Decorrelated state estimation for distributed tracking in cluttered environments.

摘要

In this dissertation, a tracking algorithm based on decorrelated state estimation is developed for distributed fusion architectures in cluttered environments. The decorrelation approach is extended such that it can be incorporated with the sequential Multi-Sensor Joint Probabilistic Data Association (MSJPDA) algorithm. In particular, these extensions include the existence of clutter, interacting targets, multiple sensors, and feedback. The decorrelation approach constructs decorrelated state sequences which can be used as measurements at the global processor. Because the decorrelation approach addresses the correlation problem prior to track fusion, many well-known centralized processing algorithms can be used at the global processor of distributed tracking. The implementation at the global processor is addressed including the roles of decorrelated state sequences in the filtering process. To demonstrate the tracking performance (such as root mean square error of the estimated target state and track lifetime) of these proposed algorithms, Monte-Carlo simulations are used. The computational complexity and communication requirements of these algorithms are evaluated to demonstrate tracking performance tradeoffs between various distributed fusion architecture configurations.