Sensor placement optimization for distributed acoustic source localization system using semidefinite programming

Abstract

The sensor-source geometry is one of the key factors affecting the performance of a distributed acoustic source localization system. In this paper, an approximate optimal routing method is proposed by minimizing the determinant of the Cramer Rao Lower Bound (CRLB) for the source location estimate. At first the sensors are equally distributed in the test field, the problem is relaxed to a combinatorial optimization problem, which can be solved by semidefinite programming (SDP) [9]. We sort the sensors by the λ∗ and then select the highest probability of the occurrence for the globally optimal placement. The simulation compares the performances of the TDOA-based acoustic source localization system by using the different placement schemes obtained from the adaptive genetic algorithm (AGA) and the SDP, respectively. The results show that the localization performance using SDP-based scheme is better than that using AGA-placement scheme when the probability of uncertain source location follows uniform distribution.