Unpredictable routes may be generated by varying the arrival time at each customer over successive visits. Inspired by a real-life case in cash distribution, this study presents an efficient solution approach for the vehicle routing problem with arrival time diversification by formulating it as a vehicle routing problem with multiple time windows in a rolling horizon framework. Because waiting times are not allowed, a novel algorithm is developed to efficiently determine whether routes or local search operations are time window feasible. To allow infeasible solutions during the heuristic search, four different penalty methods are proposed. The proposed algorithm and penalty methods are evaluated in a simple iterated granular tabu search that obtains new best-known solutions for all benchmark instances from the literature, decreasing average distance by 29% and reducing computation time by 93%. A case study is conducted to illustrate the practical relevance of the proposed model and to examine the trade-off between arrival time diversification and transportation cost.
- Multiple time windows
- Security constraints