Looking Ahead with the Pilot Method

The pilot method as a meta-heuristic is a tempered greedy method aimed at obtaining better solutions while avoiding the greedy trap by looking ahead for each possible choice. Repeatedly a master solution is modified; each time in a minimal fashion to account for best choices, where choices are judged by means of a separate heuristic result, the pilot solution. The pilot method may be seen as a meta-heuristic enhancing the quality of (any) heuristic in a system for heuristic repetition. Experiments show that the pilot method as well as similar methods can behave quite competitively in comparison with well-known and accepted meta-heuristics. In this paper we review some less known results. As a higher time complexity is usually associated with repetition, we investigate a simple short-cut policy to reduce the running times, while retaining an enhanced solution quality. Furthermore, we report successful experiments that incorporate a distinguishing feature of the pilot method, which is the extension of neighborhoods into “local” search, creating tabu search hybrids.     

A Hybrid Meta-Heuristic for the Batching Problem in Just-In-Time Flow Shops

This paper is concerned with a batching problem encountered in the context of production smoothing in just-in-time manufacturing systems. The manufacturing system of interest is a multi-level system with a flow-shop at the final level. We develop a hybrid meta-heuristic method to solve the batching problem, which is known to be NP-hard. We hybridize strategic oscillation (SO) and path re-linking (PR) methods and compare the hybrid method's performance to two benchmark methods: a bounded dynamic programming method developed for the problem earlier and an implementation of robust tabu search (RTS) meta-heuristic. Through a computational study, we show that the proposed hybrid method is effective in solving the problem within several minutes of computer time and yielding near-optimal results.     

A Tabu Search Algorithm for the Quadratic Assignment Problem

Tabu search approach based algorithms are among the widest applied to various combinatorial optimization problems. In this paper, we propose a new version of the tabu search algorithm for the well-known problem, the quadratic assignment problem (QAP). One of the most important features of our tabu search implementation is an efficient use of mutations applied to the best solutions found so far. We tested this approach on a number of instances from the library of the QAP instances—QAPLIB. The results obtained from the experiments show that the proposed algorithm belongs to the most efficient heuristics for the QAP. The high efficiency of this algorithm is also demonstrated by the fact that the new best known solutions were found for several QAP instances.     

Rescaled Simulated Annealing—Accelerating Convergence of Simulated Annealing by Rescaling the States Energies

This paper presents a new metaheuristic, called rescaled simulated annealing (RSA) which is particularly adapted to combinatorial problems where the available computational effort to solve it is limited. Asymptotic convergence on optimal solutions is established and the results are favorably compared to the famous ones due to Mitra, Romeo, and Sangiovanni-Vincentelli (Mitra, Romeo, and Sangiovanni-Vincentelli. (1986). Adv. Appl. Prob. 18, 747–771.) for simulated annealing (SA). It is based on a generalization of the Metropolis procedure used by the SA algorithm. This generalization consists in rescaling the energies of the states candidate for a transition, before applying the Metropolis criterion. The direct consequence is an acceleration of convergence, by avoiding dives and escapes from high energy local minima. Thus, practically speaking, less transitions need to be tested with RSA to obtain a good quality solution. As a corollary, within a limited computational effort, RSA provides better quality solutions than SA and the gain of performance of RSA versus SA is all the more important since the available computational effort is reduced. An illustrative example is detailed on an instance of the Traveling Salesman Problem.