Incorporating travel time uncertainty into the design of service regions for delivery/pickup problems with time windows

Abstract

We attempt to incorporate stochastic travel time into the design of service regions for delivery/pickup problems with time windows. Different from the conventional model, our model is aimed at partitioning an entire service area into multiple service regions or clusters of customers. Each service region is served by a single vehicle. Once service regions are determined, vehicle routes within service regions can be adjusted from day to day in response to the changing traffic condition. We develop a scenario-based algorithm in which an index of “intimacy degree” is defined for grouping customers into different clusters. Two local search heuristics are constructed based on the index of “intimacy degree” to search the optimal solution. The results of our model would allow the decision makers to assess the trade-off among the number of vehicles used, the total travel cost, and the penalties incurred due to unserved customers and unbalanced workloads of drivers. Experiments are conducted to compare the performance of our algorithm with two other algorithms. The results indicate that our algorithm consistently outperforms the other two algorithms in terms of solution quality.