Learning-Based Parameter Estimation for Kalman Filters in Object Tracking for Autonomous Driving

Institut

Lehrstuhl für Fahrzeugtechnik

Typ

Masterarbeit /

Inhalt

experimentell / theoretisch /  

Beschreibung

Kalman filters are widely used in object tracking, but their performance often depends heavily on parameters that are manually set — a process that can be tedious and suboptimal. This thesis looks into learning these parameters directly from data to make the tracking system more accurate and adaptable. There are already approaches in literature available, however they have not been adapted to the domain of autonomous driving.

 

What you'll do:

• Review and analyze the leading current tracking methods used in autonomous systems – Kalman filter and learning based
• Develop and test approaches to learn model parameters (noise covariances) from a data set consisting of detected objects and the ground truth
• Compare optimized Kalman filters with standard hand-tuned filters and learning based approaches on the nuScenes benchmark

 

This project offers a unique opportunity to contribute to a novel and impactful area of research, as this optimization potential has not been used so far in the field of autonomous driving.

Voraussetzungen

This makes you the perfect candidate:

• curiosity for autonomous driving and machine learning
• some experience with Python and pytorch
• an engaged and independent working attitude

 

If you are interested in this thesis project, please introduce yourself by sending your CV and a transcript of records to cornelius.schroeder@tum.de.

Tags

FTM Studienarbeit, FTM AV, FTM Schroeder, FTM Informatik

Möglicher Beginn

sofort

Kontakt

Cornelius Schröder, M.Sc.
cornelius.schroedertum.de

Ausschreibung