Xunjiang Gu

Sucre, Bolivia – Christmas 2022.

My name is Xunjiang (Alfred) Gu, and I am a first-year Master’s student at the University of Toronto, supervised by Prof. Igor Gilitschenski. I completed my Bachelor of Applied Science in Engineering Science at the University of Toronto, specializing in Robotics with a Business minor.

Previously, I interned at Noah’s Ark Lab, Huawei Technologies Canada, under the guidance of Dr. Amir Rasouli, where I contributed to autonomous driving research. I also spent a wonderful summer at NVIDIA as a joint Product/Research intern, supervised by Dr. Urs Muller and Dr. Boris Ivanovic, focusing on autonomous driving technologies. Currently, I am working with Dr. Kashyap Chitta on modeling vehicle dynamics at vehicles’ operating limits and in various extreme scenarios.

My research interests include autonomous driving, trajectory prediction, and vehicle dynamics modeling. I am particularly interested in advancing the safety and reliability of autonomous vehicles and robotics applications for real-world deployment.

Feel free to reach out to me via email if you have any questions, or simply want to have a chat!

News

Jul 01, 2024	One paper accepted by ECCV 2024!
May 29, 2024	Producing and Leveraging Online Map Uncertainty in Trajectory Prediction is a Best Paper Award Finalist at CVPR 2024! Top 0.2% of all submissions (24/11,532)
Feb 26, 2024	One paper accepted by CVPR 2024!

Selected Publications

CVPR
Producing and Leveraging Online Map Uncertainty in Trajectory Prediction

Xunjiang Gu, Guanyu Song, Igor Gilitschenski, Marco Pavone, and Boris Ivanovic

In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024

Abs arXiv Bib Code

High-definition (HD) maps have played an integral role in the development of modern autonomous vehicle (AV) stacks, albeit with high associated labeling and maintenance costs. As a result, many recent works have proposed methods for estimating HD maps online from sensor data, enabling AVs to operate outside of previously-mapped regions. However, current online map estimation approaches are developed in isolation of their downstream tasks, complicating their integration in AV stacks. In particular, they do not produce uncertainty or confidence estimates. In this work, we extend multiple state-of-the-art online map estimation methods to additionally estimate uncertainty and show how this enables more tightly integrating online mapping with trajectory forecasting. In doing so, we find that incorporating uncertainty yields up to 50% faster training convergence and up to 15% better prediction performance on the real-world nuScenes driving dataset.
@inproceedings{gu2024producing, title = {Producing and Leveraging Online Map Uncertainty in Trajectory Prediction}, author = {Gu, Xunjiang and Song, Guanyu and Gilitschenski, Igor and Pavone, Marco and Ivanovic, Boris}, booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition}, pages = {14521--14530}, year = {2024}, }
ECCV
Accelerating Online Mapping and Behavior Prediction via Direct Bev Feature Attention

Xunjiang Gu, Guanyu Song, Igor Gilitschenski, Marco Pavone, and Boris Ivanovic

In European Conference on Computer Vision, 2025

Abs arXiv Bib Code

Understanding road geometry is a critical component of the autonomous vehicle (AV) stack. While high-definition (HD) maps can readily provide such information, they suffer from high labeling and maintenance costs. Accordingly, many recent works have proposed methods for estimating HD maps online from sensor data. The vast majority of recent approaches encode multi-camera observations into an intermediate representation, e.g., a bird’s eye view (BEV) grid, and produce vector map elements via a decoder. While this architecture is performant, it decimates much of the information encoded in the intermediate representation, preventing downstream tasks (e.g., behavior prediction) from leveraging them. In this work, we propose exposing the rich internal features of online map estimation methods and show how they enable more tightly integrating online mapping with trajectory forecasting. In doing so, we find that directly accessing internal BEV features yields up to 73% faster inference speeds and up to 29% more accurate predictions on the real-world nuScenes dataset.
@inproceedings{gu2025accelerating, title = {Accelerating Online Mapping and Behavior Prediction via Direct Bev Feature Attention}, author = {Gu, Xunjiang and Song, Guanyu and Gilitschenski, Igor and Pavone, Marco and Ivanovic, Boris}, booktitle = {European Conference on Computer Vision}, pages = {412--428}, year = {2025}, organization = {Springer}, }