Target Tracking with MIMO Radar in the Presence of Multipath

Multiple-input-multiple-output (MIMO) radars are widely used for target localization in many different fields such as airborne, remote-sensing, and autonomous driving systems. Recently,  they are also becoming popular in indoor environments for human target tracking and activity recognition. For both automotive and indoor MIMO radar, the complex urban/indoor environment leads to multipath reflections from static objects such as buildings and walls. The detections from the multipath are called ‘ghosts’, which generate objects in the radar detection list that, in fact, do not exist. In the state-of-the-art, the problem of removing multipath ghosts in MIMO radars remains unsolved. Even though there have been some solutions based on prior knowledge of the reflector position (geometry model), which are called ray-tracing techniques, to analyze or eliminate multipath, it is not a realistic assumption that this multipath geometry information is known a priori.

The goal of this master thesis is to develop algorithms to track targets with a MIMO radar in a multipath environment and jointly estimate the multipath reflectors (e.g. wall position and room boundary). In the previous research of our team, we have proposed algorithms to recognize multipath without prior knowledge of multipath geometry in an indoor environment. The master project will be based on the model and measurements that have already been achieved completed.

The task of the student will be to:

1. study the basic MIMO radar signal processing
2. understand the multipath model of MIMO radar
3. study the theory of tracking, RFS and PHD filter
4. simulate the joint target tracking and multipath estimation model for multi-targets and multi-walls with a PHD filter
5. verify the algorithm by the real measurement.
6. (depending on the ambition level) extend the multipath model to a moving platform.

The student should have basic knowledge of signal processing, detection theory, probability theory and statistics. The project will be implemented in MATLAB and Python.