Target Tracking

Course Description

This course is going to be about basic target tracking theory and techniques including

• Track initiation, maintanence and deletion
• Maneuvering target tracking
• Multiple model filtering
• Interacting multiple model (IMM) filter
• Multi target tracking
• Global nearest neighbors (GNN)
• (Joint) probabilistic data association (JPDA)
• Multiple hypothesis tracking (MHT)
• Multi sensor target tracking
• Multi sensor architectures
• Track association and fusion

In addition to these, the last lecture of the course will include an overwiew of some unconventional methods that have gained popularity during the last years in the tracking community. These might include

• Random set based approaches (PHD, CPHD)
• Probabilistic multiple hypothesis tracking (PMHT)
• Track before detect (TBD)

This is a shorter scale course than most of other graduate courses in Reglerteknik and will involve 6 hp.

Prerequisites

The engineering probability theory is essential. We are going to use (extended) Kalman filters (or alternative Bayesian filters) as subblocks in the algorithms. Therefore, the sensor fusion course or basic Kalman filter knowledge is necessary. For the computer exercises, a fair knowledge of Matlab is required.

Literature

There is unfortunately no single book suitable for all of our purposes. The following book is now a famous one that covers the most of the material relavant to the course.

• S. Blackman and R. Popoli, Design and Analysis of Modern Tracking Systems, Artech House, Norwood MA, 1999.

When we need deeper information about the covered subjects, the additional material will be distributed in the class.

Organization

The course is going to be composed of lectures and related computer exercises. There is no written exam for the course and instead students will submit the reports of their their solutions to the exercises.

Lectures

The standard two hour lectures will more or less follow the topic order given above. The program is given in the lectures page.

Computer Exercises

The course is going to include extensive computer exercises which will involve the implementation of the algorithms covered in class on simplified examples. A tentative plan of the exercises is given in the exercises page.

Project (optional)

For people who would like to do more and get more points (3hp), there is an opportunity to do a project which can take place after the course is finished. More information is given in the project page.

What is not covered

The tracking literature is vast. It is not possible to cover all different models and methodologies. So this course (except the last lecture) is going to be limited to the following basic assumptions:

• Assumption 1: At a single time, a target can result in at most one measurement
• Assumption 2: A measurement can originate from at most a single target or clutter.

The first assumption is basically a point target assumption (compared to the sensor resolution). Hence we exclude the so-called extended targets. The second assumption boils down to assuming that the targets are separated enough (compared to the sensor resolution) to yield individual measurements. Hence we exclude the so-called unresolved targets. This might look a little disappointing but the material covered in the course actually forms the foundations of the methods proposed for the cases where these assumptions are not valid.

Contact Person

Umut Orguner (tel: 013-281338, email: umut_at_isy.liu.se).