Application of Self Controlling Software Approach To Reactive Tabu Search

Abstract

In this paper the principle of self adaptation is applied to achieve a self controlling software. The software considered in this case is a heuristic search algorithm: the reactive tabu search. In reactive search algorithms, the behavior of the algorithm is evaluated and modified during the search. To improve self adaptation, two new strategies for reactive tabu search are introduced. Tire first strategy uses a control theoretic approach, treats the algorithm as a plant to be controlled and modifies the algorithm parameters to control the intensification of the search. The second strategy adjusts several parameters according to the feedback coming from the search to achieve diversification during the search. These strategies adjust the parameters of the tabu search and form the Self Controlling Tabu Search (SC-Tabu) algorithm. The performance of the algorithm is tested on different problem types of the Quadratic Assignment Problem (QAP). The results show that the algorithm adapts successfully to achieve good performance on problems with different structures.