Topics in Multisensor Maneuvering Target Tracking
Type of DegreeDissertation
Electrical and Computer Engineering
MetadataShow full item record
Tracking uses models of the real environment to estimate the past and present and even predict the future state of a moving object from noisy observations of uncertain origin. In a tracking scenario the most critical problem is that of data-association. This topic has received considerable attention in the literature and a number of solutions have been proposed. This dissertation considers the problem of tracking highly maneuvering target(s) using multiple sensors in the presence of clutter. A set of noble algorithms are developed to handle this problem. First, the basic interacting multiple model (IMM) approach has been combined with probabilistic data association (PDA) to develop an IMMPDA (interacting multiple model probabilistic data association) algorithm with simultaneous measurement update (SMU) for tracking a maneuvering target in clutter with multiple sensors. Second, we extend our noble SMU algorithm to a more practical tracking scenario, that of tracking a maneuvering target with asynchronous (in-sequence but time delayed) measurements. A state-augmented approach is developed to estimate the time delay between a local sensor (assumed to be collocated and synchronized with a central processor) and a remote sensor (assumed to be separately located and not synchronized with a central processor). Third, we address one of the most important issues for target tracking in a multisensor fusion network: out-of-sequence measurements (OOSM). However, this dissertation is not concerned with different sampling rate among sensors. Instead, we focus on a suboptimal filtering algorithm dealing with possibly time delayed, out-of-sequence measurements (OOSM) with a fixed relative time-delay (we assume that sampling rate are all the same for all sensors) among sensor measurements. A state-augmented approach is also developed to improve tracking performance with the possible presence of OOSM. The filtering algorithm is developed by OOSM updating with IMMPDA for the target. Finally, we consider tracking of multiple highly maneuvering targets using multiple sensors with possibly unresolved measurement. When multiple targets move temporarily in close formation, it possibly gives rise to a single detection due to the resolution limitations of the sensor. Assuming that there are possibly unresolved measurements from at least two targets (i.e., measurement association with more than two targets simultaneously), any measurement therefore is either associated with a target, a group of merged targets, or caused by clutter. The filtering algorithm is developed by applying the basic IMM approach and the joint probabilistic data association with merged measurements (JPDAM) technique and coupled target state estimation.