Optimizing Dynamic VRP for on-Demand Bus with Time Window and Uncertain Traveling Time
As the elderly population rises and the development of social welfare, more attention is being focused on the elders and disabilities to their transportation demand, but the traditional bus with a fixed schedule can’t offer door-to-door service. Therefore, On-demand bus was promoted. To make the problem more realistic, this paper focuses on real-time scheduling, and summarizes two main effect factors while scheduling of bus routes: (i) dynamic customers; (ii) uncertainty of traveling time; these two kinds of the problem will be solved in this paper.Problem 1 is a deterministic problem which only considered (i); Problem 2 is a stochastic problem which combined both (i) and (ii). In this paper, a multi-objective mathematical model of dynamic on-demand bus scheduling is built to balance the total cost and service levelunder uncertainty. At last, the realistic data from map is set to verify the effect of the proposed method.In problem 1, the effect of soft time window proposal is compared with a hard time window, and the solution quality of proposed method HHA is evaluated by comparing with traditional GA and GA + Nearest Neighbor method. In problem 2, the robustness and calculation time of scenario based HHA is also discussed.The results show that HHA is better than GA in deterministic problem, and scenario-based HHA is more robust than GA under the same uncertainty in the stochastic problem.
Keywords - On-demand Bus, Dynamic Customers, Stochastic Traveling Time, Genetic Algorithm, Local Search.