Loading...
Bird's eye view based multi-agent trajectory predictiob for autonomous driving with segmentation and sensor fusion
Date
2025-12
Abstract
Accurate trajectory prediction is a key component of autonomous driving systems, particularly in dynamic and complex traffic environments where anticipating future movements is essential for safety. This thesis presents a novel multi-sensor fusion framework that utilises data from cameras, LiDAR, and radar to enhance the accuracy and robustness of vehicle trajectory prediction. By leveraging the strengths of each sensor, including visual information from cameras, precise 3D data from LiDAR, and reliable velocity measurements from radar, a comprehensive representation of the driving scene is constructed. The architecture proposed in this thesis integrates these heterogeneous inputs using advanced fusion techniques that align LiDAR point clouds with camera imagery for spatial consistency. This fused data is processed by a transformer-based deep learning model that captures temporal dependencies and agent interactions, enabling the system to focus on the most salient scene elements and improve prediction in congested or uncertain environments.
Evaluations on benchmark datasets using metrics such as Average Displacement Error, Final Displacement Error, collision rate, and miss rate demonstrate that the proposed approach consistently outperforms traditional baselines. The methods developed in this thesis achieve state-of-the-art results, highlighting the impact of sensor fusion and temporal modelling in advancing safe and reliable autonomous driving.
Supervisor
Description
Publisher
University of Limerick
Citation
Collections
Files
Loading...
Sharma_2025_Birds.pdf
Adobe PDF, 89.78 MB
ULRR Identifiers
Funding code
Funding Information
Sustainable Development Goals
External Link
License
Attribution-NonCommercial-ShareAlike 4.0 International