Sensor bias estimation in multisensor systems

Date of Completion

January 2004


Engineering, Electronics and Electrical




In this dissertation research, three topics in multisensor bias estimation are discussed: sensor bias estimation based on single-frame unassociated tracks, exact dynamic multisensor bias estimation, and multisensor bias estimation for asynchronous sensors. ^ The first part of the dissertation is on the bias estimation based on the unassociated local tracks (state estimates) at a specific time without a priori association. The likelihood function of the biases is derived and it is shown that the difference of the local estimates is the sufficient statistic for estimating the biases. A least squares solution of the bias estimates and the corresponding Cramer-Rao Lower Bound (CRLB) are presented. Two approaches in the absence of known track-to-track association, namely the Maximum Likelihood estimator combined with Probabilistic Data Association (ML-PDA) and an estimator based on soft data association, are proposed. ^ In the second part, an exact solution for the dynamic multiple sensor bias estimation problem is proposed. The sensors are assumed synchronized. It is shown that the sensor bias estimates can be obtained dynamically using the local (biased) state estimates. This is accomplished by manipulating the local state estimates such that they yield pseudomeasurements of the sensor biases with additive noises that are zero-mean, white and with easily calculated covariances. Monte Carlo simulations show that this method has significant improvement in performance with reduced RMS errors of 70% compared to the commonly used decoupled Kalman filter. ^ Finally, an exact bias estimation solution for general asynchronous sensors is proposed. A novel scheme is used to transform the measurements from the different times of the sensors into pseudomeasurements of the sensor biases. These methods, both for synchronous sensors and asynchronous sensors, allow evaluation of the CRLB for the sensor bias estimates, i.e., a quantification of the available information about the sensor biases in any scenario. Monte Carlo simulation results also show that these new methods are statistically efficient, i.e., they meet their corresponding CRLBs. The compensation of state estimates and the use of these techniques for scale biases and sensor location uncertainties in addition to the usual additive biases are also presented. ^