A genetic column generation algorithm for sustainable spare part delivery: application to the Sydney DropPoint network

Abstract

Modern-day logistics companies require increasingly shorter lead-times in order to cater for the increasing popularity of on-demand services. There is consequently an urgent need for fast scheduling algorithms to provide high quality, real-time implementable solutions. In this work we model a spare part delivery problem for an on-demand logistics company, as a variant of vehicle routing problem. For large delivery networks, the optimisation solution technique of column generation has been employed successfully in a variety of vehicle routing settings and is often used in combination with exact methods for solving problems with a large number of variables. Challenges may arise when the pricing subproblem is difficult to solve in a realistic period due to complex constraints or a large number of variables. The problem may become intractable when the network structure varies daily or is known with less certainty over longer period. In such instances, a high quality heuristic solution may be more preferable than an exact solution with excessive run time. We propose an improved version of column generation approach integrating an efficient genetic algorithm to obtain fast and high-quality solutions for a sustainable spare parts delivery problem. More specifically, we propose to retain the traditional column generation iterative framework, with master problem solved exactly, but with pricing subproblem solved using a metaheuristic. Computational results on a real dataset indicate that this approach yields improved solutions compared to the current best-case business-as-usual costs. It also substantially decreases the computational time; allowing for high-quality, tractable solutions to be obtained in few minutes. We propose to strike a balance between the practical and efficient solution aspects of metaheuristic algorithms, and the exact decomposition and iterative aspect of the column generation solution technique.

Identical Parallel Machine Scheduling Considering Workload Smoothness Index

“…The genetic algorithm simulates the natural selection and genetic mechanism in Darwin's biological evolution theory. As meta-heuristic search algorithms, genetic algorithms have been extensively used to solve scheduling problems in manufacturing systems [29] and other NP-hard problems [30]. In the genetic algorithm, each chromosome represents a feasible solution [31].…”

A hybrid column-generation and genetic algorithm approach for solving large-scale multimission selective maintenance problems in serial K-out-of-n:G systems

“…Liu, Haghani, and Toobaie (2010) also apply CG to the passenger rail crew scheduling and used a GA to solve the induced subproblems. Dunbar et al (2020) propose a CG approach where the restricted master problem (RMP) is solved exactly and a GA is used to solve the subproblems. The results of this study indicate that the approach yields improved solutions compared to the current best-case costs.…”