Antabu | Enhanced Version Lil-99-1

Many methods currently used in combinatorial optimization are inspired by adaptative natural behaviors or natural systems, e.g. hill-climbing local search, genetic algorithms, simulated annealing, : : : These heuristics provide some very good results, even when the problem size makes it impossible to use more traditional exact methods (such as branch and bound, ...). Ants system algorithms belong to this class of adaptive or evolutionnary \nature inspired" algorithms, and are based on the natural behavior of ants. These algorithms have been applied successfully to many optimisation problems like travelling salesman (TSP), job shop scheduling (JSP), graph colouring, quadratic assignment problem (QAP), : : : This paper presents ANTabu, an ants system algorithm for the QAP. We have designed ANTabu with a parallel model, thus allowing it to take advantage of the power of networks of workstations. The co-operation between simulated ants is provided by a pheromone matrix that plays the role of a global memory, with a re ned pheromone update mechanism. The exploration of the search space is guided by the evolution of pheromones level, while exploitation has been boosted by a tabu local search. Special care has also been taken in the design of a diversi cation phase, based on a frequency matrix. We give results that have been obtained on benchmarks instances from the QAPlib. We show that they compare favourably with other top of the art algorithms dedicated for the QAP.