Mauro Birattari

Ant colony optimization: artificial ants as a computational intelligence technique

The introduction of ant colony optimization (ACO) and to survey its most notable applications are discussed. Ant colony optimization takes inspiration from the forging behavior of some ant species. These ants deposit Pheromone on the ground in order to mark some favorable path that should be followed by other members of the colony. The model proposed by Deneubourg and co-workers for explaining the foraging behavior of ants is the main source of inspiration for the development of ant colony optimization. In ACO a number of artificial ants build solutions to an optimization problem and exchange information on their quality through a communication scheme that is reminiscent of the one adopted by real ants. ACO algorithms is introduced and all ACO algorithms share the same idea and the ACO is formalized into a meta-heuristics for combinatorial problems. It is foreseeable that future research on ACO will focus more strongly on rich optimization problems that include stochasticity.

Swarm robotics: a review from the swarm engineering perspective

Swarm robotics is an approach to collective robotics that takes inspiration from the self-organized behaviors of social animals. Through simple rules and local interactions, swarm robotics aims at designing robust, scalable, and flexible collective behaviors for the coordination of large numbers of robots. In this paper, we analyze the literature from the point of view of swarm engineering: we focus mainly on ideas and concepts that contribute to the advancement of swarm robotics as an engineering field and that could be relevant to tackle real-world applications. Swarm engineering is an emerging discipline that aims at defining systematic and well founded procedures for modeling, designing, realizing, verifying, validating, operating, and maintaining a swarm robotics system. We propose two taxonomies: in the first taxonomy, we classify works that deal with design and analysis methods; in the second taxonomy, we classify works according to the collective behavior studied. We conclude with a discussion of the current limits of swarm robotics as an engineering discipline and with suggestions for future research directions.

Frankenstein's PSO: A Composite Particle Swarm Optimization Algorithm

During the last decade, many variants of the original particle swarm optimization (PSO) algorithm have been proposed. In many cases, the difference between two variants can be seen as an algorithmic component being present in one variant but not in the other. In the first part of the paper, we present the results and insights obtained from a detailed empirical study of several PSO variants from a component difference point of view. In the second part of the paper, we propose a new PSO algorithm that combines a number of algorithmic components that showed distinct advantages in the experimental study concerning optimization speed and reliability. We call this composite algorithm Frankensteins PSO in an analogy to the popular character of Mary Shelleys novel. Frankensteins PSO performance evaluation shows that by integrating components in novel ways effective optimizers can be designed.

Lazy Learning for Local Modelling and Control Design

This paper presents local methods for modelling and control of discrete-time unknown non-linear dynamical systems, when only input-output data are available. We propose the adoption of lazy learning, a memory-based technique for local modelling. The modelling procedure uses a query-based approach to select the best model configuration by assessing and comparing different alternatives. A new recursive technique for local model identification and validation is presented, together with an enhanced statistical method for model selection. A lso, three methods to design controllers based on the local linearization provided by the lazy learning algorithm are described. In the first method the lazy technique returns the forward and inverse models of the system which are used to compute the control action to take. The second is an indirect method inspired by self-tuning regulators where recursive least squares estimation is replaced by a local approximator. The third method combines the linearization provided by t...

F-Race and Iterated F-Race: An Overview

Algorithms for solving hard optimization problems typically have several parameters that need to be set appropriately such that some aspect of performance is optimized. In this chapter, we review F-Race, a racing algorithm for the task of automatic algorithm configuration. F-Race is based on a statistical approach for selecting the best configuration out of a set of candidate configurations under stochastic evaluations. We review the ideas underlying this technique and discuss an extension of the initial F-Race algorithm, which leads to a family of algorithms that we call iterated F-Race. Experimental results comparing one specific implementation of iterated F-Race to the original F-Race algorithm confirm the potential of this family of algorithms.

