Solving multi-objective dynamic vehicle routing problem with time windows using multi-objective algorithm

Khoo, Thau Soon (2022) Solving multi-objective dynamic vehicle routing problem with time windows using multi-objective algorithm. PhD thesis, UTAR.

Preview

PDF Download (8Mb) | Preview

Abstract

Logistics plays a very important role in the business economy. It is over a trillion of dollars in revenue annually and increase exponentially over the years One of the current trends is to solve the last mile is to optimize the delivery routes. One of the best ways to optimize the delivery routes is to study and implement the multi-objective dynamic vehicle routing problem with time windows because it resembles the online delivery services that are ubiquitous and propagate over the year, especially during the COVID-19 pandemic. During the past decade, there is an increasing trend of published papers dealing with dynamic vehicle routing problems with time windows (DVRPTW) but not on multi-objective dynamic vehicle routing problems with time windows (MODVRPTW). Therefore, it brings a significant contribution if this study can be carried out because it represents the daily real-life problem in transportation. To solve this problem, it needs to be modelled and an algorithm is needed to be developed and tested to ascertain its efficiency and effectiveness. It is difficult and challenging to develop an algorithm that can produce consistent near-optimal solutions even after many runs, average near-optimal solutions that have the least difference in magnitude, broader Pareto set, and iii achieve near-optimal solutions but highly sought after if it is commercially viable. Our algorithm uses non-fitness evolutionary distributed parallelized adaptive large neighbourhood search (NEDPALNS). The non-fitness evolutionary distributed (NED) takes advantage of the exploitation of the search space and the parallelized adaptive large neighbourhood search (PALNS) makes full use of the exploration and exploitation of its inner strength. These combinations achieve near-optimal solutions consistently. We compare our results using hypothetical datasets and real datasets. Our results are competitive and outperform other published algorithms and best-known solutions in both static and dynamic environments.

Actions (login required)