Counter-UAV radar: system design, signal processing and machine learning

Unmanned aerial vehicles (UAVs) are growing in popularity across civilian and defense uses. However, it is becoming necessary to deploy sensors to monitor the use of UAVs to ensure that they are used safely. Such sensors struggle to detect and classify UAVs at long distances.

Radars are good candidate for this application. Their all-weather ability makes them good candidate for small objects detection in all environmental and lighting conditions. The combination of high range and angular resolution with the ability to measure the Doppler effect provides accurate information about targets and the surrounding environment.

A major challenge in the detection of small drones is the combination of low reflectivity, high required detection range and short capture time due to their high velocity—difficulties that are further amplified in beyond visual line of sight (BVLOS) scenarios, where the drone cannot be seen by vision.. These characteristics complicate radar detection and tracking algorithms. Part of those challenges can be solved by using heavy, power hungry ground-based radar. But the maximum detection range of those radars is limited for low altitude objects (such as UAVs) as the relief can easily block the view. A solution is then to mount  miniature radars in other drones. This offers better view of the environment and can cover a large area, particularly in BVLOS operations, as drones are easy to deploy. In that configuration, we are limited in weight and power consumption. The student will propose radar modulation techniques and algorithms to allow long range detection with limited weight and power consumption. Multi-radar fusion reconstruction algorithms and multi-radar synchronization techniques will also be proposed to combine multiple radar measurements in a single high-resolution representation of the environment. In addition, the student will investigate implementation challenges such as synchronization between radars and impact/mitigation of hardware non-idealities. Finally, machine learning-based algorithm will also be investigated to classify the detected targets and improve radar performances.

The successful PhD candidate will be part of a large imec team working on the research, implementation and prototyping of future radar systems, composed of experts in digital and analog, radar and wireless communication systems, signal processing and machine learning algorithms. This is a unique opportunity to actively contribute and develop breakthrough technology and build-up future radar sensors.