Projection-Based Constrained Fusion Performance with Link Loss and Measurement Bias

Abstract

Sensor networks have been deployed in many real-world applications for tracking and monitoring of dynamic targets over large geographical areas. We consider a scenario where a number of ground and/or airborne sensors are used to jointly track a ground target with certain constraints on its motion dynamics, such as those defined by roadways. The state estimates from these sensors are sent over a lossy communications network to a remote fusion center for fusion. We propose both centralized and distributed projection-based methods to incorporate known constraints into the estimation and fusion process. In addition to presenting simple closed-form projected fusion solutions, we also show how to utilize knowledge of the constraints into the fusion process by two distance-based weighting methods. We also explore the effect of link loss and sensor measurement bias on tracking accuracy performance of constrained fusion using illustrative tracking examples. Performance evaluation results demonstrate the potential of adopting weighted average fusers in improving the overall tracking performance in the presence of link loss and measurement bias.