Multisensor-multitarget data association for tracking

Date of Completion

January 1994


Engineering, Electronics and Electrical|Operations Research




In this dissertation we develop an algorithm to solve the multisensor multi-target state estimation problem. The problem is to associate the measurements from one or more, possibly heterogeneous, sensors to identify the real targets, and to estimate their positions at any given time. Such problems arise in surveillance and tracking systems estimating the states (target positions and velocities) of an unknown number of targets. The sensors could be active (3D radars measuring full target position; or 2D radars measuring target azimuth and slant range) or passive (Electro-optic sensors measuring the azimuth and elevation angles of the source; or high frequency direction finders measuring only the azimuth angles of the source). The sensors could be stationary or mounted on a moving platform, such as a low orbiting satellite. The targets may be in motion, but are assumed to be non-maneuvering. The source of a detection can be either a real target, in which case the measurement is the true observed variable of the target plus measurement noise, or a spurious one, i.e., a false alarm. In addition, the sensors may have nonunity detection probabilities.^ The central problem in a multisensor-multitarget state estimation problem is that of data association--the problem of determining from which target, if any, a particular measurement originated. The data-association problem is formulated as a generalized S-dimensional (S-D) assignment problem, which is NP-hard for 3 or more sensor scans (S $\ge$ 3). In this dissertation, we present an efficient and recursive generalized S-D assignment algorithm (S $\ge$ 3) employing a successive Lagrangian relaxation technique, with application to a wide variety of scenarios. An important feature of this algorithm is that it improves its solution iteratively, but all the intermediate solutions are feasible. Thus, one may execute this algorithm up to some pre-specified deadline and obtain the best solution at termination. A second useful feature is that it also provides a measure of how close the current solution is from the (perhaps unknowable) optimal solution. A sliding window version of the algorithm is also presented, with application to a simple multisensor multitarget tracking example. ^