Robot Control
Practical approaches to perception, motion and whole-body control for real robotic systems.
WORKBUILT
BUILT
TO MOVE.
We tackle real, unresolved problems in robotics, building practical methods across perception, decision-making and control, then validating them on physical systems in complex environments.
Aerial Robotics
Motion planning and control for agile, autonomous aerial robots.
Multi-Agent
Learning to coordinate missions and reason about interactions among agents.
Robotic Task Planning
Learning efficient task sequences and collision-free paths for robotic systems.
Multimodal Fused Learning for Solving the Generalized Traveling Salesman Problem in Robotic Task Planning
Graph and image representations are fused to solve generalized traveling salesman problems for real-time robotic task planning.
Attention-based Priority Learning for Limited Time Multi-Agent Path Finding
Attention-based reinforcement learning learns agent priorities for scalable, collision-free path planning under limited computation time.
Autonomous Vehicles
Trajectory prediction and safe planning for autonomous vehicles.
VLA
Vision-language-action models for cross-embodiment robot learning.
Lifelong Learning
Learning new knowledge over time while mitigating catastrophic forgetting.
Naming to Learn: Class Incremental Learning for Vision-Language Model with Unlabeled Data
A class-incremental learning approach for vision-language models that leverages unlabeled data through a naming-and-filtering mechanism to continually learn new classes while mitigating catastrophic forgetting.