration framework is adopted and extended by applying it to a practical-oriented routing problem, the Pickup and Delivery Problem with Time Windows and Heterogeneous Vehicle Fleets. Particularly, we analyse the bundle selection process made by the auctioneers, which is a stochastic problem and specify which requests are supposed to be offered together in a bundle to the cooperating carriers. For the purpose of solving the selection problem properly, we implement a new procedure required due to the characteristics of the considered routing problem: a scenario-based bundle selection approach. In order to make this approach applicable, two pre-selection techniques (cluster- and neural network-based) are developed. Our approach is evaluated on 240 collaboration network instances created from well-known pickup and delivery research data sets generated by Li and Lim (2001). The collaboration framework results are compared with respect to the profit to individual transportation planning of each carrier (lower threshold) as well as centralized transportation planning of all carriers (upper threshold). It can be shown that the auction-based collaboration approach is up to 43.49% better than the individual planning as well as exhausts at least 53.5% of the centralized transportation planning potential on average.