Management policies in a dynamic multi-period routing problem

Summary: In this paper we analyze the Dynamic Multi-Period Routing Problem (DMPRP), where a fleet of uncapacitated vehicles has to satisfy customers' pick-up requests. The service of each customer can take place the day the request is issued or the day after. At the beginning of a day a set of requests are already known and have to be served during the day. Additional requests may arrive during the day while the vehicles are traveling. In this context we perform different types of analysis, each one characterized by the comparison of alternative management policies. The first analysis compares a policy which decides, at the time the request is issued, whether to accept or reject it to a policy that accepts all the requests and decides, at a later time, which ones to forward to a back-up service company. The second evaluates the advantages of a collaborative service policy where a fleet of vehicles is managed by a unique decision maker with respect to a policy where the same vehicles are managed independently. Finally, in the last analysis a policy where each new request is taken into account as soon as it is issued is compared to a policy where all the requests issued during a day are analyzed at the end of the day. Extensive computational results evaluating the number of lost requests and the distance traveled provide interesting insights.