Sensor Management for Large-Scale MultisensorMultitarget Tracking

Abstract

<p>In this thesis we consider the problem of managing an array of sensors in order to track multiple targets in the presence of clutter in centralized, distributed and decentralized architectures. As a result of recent technological advances, a large number of sensors can be deployed and used for multitarget tracking purposes. Even though a large number of sensors are available, due to frequency, power and other physical limitations, only a few of them can be used at anyone time. The problem is then to select sensor subsets that should be used by fusion centers at each measurement time step in order to optimize the tracking performance subject to their operational constraints.</p> <p>In general, sensor management is performed based on the predicted tracking performance at the future time steps. In this thesis, the Posterior Cntmer-Rao lower bound (PCRLB), which provides a measure of the optimal achievable accuracy of target state estimation, is used as the performance measure. We derive the multitarget PCRLB and show the existence of a multitarget information reduction matrix (IRM), which can be calculated off-line in most cases. First, the sensor subset selection problem for centralized architecture is considered for two different scenarios: (i) fixed and known llumber of targets; (ii) varying number of targets. Then, in the distributed architecture, in addition to assigning sensor subsets to local fusion centers (LFCs), the transmission frequencies and powers of active sensors need to be assigned. In this thesis, we assume that the transmission power of the sensors will be software controllable within certain lower and upper limits. Finally, we consider the decentralized architecture in which there is no central fusion center (CFC), each fusion center (FC) communicates only with the neighboring FCs, and communications are restricted. In this case, each FC has to decide which sensors should to be used by itself at each measurement time step by considering which sensors may be used by neighboring FCs.</p> <p>We give the optimal formulations for all of the above problems. Finding the optimal solutions to the above problems in real time is very hard in large scale scenarios. We present algorithms to find suboptimal solutions in real time. Simulation results illustrate the performance of the algorithms, both in terms of their real-time capability for large scale problems and the resulting estimation accuracy.</p>