Core Concepts
The authors propose improved approximation algorithms for the Multidepot Capacitated Vehicle Routing Problem (MCVRP), achieving better approximation ratios compared to previous work.
Abstract
The Multidepot Capacitated Vehicle Routing Problem (MCVRP) is a well-known variant of the classic Capacitated Vehicle Routing Problem (CVRP), where vehicles located in multiple depots need to serve customers' demand while minimizing the total traveling distance. There are three versions of MCVRP based on the demand property: unit-demand, splittable, and unsplittable.
The authors review previous approximation algorithms for MCVRP and CVRP, and then make the following contributions:
They propose a refined tree-partition algorithm based on the idea in [14], which has a better approximation ratio for the case when the vehicle capacity k is fixed.
By combining the refined tree-partition algorithm with recent progress in approximating CVRP [9], they obtain an improved (4 - 1/1500)-approximation algorithm for splittable and unit-demand k-MCVRP.
They design an LP-based cycle-partition algorithm for unsplittable k-MCVRP, which achieves an improved (4 - 1/50000)-approximation ratio by leveraging the result in [10].
For the case when k is fixed, they propose an LP-based tree-partition algorithm that achieves a (3 + ln 2 - Θ(1/√k))-approximation ratio for all three versions of k-MCVRP, which is better than the current-best ratios for any k > 11.
By further combining the LP-based tree-partition algorithm with the result in [9], they show that the approximation ratio can be improved to 3 + ln 2 - max{Θ(1/√k), 1/9000} for splittable and unit-demand k-MCVRP.