Tracking & Data Fusion

Fundamental Tracking

Research collaboration on fundamental issues in radar tracking implementing particle filtering techniques in order to approximate the optimal filtering solution, which is known to be computationally infeasible. Particle filters appear to provide a more computationally reliable manner to approach this problem. Successful implementation of particle filtering techniques that appear to be able to compete with the best approximate solutions known to date have been realised.

Tracking and Data Fusion Laboratory (TDFL)

The Tracking and Data Fusion Laboratory is a joint research collaboration between DSTO - Australian Defence Science & Technology Organisation and CSSIP. This collaboration considers research and development of advanced algorithms and architectures for distributed surveillance systems, specially ones arising in multi-sensor multi-target tracking and identification. The problem space includes unwanted measurements (clutter) and unreliable and noisy target measurements (detections), dissimilar sensors with unreliable orientation information, and unreliable communications between sensors. The desired outcome is estimating the number and trajectories of targets, as well as their identity in this environment.

The TDFL Distributed Surveillance Systems Project has a number of sub-projects as follows:

Title: Distributed Sensor Systems

Description: This research examines the relationship between fusion architectures, fusion algorithms and system performance in distributed surveillance networks, focusing on the data incest problem and surveillance picture consistency.

Title: Distributed Sensor Fusion with Network Constraints

Description: This research accommodates restrictive communication channels between distributed sensors. In real life communications, records can be lost or received out of sequence. Communication channels also have a limited and time variable bandwidth availability. This research concentrates on the target state estimation when sensor data is received out of sequence and with limited bandwidth.

Title: Tracking a Group of Targets

Description: This research investigates the problem of tracking a group of targets. A formation of targets poses a number of problems to the tracking system, including, and not limited to, computational complexity, track divergence and coalescence problem, as well as erratic tracking. Research in this year has concentrated on solving the computational complexity problem, which will immediately lift the capacity of modern tracking systems.

Title: Sensor Registration

Description: Sensor registration is a problem of unknown orientation and measurement biases of sensors. Building on previous results of simultaneous tracking and registration, this research investigates the problem of bias observability in multi sensor tracking systems, and aims to establish fundamental limits of registration and track estimation.

Title: Performance Assessment of Distributed Multi Sensor Data Fusion Systems

Description: Modern target tracking algorithms are described using various probabilistic instruments. This research aims to unify various measures of uncertainty employed by different probabilistic instruments.

Situation Awareness

This project is to carry out fundamental research of situation awareness and its applications to air defence and exploits the ideas of Bayesian inference methods.

Multisensor Integration [PhD project]

The Research Project is to carry out research into Multi Sensor Integration in relation to an Airborne Early Warning and Control application. Research involves data incest problems and how to overcome data incest problems in sensor networks.