Master thesis - Robust 4D Radar Perception for Autonomous Driving: Detect and Classify Multipath

30 credits - Robust 4D Radar Perception for Autonomous Driving: Detecting and Classifying Multipath Effects 

Background 

4D radar is becoming an increasingly important sensing modality in autonomous driving. New generations of radar offer better range and finer spatial resolution, making radar useful not only for geometric scene understanding, but also for motion reasoning through Doppler measurements. 

At the same time, 4D radar data can contain significant multipath artifacts. These effects can produce false detections with incorrect position, incorrect Doppler, or both, often appearing as ghost objects. Such errors can degrade downstream perception and motion estimation if not handled robustly. 

This creates an important research opportunity: developing methods that identify and suppress multipath-induced false positives while preserving true dynamic and static targets. 

Objective 

The objective of this thesis is to investigate methods for classifying and removing false 4D radar returns caused by multipath propagation in autonomous driving scenarios. The final research questions will be defined together with the student based on literature, available data, and project direction. Possible directions include: 

• Characterizing multipath failure modes in modern 4D radar data. 

• Designing models or feature pipelines to distinguish valid returns from ghosts. 

• Leveraging spatial, Doppler and temporal cues for robust filtering. 

• Comparing classical signal-processing and machine-learning approaches. 

• Evaluating the impact of filtering on downstream tasks such as odometry, motion estimation, or occupancy prediction. 

The Project Offers 

• Work on a high-impact topic at the intersection of radar sensing and autonomous driving. 

• Access to real-world driving data and computational resources. 

• Close supervision and collaboration in an active research environment. 

• Opportunity to contribute methods that can improve robustness of future AD systems. 

• Potential for scientific publication depending on outcomes. 

Who are we looking for? 

We are looking for one or two motivated Master’s students in Computer Science, Electrical Engineering, Engineering Physics, Robotics, Applied Mathematics, or related fields. 

Experience in one or more of the following is beneficial: 

• Machine learning or deep learning 

• Signal processing 

• Computer vision or multimodal perception 

• Probabilistic modeling and data analysis 

• Python and PyTorch 

• Autonomous driving, robotics, or sensor fusion 
  
  

The planned thesis start is January 2027. 

Number of students: 1 

Start date for the thesis work: [To be agreed] 

Estimated time required: 20 weeks, full time (30 credits) 

Contact persons and supervisors 

Ajinkya Khoche, ajinkya.khoche@scania.com; Jonny Andersson, jonny.andersson@scania.com 

Hiring Manager: Magnus Granström, magnus.granstrom@scania.com 

Application 

Your application must include a CV, personal letter, and transcript of grades. 

A background check might be conducted for this position. We are conducting interviews continuously and may close the recruitment earlier than the date specified. 

References 

1. Robometer: Scaling General-Purpose Robotic Reward Models via Trajectory Comparisons