Decomposition Algorithm for the Multi-Trip Single Vehicle Routing Problem with AND-type Precedence Constraints

TL;DR

This paper introduces a novel decomposition algorithm for a complex vehicle routing problem with AND-type precedence constraints, proposing new models and solution methods that improve efficiency and solution quality.

Contribution

It develops three MIP models and a Logic-Based Benders Decomposition approach with a new optimality cut for the multi-trip vehicle routing problem with precedence constraints.

Findings

The LBBD algorithm outperforms existing methods in solution speed.

The new optimality cut accelerates convergence.

The models effectively handle complex precedence constraints.

Abstract

This paper addresses a new variant of the multi-trip single vehicle routing problem where the nodes are related to each other through AND-type precedence constraints. The problem aims at determining a sequence of trips to visit all the nodes respecting every precedence constraint within and among the routes so to minimize the total traveling cost. Our motivation comes from routing problems where a node may have a set of predecessors (not just single one proposed in the dial-a-ride or pickup and delivery problems) resulting in a set of pairwise relations that specify which customers need to be visited before which other ones. We develop three Mixed Integer Programming (MIP) models to formulate the problem. The models are experimentally compared to determine the best one. Moreover, a solution approach based on the Logic-Based Benders Decomposition (LBBD) algorithm is developed which allows to decompose the original problem into an assignment master problem and independent sequencing subproblems. A new valid optimality cut is devised to achieve faster convergence. The cut performance is experimentally investigated by comparing with a recently proposed one in the literature. We further relax the algorithm to find the sub-optimal solution and demonstrate its efficiency. Extensive computational experiments are conducted to assess the proposed algorithms in terms of solution quality and CPU time.