Reactive Tabu Search and Sensor Selection in Active Structural Acoustic Control Problems

Abstract

A Reactive Tabu Search (RTS) is examined. In addition to a dynamic tabu tenure RTS also detects when the search has entered an unproductive area and restarts RTS based on distinctive features of the unproductive area. We explore the effectiveness of RTS over a static tabu list (of a kind used in many implementations) for a two variable unconstrained discrete optimization model with 9 nearly identical minima and 513 other local minima. One of the key features of this problem is that the two-dimensional domain allows us to provide graphical descriptions of the performance of RTS. We then apply RTS to a sensor selection problem in active structural acoustic control. The objective in this problem is to select a set of 8 sensors out of 462 potential sensor locations so that the noise measured at the 8 chosen sensors is as close as possible to the noise measured at all 462. Computational experiments for data taken from a laboratory test article at NASA Langley Research Center are provided.