Search

Following decades of sustained improvement, metaheuristics are one of the great success stories of optimization research. However, in order for research in metaheuristics to avoid fragmentation and a lack of reproducibility, there is a pressing need for stronger scientific and computational infrastructure to support the development, analysis and comparison of new approaches. To this end, we present the vision and progress of the Metaheuristics. In the Large project. The conceptual underpinnings of the project are: truly extensible algorithm templates that support reuse without modification, white box problem descriptions that provide generic support for the injection of domain specific knowledge, and remotely accessible frameworks, components and problems that will enhance reproducibility and accelerate the field’s progress. We argue that, via such principled choice of infrastructure support, the field can pursue a higher level of scientific enquiry. We describe our vision and report on progress, showing how the adoption of common protocols for all metaheuristics can help liberate the potential of the field, easing the exploration of the design space of metaheuristics.

This work tackles a rich vehicle routing problem (VRP) problem integrating a capacitated vehicle routing problem with time windows (CVRPTW), and a service technician routing and scheduling problem (STRSP) for delivering various equipment based on customers' requests, and the subsequent installation by a number of technicians. The main objective is to reduce the overall costs of hired resources, and the total transportation costs of trucks/technicians. The problem was the topic of the fourth edition of the VeRoLog Solver Challenge in cooperation with the ORTEC company. Our contribution to research is the development of a mathematical model for this problem and a novel hyper?heuristic algorithm to solve the problem based on a population of solutions. Experimental results on two datasets of small and real?world size revealed the success of the hyper?heuristic approach in finding optimal solutions in a shorter computational time, when compared to our exact model. The results of the large size dataset were also compared to the results of the eight finalists in the competition and were found to be competitive, proving the potential of our developed hyper?heuristic framework.

This paper presents the results of the second edition of the Wind Farm Layout Optimization Competition, which was held at the 22nd Genetic and Evolutionary Computation COnference (GECCO) in 2015. During this competition, competitors were tasked with optimizing the layouts of five generated wind farms based on a simplified cost of energy evaluation function of the wind farm layouts. Online and offline APIs were implemented in C++, Java, Matlab and Python for this competition to offer a common framework for the competitors. The top four approaches out of eight participating teams are presented in this paper and their results are compared. All of the competitors' algorithms use evolutionary computation, the research field of the conference at which the competition was held. Competitors were able to downscale the optimization problem size (number of parameters) by casting the wind farm layout problem as a geometric optimization problem. This strongly reduces the number of evaluations (limited in the scope of this competition) with extremely promising results.

Multi-mode resource and precedence-constrained project scheduling is a well-known challenging real-world optimisation problem. An important variant of the problem requires scheduling of activities for multiple projects considering availability of local and global resources while respecting a range of constraints. A critical aspect of the benchmarks addressed in this paper is that the primary objective is to minimise the sum of the project completion times, with the usual makespan minimisation as a secondary objective. We observe that this leads to an expected different overall structure of good solutions and discuss the effects this has on the algorithm design. This paper presents a carefully-designed hybrid of Monte-Carlo tree search, novel neighbourhood moves, memetic algorithms, and hyper-heuristic methods. The implementation is also engineered to increase the speed with which iterations are performed, and to exploit the computing power of multicore machines. Empirical evaluation shows that the resulting information-sharing multi-component algorithm significantly outperforms other solvers on a set of “hidden” instances, i.e. instances not available at the algorithm design phase.

There is a growing interest towards the design of reusable general purpose search methods that are applicable to di?erent problems instead of tailored solutions to a single particular problem. Hyper-heuristics have emerged as such high level methods that explore the space formed by a set of heuristics (move operators) or heuristic components for solving computationally hard problems. A selection hyper-heuristic mixes and controls a prede?ned set of low level heuristics with the goal of improving an initially generated solution by choosing and applying an appropriate heuristic to a solution in hand and deciding whether to accept or reject the new solution at each step under an iterative framework. Designing an adaptive control mechanism for the heuristic selection and combining it with a suitable acceptance method is a major challenge, because both components can in?uence the overall performance of a selection hyper-heuristic. In this study, we describe a novel iterated multi-stage hyper-heuristic approach which cycles through two interacting hyper-heuristics and operates based on the principle that not all low level heuristics for a problem domain would be useful at any point of the search process. The empirical results on a hyper-heuristic benchmark indicate the success of the proposed selection hyper-heuristic across six problem domains beating the state-of-the-art approach.

