Thesis work: 30 hp - Context-Aware Prediction of Pedestrians Using Transformer-based Models

30 credits – Context-Aware Prediction of Pedestrians in Urban Scenarios Using Transformer-based Models

Traton R&D develops core technologies for the global brands Scania, MAN, and International. A master’s thesis with us is a great way to build your network and work at the forefront of advanced driver-assistance systems (ADAS) and autonomous driving. Many of our colleagues began their careers here through a thesis project.

Background:

Accurately estimating the state and tracking the movements of pedestrians is critical for ADAS and autonomous vehicles. Robust perception enables a vehicle to anticipate and respond to safety-critical situations, improving protection for drivers, passengers, and other road users.

Predicting pedestrian intention and behavior in complex urban scenarios remains challenging due to the many degrees of freedom in human motion. Pedestrians share space with vehicles and may intersect vehicle paths in varied and often abrupt ways. Traditional prediction techniques that often rely on physical states (e.g. position, velocity) can be unreliable when future motion depends on environmental and contextual cues, such as traffic signals, crosswalks, group behavior or sidewalk geometry. Incorporating context-aware predictions offers the potential to significantly improve the safety of pedestrians in real traffic.

Target:

Develop a transformer-based framework that can reliably predict pedestrian intentions based on contextual awareness in simple traffic scenarios. The framework should be capable of producing real-time predictions that can integrate with onboard decision-making module in ADAS and autonomous vehicles.

Assignment:

The assignment is divided into these sub-tasks:

1. Investigate methods for predicting pedestrians’ intentions in urban scenarios.

2. Design and implement a framework.

3. Train the framework on datasets.

4. Test the framework in simulation and/or in real experiments using research and prototype vehicles.

Education:

Master (civilingenjör) in computer science, robotics, engineering physics, electrical engineering, or applied mathematics, preferably with specialization in applied estimation, control theory, optimization, and machine learning. Knowledge of reachability analysis, system modeling and programming are a plus.
Number of students: 1-2
Start date: January 2026
Estimated time needed: 20 weeks

Contact persons and supervisors:
Jacob Wallersköld, ADAS developer at Automated Driving Department,

jacob.wallerskold@scania.com, +46 76-53 15 28

Vandana Narri, Industrial PhD student in Post-Perception for Connected and Automated Vehicle,

vandana.narri@scania.com , 08 – 553 822 97

Application:

Enclose CV, cover letter and transcript of records.

A background check may be conducted for this position. We conduct interviews on an ongoing basis and may close recruitment earlier than the stated date.