An effective memetic algorithm for the generalized bike-sharing rebalancing problem
The generalized bike-sharing rebalancing problem (BRP) entails driving a fleet of capacitated vehicles to rebalance bicycles among bike-sharing system stations at a minimum cost. To solve this NP-hard problem, we present a highly effective memetic algorithm that combines (i) a randomized greedy cons...
Saved in:
| Published in: | Engineering applications of artificial intelligence 2020-10, Vol.95, p.103890, Article 103890 |
|---|---|
| Main Authors: | , , |
| Format: | Article |
| Language: | eng |
| Subjects: | |
| Online Access: | Get full text |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The generalized bike-sharing rebalancing problem (BRP) entails driving a fleet of capacitated vehicles to rebalance bicycles among bike-sharing system stations at a minimum cost. To solve this NP-hard problem, we present a highly effective memetic algorithm that combines (i) a randomized greedy construction method for initial solution generation, (ii) a route-copy-based crossover operator for solution recombination, and (iii) an effective evolutionary local search for solution improvement integrating an adaptive randomized mutation procedure. Computational experiments on real-world benchmark instances indicate a remarkable performance of the proposed approach with an improvement in the best-known results (new upper bounds) in more than 46% of the cases. In terms of the computational efficiency, the proposed algorithm shows to be nearly two to six times faster when compared to the existing state-of-the-art heuristics. In addition to the generalized BRP, the algorithm can be easily adapted to solve the one-commodity pickup-and-delivery vehicle routing problem with distance constraints, as well as the multi-commodity many-to-many vehicle routing problem with simultaneous pickup and delivery.
•A highly effective memetic algorithm (MA) for the bike-sharing rebalancing problem.•MA combines evolutionary local search with a route-copy-based crossover.•Improved best upper bounds are reported for 42 medium and large instances.•Adaptations of MA are able to provide competitive results for the related 1-PDVRPD and M-M-VRPSPD problems. |
|---|---|
| ISSN: | 0952-1976 1873-6769 |