Improvement strategies for the F-Race algorithm: sampling design and iterative refinement

Finding appropriate values for the parameters of an algorithm is a challenging, important, and time consuming task. While typically parameters are tuned by hand, recent studies have shown that automatic tuning procedures can effectively handle this task and often find better parameter settings. F-Race has been proposed specifically for this purpose and it has proven to be very effective in a number of cases. F-Race is a racing algorithm that starts by considering a number of candidate parameter settings and eliminates inferior ones as soon as enough statistical evidence arises against them. In this paper, we propose two modifications to the usual way of applying F-Race that on the one hand, make it suitable for tuning tasks with a very large number of initial candidate parameter settings and, on the other hand, allow a significant reduction of the number of function evaluations without any major loss in solution quality. We evaluate the proposed modifications on a number of stochastic local search algorithms and we show their effectiveness.

An effective hybrid algorithm for university course timetabling

The university course timetabling problem is an optimisation problem in which a set of events has to be scheduled in timeslots and located in suitable rooms. Recently, a set of benchmark instances was introduced and used for an ‘International Timetabling Competition’ to which 24 algorithms were submitted by various research groups active in the field of timetabling. We describe and analyse a hybrid metaheuristic algorithm which was developed under the very same rules and deadlines imposed by the competition and outperformed the official winner. It combines various construction heuristics, tabu search, variable neighbourhood descent and simulated annealing. Due to the complexity of developing hybrid metaheuristics, we strongly relied on an experimental methodology for configuring the algorithms as well as for choosing proper parameter settings. In particular, we used racing procedures that allow an automatic or semi-automatic configuration of algorithms with a good save in time. Our successful example shows that the systematic design of hybrid algorithms through an experimental methodology leads to high performing algorithms for hard combinatorial optimisation problems.

Hybrid Metaheuristics for the Vehicle Routing Problem with Stochastic Demands

This article analyzes the performance of metaheuristics on the vehicle routing problem with stochastic demands (VRPSD). The problem is known to have a computationally demanding objective function, which could turn to be infeasible when large instances are considered. Fast approximations of the objective function are therefore appealing because they would allow for an extended exploration of the search space. We explore the hybridization of the metaheuristic by means of two objective functions which are surrogate measures of the exact solution quality. Particularly helpful for some metaheuristics is the objective function derived from the traveling salesman problem (TSP), a closely related problem. In the light of this observation, we analyze possible extensions of the metaheuristics which take the hybridized solution approach VRPSD-TSP even further and report about experimental results on different types of instances. We show that, for the instances tested, two hybridized versions of iterated local search and evolutionary algorithm attain better solutions than state-of-the-art algorithms.

Parallel ant colony optimization for the traveling salesman problem

There are two reasons for parallelizing a metaheuristic if one is interested in performance: (i) given a fixed time to search, the aim is to increase the quality of the solutions found in that time; (ii) given a fixed solution quality, the aim is to reduce the time needed to find a solution not worse than that quality. In this article, we study the impact of communication when we parallelize a high-performing ant colony optimization (ACO) algorithm for the traveling salesman problem using message passing libraries. In particular, we examine synchronous and asynchronous communications on different interconnection topologies. We find that the simplest way of parallelizing the ACO algorithms, based on parallel independent runs, is surprisingly effective; we give some reasons as to why this is the case.

Parameter Adaptation in Ant Colony Optimization

This chapter reviews the approaches that have been studied for the online adaptation of the parameters of ant colony optimization (ACO) algorithms, that is, the variation of parameter settings while solving an instance of a problem. We classify these approaches according to the main classes of online parameter-adaptation techniques. One conclusion of this review is that the available approaches do not exploit an in-depth understanding of the effect of individual parameters on the behavior of ACO algorithms. Therefore, this chapter also presents results of an empirical study of the solution quality over computation time for Ant Colony System and MAX-MIN Ant System, two well-known ACO algorithms. The first part of this study provides insights on the behaviour of the algorithms in dependence of fixed parameter settings. One conclusion is that the best fixed parameter settings of MAX-MIN Ant System depend strongly on the available computation time. The second part of the study uses these insights to propose simple, pre-scheduled parameter variations. Our experimental results show that such pre-scheduled parameter variations can dramatically improve the anytime performance of MAX-MIN Ant System.