The dynamic pickup and delivery problem with time windows

This thesis investigates the large-scale dynamic pickup and delivery problem with time windows, and proposes heuristic solution approaches. The pickup and delivery problem with time windows deals with finding a set of optimal routes for a fleet of vehicles in order to serve transportation requests. Each transportation request requires pickup of a load at the pickup location and delivery of the load at the delivery location. The load is too small for capacity constraints to be relevant, and each stop location has to be served within a given time window. The dynamic pickup and delivery problem with time windows considers the process of solving the problem in an environment where not all requests are known in advance. While vehicles are serving requests, new requests are coming in, and their assignment is to be done on-line. Our focus is on the pure dynamic problem for which the future requests are not stochastically modelled or predicted. This models the problem faced by dispatchers in courier companies that serve same-day pickup and delivery requests for transporting letters and small parcels. The major results of our research include: a two-goal dynamic model for a dynamic routing problem, a two-strategy heuristic (a framework in which any heuristic or metaheuristic can be embedded), a dynamic utility function for the evaluation of one location insertion, recognizing the importance of developing a schedule in a dynamic environment (as opposed to a static environment), analysis of simple waiting strategies, design of complex waiting strategies for solving the route scheduling problem, and the use of precedence graphs for testing feasibility of insertion of a request into a route in the pickup and delivery problem with time windows. (The precedence graphs can also be used for finding bounds on the number of vehicles for the multiple traveling salesman problem with time windows.)