Member
About/Bio
 
 
Signal Processing and Machine Learning for Millimeter-Wave Radar 3D Imaging 

Radars are being widely employed as imagers in modern vehicles towards achieving full autonomy. The main advantages of microwave imagers over optical imagers are the all-weather operation capability as well as the ambient-light independence. However, microwave imagers come inferior to optical imagers when it comes to angular resolution.

The goal of my PhD research is to develop a suite of signal processing and machine learning algorithms for enabling a high-resolution forward-looking radar imaging system. Such system should offer specifications that meet the market’s requirements (i.e., standard regulations and cost effectiveness).

For autonomous vehicles, high range, angular, and doppler resolution are required. While, better range and doppler resolution can be achieved with larger bandwidth and longer observation time, respectively, higher angular resolution is achieved with larger antenna apertures, which can be expensive or infeasible. To circumvent this problem, the concept of synthetic aperture radar (SAR), along with sparse antenna arrays, can be used instead of large fully populated (dense) arrays.

"Greatness is a lot of small things done well."

The deployment of a high-resolution forward-looking radar imager, that exploits a sparse array-based SAR, in an autonomous vehicle poses the following challenges:

1) Adaptation of the image reconstruction algorithms for forward-looking radars:
     a. Robust image reconstruction algorithms are needed to address the unpredictable maneuvers of the platform.
     b. Fast and computationally efficient algorithms are needed for enabling online applications.

2) Sub-Nyquist and non-uniform sampling:
     a. Aliasing and phase error in the formed image due to the sparsity of the antenna array.
     b. High angular sidelobes.

3) Defocusing of moving targets due to the range migration phenomenon:
     a. Moving targets’ signal separation from clutter.
     b. Motion parameter estimation.
     c. Image focusing and reconstruction of the scene.

4) Radar imaging in urban areas suffer from complicated multipath scattering.

5) Synchronization between cascaded radars (which form the long baseline aperture).

My PhD research aims at addressing mainly challenges 1 to 4, with the remaining challenge (5) kept as tentative work.

Achievements (Honors & Awards) 
  • IMEC PhD scholarship
    • [15-10-2020]
  • KU Leuven scholarship for international scholars, Faculty of Engineering Science
    • [15-9-2019]
Memberships