Probabilistic Data Association

Over the recent past, significant attention has been focused on the use of multiple sensors for target tracking over a large geographic area, as using a single sensor with a very large range is highly impractical. In addition, data from multiple sensors can be combined to provide improved estimation. Such Multi Sensor Data Fusion (MSDF) is similar to how humans and animals use multiple senses for perception. Probabilistic Data Association algorithms for Multiple Target Tracking (MTT), such as PDA and JPDA, use the statistical knowledge of the target dynamics and sensor measurement to obtain improved tracking. In this work, Probabilistic Data Association filters (i.e., both PDA and JPDA filters) for Complete Object Refinement have been designed, developed and studied. The designed PDA algorithms have been developed using a Track Oriented approach. For JPDA, association matrices are determined at each scan based on the current targets and measurements; from this the set of event matrices are formed which is observed to be the most computationally intensive part of the JPDA. A radar simulator for generating synthetic 3D target measurements along with false measurements is developed. The system has been developed using an Object Oriented Development scheme and implemented as a plug-in architecture in which new features can be added. The developed PDA and JPDA filters are extensively tested using synthetic track measurement data produced using the radar simulator. It was observed from the test results that PDA has lesser computation load in comparison to JPDA, but fails in the case of multiple interfering targets in a dense clutter. JPDA is successful in tracking such interfering targets due its joint association between targets and measurements. Due to JPDA's computation load only about 10 targets could be tracked, hence, clustering has to be used to decrease the computational load. PDA on the other hand is able to track more than 150 targets per scan. The work carried out can be used as a foundation for further development of target tracking systems with complex tracking needs.

In many radar or sonar tracking systems, where the state of interest typically includes target position and velocity components, target Doppler measurements may be available in addition to the target position measurements. Using measurements of Doppler information directly in target trajectory estimation leads to a nonlinear filter implementation. In this paper we investigate the methods of using Doppler measurements to improve tracking performance while maintaining the tracker implementation in the linear domain. Two new methods, namely, the one point track initialisation and the Doppler data association are proposed. By applying these two methods on the linear multi-target integrated probabilistic data association algorithm (LMIPDA), we improve tracking performance in terms of reduced number of false tracks and computational load. The merits of the proposed approaches are demonstrated via an active sonar underwater multi-target tracking scenario.

A two-dimensional target can be effectively and efficiently tracked from angle-only measurements only if the observing platform is sufficiently maneuvering. Covert passive sonar TMA problem is a common application, and, for that, a two-leg platform pattern is appealing both in observability and simplicity. Now, suppose communication relating to the platform's estimate of the target is intercepted by the target: Can the target therefrom infer where the (covert) platform must have been? This paper both answers the question and gives algorithms to solve it.

Modern radar systems have considerable flexibility in their modes of operation. In particular, it is possible to modify the waveform on a pulse to pulse basis, and electronically steered phased arrays can quickly point the radar beam in any feasible direction. Such flexibility calls for new methods of scheduling, both of the waveform and the beam direction so as to optimize the radar performance. We consider a radar system capable of rapid beam steering and of waveform switching. The transmit waveform is chosen from a small library. ...

Approximately a decade ago the maximum likelihood probabilistic data association (MLPDA) tracking architecture was proposed; it was found, via simulation, to be a very effective (perhaps the only) way to track very low-observable contacts - down to 7-8dB (post signal processing), meaning that in a large field of view a given true contact might only rank consistently among the top five or six of those obtained. The MLPDA is based on maximizing statistical likelihood according to a precise model. Its application to real multistatic data (which is unlikely to fit MLPDA assumptions) is both a test of its effectiveness and an opportunity to explore new ways to improve it. We shall see in what follows: bistatic operation; automatic registration via the "direct blast"; the use of the Savage amplitude feature; false-alarm density modification; and a smart grid-search

Abstract—This paper presents sensor and data rate control algo-rithms for tracking maneuvering targets. The manuevering target is modeled as a jump Markov linear system. We present novel extensions of the Interacting Multiple Model (IMM), Particle filter tacker, and Probabilistic Data Association (PDA) algorithms to handle sensor and data rate control. Numerical studies illustrate the performance of these sensor and data rate control algorithms. Index Terms—IMM algorithm, particle filters, PDA algorithm, sensor selection, target tracking.

This paper describes a multisensor single target tracking simulator “MUST” developed at CSSIP. MUST is based on a multisensor extended Kalman filter (EKF) which can handle asynchronous nonlinear multiple measurements of target parameters such as range, bearing, range rate and elevation angle. Multiple measurements from each sensor are handled by the probabilistic data association (PDA) filter. MUST provides a flexible platform where the performance of the EKF-PDA tracker can be assessed for different sensor, target and environment scenarios. This paper also presents some simulation results for different tracking scenarios and gives an initialization algorithm for the EKF-PDA algorithm

The probability hypothesis density (PHD) filter is a practical alternative to the optimal Bayesian multi-target filter based on finite set statistics. It propagates the PHD function, a first order moment of the full multi-target posterior density. The peaks of the PHD function give estimates of the target states. However, the PHD filter keeps no record of target identities and hence

In tracking a single target in clutter, many algorithms have been developed ranging in complexity from nearest neighbor (NN) and probabilistic data association (PDA) to the optimal Bayesian filter. In multiple-target tracking, a number of the techniques have been exercised such as the JPDA and the multiple hypothesis (MHT) schemes. Sub-optimal algorithms, such as the PDA filter, have been used

Most target tracking algorithms implicitly assume that target exists. There are only a few techniques that address the target existence problem along with target tracking. For example, (Integrated Probabilistic Data Association) IPDA filter addresses the target tracking and target existence problems simultaneously and it does so under at most one target assumption. In recent times random sets have been proposed as a general framework for multiple target tracking problem. However, its relationship to well understood existing tracking algorithms like IPDA has not been explored. In this paper, we show that under appropriate conditions random sets provide appropriate mathematical framework for solving the joint target existence and state estimation problem and subsequently show that it results in IPDA under appropriate simplifying assumptions.