Demos
Task-level demonstrations from RAIDS research programs—spanning embodied data acquisition, aerial manipulation, human–robot collaboration, and digital-twin-enabled manufacturing. All videos play inline; no external links required.
A novel aerial embodied-intelligence platform for dexterous manipulation and interaction in 3D space. It combines aerial mobility with robotic arm manipulation, supports embodied task learning in both simulation and real-world environments, and enables a Real-2-Sim-2-Real workflow for developing transferable aerial manipulation policies.
An AR-guided intelligent drilling system for industrial robotic machining. Target positions are overlaid onto real workpieces for intuitive hole-location confirmation; large language models translate natural-language instructions into structured machining parameters—enabling more efficient, interactive, and flexible human-robot collaboration in drilling tasks.
A bimanual package manipulation demo developed in collaboration with HKPC using the Flexiv AICO 2 robot. The system opens a sealed box and retrieves its contents by learning from teleoperation demonstrations, with stage-wise decomposition and recovery data supporting robust long-horizon execution.
A competition demonstration for the Hangzhou International Embodied Robot Scenario Application Competition, showcasing ORCA-based simulation for humanoid task execution in realistic industrial scenarios and validating embodied policies before real-world deployment.
A versatile industrial embodied-intelligence data acquisition terminal for vision–tactile multimodal capture. It integrates readily with diverse general-purpose robotic arms and dexterous hands, and supports egocentric data collection.
An AR-guided robotic drilling system that combines multimodal human–machine interaction with low-code task programming. Operators can configure, validate, and execute precision drilling workflows more intuitively on the shop floor.
A human–robot collaborative system for retired electric vehicle battery recycling and disassembly. It integrates operator guidance with autonomous robotic handling for safer, more efficient end-of-life battery processing.
A multi-sensor fusion framework for real-time monitoring of automated drilling processes. Force, vision, and process signals are combined to detect anomalies and support quality-assured robotic machining.
The Glovity multimodal data acquisition glove for dexterous manipulation learning—combining vision, tactile, and motion capture in an egocentric wearable form factor to translate expert industrial skill into deployable autonomy.
A human-to-robot skill transfer framework for bimanual dexterous manipulation. MR-assisted teleoperation enables intuitive human demonstrations, while force/tactile-based imitation learning helps the robot acquire and reproduce complex manipulation skills.
A deep reinforcement learning approach to robotic motion planning with interactive trajectory preview, enabling operators to evaluate and refine learned policies before deployment in human–robot collaborative tasks.
An AR-assisted 3D modelling and interactive design system that overlays digital geometry onto physical workspaces, supporting rapid design iteration and human–robot collaborative manufacturing preparation.
A dexterous manipulation learning demonstration integrating vision-based perception and robot control, illustrating how embodied intelligence methods support adaptive interaction in contact-rich industrial tasks.
A VR-assisted teleoperation system for intuitive robot control and embodied data acquisition, enabling immersive demonstration capture for downstream learning and human–robot skill transfer.
An integrated hand–object 6D pose estimation approach with explicit occlusion handling, supporting perception-driven human–robot collaborative manipulation and disassembly tasks.
A human–robot collaborative manufacturing demonstration exploring adaptive interaction between operators and robotic systems, illustrating how intelligent coordination improves flexibility on representative industrial tasks.
A digital twin system demonstration integrating physical production assets with virtual models for monitoring, simulation, and decision support in smart manufacturing environments.
