A Concentration-Based Approach for Optimizing the Estimation Performance in Stochastic Sensor Selection

Abstract: In this work, we consider a sensor selection drawn at random by a sampling with replacement policy for a linear time-invariant dynamical system subject to process and measurement noise. We employ the Kalman filter to estimate the state of the system. However, the statistical properties of the filter are not deterministic due to the stochastic selection of sensors. As a consequence, we derive concentration inequalities to bound the estimation error covariance of the Kalman filter in the semi-definite sense. Concentration inequalities provide a framework for deriving semi-definite bounds that hold in a probabilistic sense. Our main contributions are three-fold. First, we develop algorithmic tools to aid in the implementation of a matrix concentration inequality. Second, we derive concentration-based bounds for three types of stochastic selections. Third, we propose a polynomial-time procedure for finding a sampling distribution that indirectly minimizes the maximum eigenvalue of the estimation error covariance. Our proposed sampling policy is also shown to empirically outperform three other sampling policies: uniform, deterministic greedy, and randomized greedy.