Understanding the 3D world is fundamental to autonomous systems. Our lab develops cutting-edge algorithms for 3D object detection, occupancy prediction, and high-definition (HD) map construction, enabling vehicles to perceive their surroundings with unprecedented accuracy. By integrating multiple sensor modalities through sensor fusion, we enhance spatial awareness, ensuring robust and reliable scene understanding even in complex and dynamic environments. Our research paves the way for safer and more efficient autonomous driving in real-world conditions.

Autonomous vehicles must not only perceive the world but also anticipate future events and act accordingly. Our lab focuses on motion prediction—understanding how other vehicles, pedestrians, and cyclists will move in the next few seconds—and motion planning, ensuring that autonomous agents can navigate safely and efficiently. By leveraging deep learning, probabilistic models, and optimization techniques, we develop scalable solutions that allow autonomous systems to make intelligent, real-time decisions in diverse traffic scenarios.

Building robust AI for self-driving systems requires data-efficient and continuously improving models. Our research tackles this challenge through active learning, self-supervised learning, and continual learning, enabling AI to learn from limited labeled data, adapt to new environments, and retain knowledge over time. By pushing the boundaries of scalable AI, we aim to develop autonomous driving systems that generalize effectively across diverse geographies, weather conditions, and traffic patterns, making widespread deployment feasible.

We envision a future where self-driving vehicles operate with minimal human intervention. Our lab explores end-to-end learning-based approaches that unify perception, prediction, and planning into a single AI-driven framework. By harnessing the power of deep learning and reinforcement learning, we strive to create autonomous systems that can seamlessly navigate urban landscapes, highways, and unstructured environments. Our goal is to bridge the gap between research and real-world deployment, bringing fully autonomous driving closer to reality.