Product Features
- Physical AI robot training equipment for imitation learning and reinforcement learning consisting of two 6-axis robot arms, intelligent robot controller, work cell base
- Enhance mobility and scalability by placing leader robot and follower robot on separate aluminum profile base
- GUI-based data view labeling replay is possible through monitor · camera · microphone included in leader robot
- Configure view calibration·AR marker · segmentation · pose estimation pipeline through follower robot and two cameras in follower robot
- Establish strong grip against lightingㆍlayout change through camera-based vision grasping
- Support imitation learning and reinforcement learning on both leader and follower robot using intelligent robot controller integrated with ROS 2 open stack
- Provide modelㆍdatasetㆍtoolchain integrated into PyTorch to easily reproduce and share real-world robotics AI
- Consistently manage teleoperationㆍdataset standardizationㆍpolicy learningㆍevaluationㆍreproduction in a single framework
- Support training experiment based on safety training, synthetic data generation, domain randomization through two-way synchronized simulation with real equipment
- Starting with synchronized multimodal data, validating it on digital twin, and supporting safe learning and deployment on real equipment
- Generative AI integration is possible, including diffusion policy, VLM/VLA, prompt-based control
- Provide block pick-and-place examples based on teleoperation data collection and Behavior Cloning
- Provide digital twin examples based on NVIDIA Isaac Sim
Layout
Training Contents
- Physical AI & PhysicAI-Arm Pipeline Overview
- ROS2 Basic & Interface Standardization
- Vision Pipeline I — Camera Calibration
- Data Collection Strategy & Quality Control
- Behavior Cloning (BC) Baseline
- Model Evaluation & Experiment Management
- Diffusion Policy & Action Smoothing
- Vision Pipeline II — Grasping Core
- Isaac Sim Integration & Digital Twin
- ROS2-based Deployment & Stabilization
- Generative AI Integration
Components
