Exploring Heuristics for the Vehicle Routing Problem with Split Deliveries and Time Windows.

摘要

This dissertation investigates the Vehicle Routing Problem with Split Deliveries and Time Windows. This problem assumes a depot of homogeneous vehicles and set of customers with deterministic demands requiring delivery. Split deliveries allow multiple visits to a customer and time windows restrict the time during which a delivery can be made. Several construction and local search heuristics are tested to determine their relative usefulness in generating solutions for this problem. This research shows a particular subset of the local search operators is particularly influential on solution quality and run time. Conversely, the construction heuristics tested do not significantly impact either. Several problem features are also investigated to determine their impact. Of the features explored, the ratio of customer demand to vehicle ratio revealed a significant impact on solution quality and influence on the effectiveness of the heuristics tested. Finally, this research introduces an ant colony metaheuristic coupled with a local search heuristic embedded within a dynamic program seeking to solve a Military Inventory Routing Problem with multiple-customer routes, stochastic supply, and deterministic demand. Also proposed is a suite of test problems for the Military Inventory Routing Problem.