The value of delayed information in tracking with distributed sensor networks

Abstract

Tracking the movement of objects using multiple sensors (such as tracking vehicles or people using a distributed set of video cameras) is a topic of importance in security and automation applications. A prime consideration in these multi-sensor tracking networks is the timeliness of estimates, as tracking is a time-sensitive task. Current methods for handling delayed observations can be computationally intensive while minimally improving the estimate. This work aims to develop methods for estimating the value of delayed information. The value of observations from a variety of sensor locations at time points in the past are evaluated using the proposed strategy. We derive the relationships and trends that relate the 'age' of past sensor data to its marginal contribution toward improving the accuracy of an estimate of the current location of an object. Low-complexity methods for assessing the value of delayed sensor information can impact applications in defense and national security, personal electronics, and financial engineering where timely decisions are paramount.