Automating high school timetabling is a challenging task. This problem is a well known hard computational problem which has been of interest to practitioners as well as researchers. High schools need to timetable their regular activities once per year, or even more frequently. The exact solvers might fail to find a solution for a given instance of the problem. A selection hyper-heuristic can be defined as an easy-to-implement, easy-to-maintain and effective 'heuristic to choose heuristics' to solve such computationally hard problems. This paper describes the approach of the team hyper-heuristic search strategies and timetabling (HySST) to high school timetabling which competed in all three rounds of the third international timetabling competition. HySST generated the best new solutions for three given instances in Round 1 and gained the second place in Rounds 2 and 3. It achieved this by using a fairly standard stochastic search method but significantly enhanced by a selection hyper-heuristic with an adaptive acceptance mechanism. © 2014 Springer Science+Business Media New York.

High school timetabling is one of those recurring NP-hard real-world combinatorial optimisation problems that has to be dealt with by many educational institutions periodically, and so has been of interest to practitioners and researchers. Solving a high school timetabling problem requires scheduling of resources and events into time slots subject to a set of constraints. Recently, an international competition, referred to as ITC 2011 was organised to determine the state-of-the-art approach for high school timetabling. The problem instances, obtained from eight different countries across the world used in this competition became a benchmark for further research in the field. Selection hyper-heuristics are general-purpose improvement methodologies that control/mix a given set of low level heuristics during the search process. In this study, we evaluate the performance of a range of selection hyper-heuristics combining different reusable components for high school timetabling. The empirical results show the success of the approach which embeds an adaptive great-deluge move acceptance method on the ITC 2011 benchmark instances. This selection hyper-heuristic ranks the second among the previously proposed approaches including the ones competed at ITC 2011.

A square matrix of distinct numbers in which every row, column and both diagonals have the same total is referred to as a magic square. Constructing a magic square of a given order is considered a difficult computational problem, particularly when additional constraints are imposed. Hyper-heuristics are emerging high-level search methodologies that explore the space of heuristics for solving a given problem. In this study, we present a range of effective selection hyper-heuristics mixing perturbative low-level heuristics for constructing the constrained version of magic squares. The results show that selection hyper-heuristics, even the non-learning ones deliver an outstanding performance, beating the best-known heuristic solution on average.

There is a growing interest towards self configuring/tuning automated general-purpose reusable heuristic approaches for combinatorial optimisation, such as, hyper-heuristics. Hyper-heuristics are search methodologies which explore the space of heuristics rather than the solutions to solve a broad range of hard computational problems without requiring any expert intervention. There are two common types of hyper-heuristics in the literature: selection and generation methodologies. This work focuses on the former type of hyper-heuristics. Almost all selection hyper-heuristics perform a single point based iterative search over the space of heuristics by selecting and applying a suitable heuristic to the solution in hand at each decision point. Then the newly generated solution is either accepted or rejected using an acceptance method. This improvement process is repeated starting from an initial solution until a set of termination criteria is satisfied. The number of studies on the design of hyper-heuristic methodologies has been rapidly increasing and currently, we already have a variety of approaches, each with their own strengths and weaknesses. It has been observed that different hyper-heuristics perform differently on a given subset of problem instances and more importantly, a hyper-heuristic performs differently as the set of low level heuristics vary. This thesis introduces a general "multi-stage" hyper-heuristic framework enabling the use and exploitation of multiple selection hyper-heuristics at different stages during the search process. The goal is designing an approach utilising multiple hyper-heuristics for a more effective and efficient overall performance when compared to the performance of each constituent selection hyper-heuristic. The level of generality that a hyper-heuristic can achieve has always been of interest to the hyper-heuristic researchers. Hence, a variety of multi-stage hyper-heuristics based on the framework are not only applied to the real-world combinatorial optimisation problems of high school timetabling, multi-mode resource-constrained multi-project scheduling and construction of magic squares, but also tested on the well known hyper-heuristic benchmark of CHeSC 2011. The empirical results show that the multi-stage hyper-heuristics designed based on the proposed framework are still inherently general, easy-to-implement, adaptive and reusable. They can be extremely effective solvers considering their success in the competitions of ITC 2011 and MISTA 2013. Moreover, a particular multi-stage hyper-heuristic outperformed the state-of-the-art selection hyper-heuristic from CHeSC 2011.