.png)
AI 機器人與具身智慧
Robotics & Embodied AI
Humanoid Robot Training, Teleoperation & Embodied AI Integration
Integrating humanoids, motion capture, data gloves, dexterous hands, and simulation to build an end-to-end pipeline—from human demonstration to robot learning and real-world deployment. Powered by ROS 2 and AI models, we empower enterprises and research teams in Embodied AI development.
Four Key Modules for Embodied AI
Teleoperation
Combines tactile haptic gloves with immersive headsets to collect human teleoperation demonstration data in real time.
NVIDIA Isaac Sim
Leverages high-fidelity physics-based synthetic data generation to accelerate Sim-to-Real transfer for reinforcement learning and vision models.
Millimeter-Level Precision
Tracks and annotates robot joints and environmental objects, utilizing high-precision motion capture data from XSENS and MANUS for robot training.
VLA Model Validation
Integrates Vision-Language-Action foundation models to enable closed-loop control of dual-arm humanoid motion via natural language commands.
Core Application Modules for Embodied AI
Precise Motion Capture Authentic Human Realism
Combining MANUS Metagloves with Movella Xsens motion capture and Tesollo dexterous hands, we deliver a comprehensive human-robot interaction and AI training solution. Through real-time human data collection and high-precision control, natural human movements are seamlessly transformed into intelligent robotic behavior.
Capturing high-precision finger movements with MANUS Metagloves and streaming real-time data to ROS/SDK for seamless control of robotic hands and manipulators.
Wear the Xsens motion capture system to track full-body movement for teleoperation. Real-time data is streamed directly to the robot and fused with its sensory inputs to train AI and deep learning models.
AI-Assisted Acceleration of Robot Intelligence. Through human demonstrations, AI learns operational logic from massive datasets and deconstructs complex behaviors into repeatable commands. This enables the AI to automatically adapt actions across different scenarios, enhancing consistency and efficiency.
Executing Actions with Tesollo Dexterous Hands. High-precision finger motions are captured by MANUS Metagloves and streamed to ROS. After data parsing and transformation, ROS sends the commands to Tesollo dexterous hands for real-world execution.
MANUS VR Gloves: Sub-millimeter finger tracking with support for ROS and multiple data output formats.
Movella Xsens: Highly reliable full-body motion capture that completely faithfully captures human pose, velocity, and dynamics.
Motion Capture with Xsens
.png)
MANUS Applications
.png)