New Approaches to State Estimation in Sensor Network Systems with Uncertain Sensor Measurements

One of the key challenges in optimal filtering for sensor network systems is to perform estimation in uncertain conditions in addition to the measurement noise. This is typically accomplished by incorporating models that account for some of the most important uncertainties, such as sensor failure and sensor delay that affect sensor systems in a network environment. Sensor failure and delay phenomena are modeled by using multiplicative noise in the measurement equation. The existing solutions for this problem only model sensors with the same failure or delay characteristics. However, sensors in a networked environment may be connected to different network segments, and while there is heavy traffic congestion in some segments, others may have very light network traffic. So, there may be different delay characteristics. Even sensors manufactured in identical factory conditions with identical design specifications are not guaranteed to fail in the same manner. Due to these facts, sensors in a network system may not have the same characteristics and earlier models need to be generalized. Our goal in this research is to perform minimum variance optimal state estimation using system models that account for the above mentioned differences in sensor characteristics in a network environment. An adaptive scheme is also presented to address the case of unknown or time-varying sensor characteristics