ABSTRACT
In this chapter, the robust sensor location estimation problem is considered as a significant issue faced by several practical applications in wireless sensor networks (WSNs). In general, sensors are classified into two categories: location-aware and location-unaware sensors. For the position estimation of location-unaware sensors, exact positions are often assumed precisely available for location-aware sensors. However, in practice, such precise data may not be available. Therefore, in this chapter, how to determine the positions of location-unaware sensors in the presence of inexact positions of location-aware sensors is the primary focus. A robust min-max quadratic game method is proposed for the sensor location estimation problem by minimizing the worst-case estimation error from the two-person min-max l2 game perspective. The corresponding optimization problem is originally nonconvex, but after it is transformed into a convex semidefinite program (SDP), it can be solved by existing numerical techniques. In the case of inexact positions of location-aware sensors, the estimation robustness of the proposed min-max l2 game approach is validated by simulations under different WSN topologies. For the performance validation, the modified maximum-likelihood (ML) estimation and second-order cone programming (SOCP) relaxation methods are also used for the relative location estimation in comparison with the proposed min-max l2 game approach.
