An investigation of the ant-based hyper-heuristic for capacitated vehicle routing problem and traveling salesman problem

Abd Aziz, Zalilah (2013) An investigation of the ant-based hyper-heuristic for capacitated vehicle routing problem and traveling salesman problem. PhD thesis, University of Nottingham.

PDF - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
Download (7MB) | Preview


A brief observation on recent research of routing problems shows that most of the methods used to tackle the problems are using heuristics and metaheuristics; and they often use problem specific knowledge to build or improve solutions. In the last few years, research on hyper-heuristic has been investigated which aims to raise the generality of optimisation systems. This thesis is concerned with the investigation of ant-based hyper-heuristic. Ant algorithms have been applied to vehicle routing problems and have produced competitive results. Therefore, it is assumed that there is a reasonable possibility that ant-based hyperheuristic could perform well for the problem.

The thesis first surveys the literature for some common solution methodologies for optimisation problems and explores in some detail the ant algorithms and ant algorithm hyperheuristic methods. Furthermore, the literature specifically concerns with routing problems; the capacitated routing problem (CVRP) and the travelling salesman problem (TSP). The thesis studies the ant system algorithm and further proposes the ant algorithm hyper-heuristic, which introduces a new pheromone update rule in order to improve its performance. The proposed approach, called the ant-based hyper-heuristic is tested to two routing problems; the CVRP and TSP. Although it does not produce any best known results, the experimental results have shown that it is competitive with other methods. Most importantly, it demonstrates how simple and easy to implement low level heuristics, with no extensive parameter tuning. Further analysis shows that the approach possesses learning mechanism when compared to random hyper-heuristic. The approach investigates the number of low level heuristics appropriate and found out that the more low level heuristics used, the better solution is generated. In addition an ACO hyper-heuristic which has two categories of pheromone updates is developed. However, ant-based hyper-heuristic performs better and this is inconsistent with the performance of ACO algorithm in the literature.

In TSP, we utilise two different categories of low level heuristics, the TSP heuristics and the CVRP heuristics that were previously used for the CVRP. From the observation, it can be seen that by using any heuristics for the same class of problems, ant-based hyper-heuristic is seen to be able to produce competitive results. This has demonstrated that the ant-based hyper-heuristic is a reusable method. One major advantage of this work is the usage of the same parameter for all problem instances with simple moves and swap procedures. It is hoped that in the future, results obtained will be better than current results by using better intelligent low level heuristics.

Item Type: Thesis (University of Nottingham only) (PhD)
Supervisors: Kendall, G.
Keywords: routing, vehicle, traveling salesman problem, travelling salesman problem, ant-based, program transformation, heuristic programming, computer networks
Subjects: T Technology > T Technology (General)
T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5101 Telecommunication
Faculties/Schools: UK Campuses > Faculty of Science > School of Computer Science
Item ID: 28513
Depositing User: Jacob, Mr Tim
Date Deposited: 10 Mar 2015 09:25
Last Modified: 15 Dec 2017 10:17

Actions (Archive Staff Only)

Edit View Edit View