CopilotTele: A Shared-Autonomy Teleoperation System for Bimanual Manufacturing Assembly with Risk-Guided Arbitration
Risk-guided shared autonomy for bimanual wire-harness assembly, combining DiTFlow action assistance with on-demand human takeover.
Embodied AI · Foundation Models · Robot Learning
Frontier Intelligent
Robotics Lab
We work on embodied intelligence: building embodied foundation models such as VLA and WAM that let robots understand tasks and the physical world, and act reliably in real environments.
01 / 07
Research Overview
Research areas >Frontier Intelligent Robotics Lab (FIR Lab) at HKISI-CAS focuses on embodied intelligence. We study embodied foundation models — vision-language-action (VLA) and world-action models (WAM) — together with world models and robot learning methods that turn them into dependable skills on real robots.
Our work closes the loop between models, data and deployment: models produce actions, real-world execution produces data, and that data improves the models in turn.
RESEARCH QUESTION
How can robots understand the physical world and turn that understanding into reliable, generalizable manipulation skills?
Selected Publications
All publications >
AssemblyMate: An Interactive Context-Aware AI-XR Co-worker with Multimodal Spatial-Temporal Reasoning for Manufacturing Assembly
An AI-XR co-worker that combines multimodal and spatial-temporal reasoning agents to guide operators, evaluate assembly actions and coordinate a collaborative robot.
Augmented Reality-Assisted Cyber-Physical System for Manual Spray Painting of Complex Geometries
AR feedback for manual spray painting: a sensorized spray gun, coating-deposition simulation and an AR headset show coverage on complex surfaces in real time.
VLAbot: A Human Vision-Language-Action Models Interaction Framework for Robotic Assembly
A human-robot collaboration framework that combines multi-expert agents, VLA models, and augmented reality interaction for long-horizon assembly tasks.
Indoor Depth Completion via Geometry Processing with RGB Image
A geometry-processing approach for accurate indoor depth completion, producing sharper object boundaries and reducing distortion in manufacturing environments.
Lab News
All news >- 2026CopilotTele is accepted to IEEE CASE 2026 and selected as a Best Student Paper Award Finalist.
- 2026AssemblyMate is published in npj Advanced Manufacturing (Nature Portfolio).
- 2026AR-CPS receives the ASME IDETC/CIE 2026 VARE Technical Committee Best Paper Award.
- 2026Prof. Zhang gives an invited talk on embodied physical intelligence for smart manufacturing at OFweek 2026, Shenzhen.
- 2026FIR Lab will join the Hong Kong Institute of Science & Innovation, Chinese Academy of Sciences (HKISI-CAS) in June 2026.
- 2026VLAbot published in Robotics and Computer-Integrated Manufacturing (RCIM).
- 2026Indoor depth completion via geometry processing published in Expert Systems With Applications.
- 2025RoPESim accepted to IEEE CASE 2025.
Lab Events & Activities
View Events >System Demonstrations
Live demonstrations of embodied AI systems and autonomous robot manipulation.
Academic Seminars
Weekly group reading sessions and technical discussions on VLA models, foundation policies, and sim-to-real transfer.
Milestones & Gatherings
Project milestone reviews, team gatherings, and collaborative sessions across robotics and AI disciplines.