PhD position Applied Signal Processing & Machine Learning

The focus of this position is on the development of hybrid AI models for automotive radar data.

This means that you will conduct state-of-the art research at the interface between classical physics-based signal processing and data-driven machine learning and publish at high ranking conferences and journals.

The project itself aims to develop an AI-aided radar system that is able to fuse radar information on a symbolic level.

During the project we will focus on the following topics concerning:

• Enhancing performance at sensor-level: Develop multi-static radar signal models that incorporate angular, range (delay), and velocity (Doppler) parameters for coherent data fusion in distributed automotive radar sensors.
• Design machine learning enhanced radar preprocessing methods incorporating calibrated uncertainties.
• Reliability by higher-level integration: Merge classical Bayesian graphical models with AI components to enable scalable algorithms that account for uncertainties in a fully probabilistic manner and are able to fuse data at a higher level to improve system reliability in complex scenarios.
• Situation-aware sensor configuration: Develop active adaptation algorithms for next generation automotive radar systems guided by higher-level information.

This will allow the system to concentrate on regions of high measurement uncertainty or critical relevance.

The hybrid nature of these ML-enhanced, model-based algorithms naturally accommodates uncertainty and provides a high degree of flexibility, crucial for safety-critical applications like automotive radar.

In collaboration with Infineon, the opportunity is given to evaluate the developed hybrid methods on real automotive radar sensor architectures.