Embodied Intelligent Robot Development Platform

Our embodied AI robot development platform supports data collection in real-world scenarios with cloud-based training and inference. It provides underlying SDKs and ROS Hardware source code, and is compatible with VLA frameworks like lerobot, comprehensively facilitating secondary development of embodied intelligence, enabling scenario implementation, and application in commercial retail, warehousing logistics, scientific research, education, and other scenarios.

· Semantic Understanding and Decision-Making: Large language models empower robots to understand human natural language instructions, such as adjusting cleaning modes or planning complex task workflows through commands.

· Multimodal Perception Fusion: Vision-Language-Action (VLA) models integrate sensor data (e.g., visual, tactile), enabling robots to dynamically adapt to environmental changes, such as recognizing ground textures to adjust gait.

· Training Optimization: Large models design reward functions for reinforcement learning, accelerating the efficiency of robot skill acquisition.

· Physical Validation Scenarios: Data generated from robot interactions in real environments (e.g., grasping force feedback, environmental obstacle information) can correct "hallucination" issues in large models, enhancing decision-making reliability.

· Cross-Domain Data Generation: Practical data from scenarios like commercial retail, warehousing logistics, industry, healthcare, and scientific research provides scarce training samples for large models, promoting the evolution of general capabilities.

Embodied AI Robot Development Platform Specifications:

Dimensions: 600 x 664 x 1800 mm;

Arm Span: 1710 mm;

Total Degrees of Freedom: 2 + 1 + (7+1) x 2 = 19 DoF;

Battery Capacity: 48V 60AH;

Single Arm Degrees of Freedom: 7 DoF (excluding gripper);

Lifting Range: 0.5 ~ 1.7 m;

Maximum Speed: 1.5 m/s;

System Environment: Ubuntu + Docker;

Exoskeleton Configuration: Dual-arm seven-axis isomorphic;

All Five Core Modules of the Embodied AI Robot Body Are Open:

Underlying source code open;

Supports force-position hybrid control;

7-DOF robotic arm explores more operational space;

Intelligent wire-controlled chassis;

Nonlinear calibrated two-finger gripper;

Servo lifting platform;

Multimodal perception;

Embodied AI Robot Development Platform: Hardware-Software Integrated Coordination:

Wheeled semi-humanoid embodied AI robot design adapts to various scenarios, facilitating rapid deployment;

Developer toolkit supports integrators' secondary development, enabling scenario implementation;

Exoskeleton teleoperation system provides smoother, more precise, and comfortable operation experience;

Visual operation panel enables ultra-low latency monitoring of multi-view teleoperation feeds;