
RCTHB02 is a high-performance AI Robotics Controller built on the NVIDIA Thor platform, designed to serve as the intelligent computing core for next-generation robots.
Combining powerful edge AI computing, advanced multimodal sensor fusion, high-speed connectivity, and the NVIDIA robotics software ecosystem, RCTHB02 integrates perception, AI inference, decision-making, and motion-control capabilities into a compact embedded platform.
Designed for humanoid robots, embodied AI systems, industrial dual-arm collaborative robots, inspection robots, and next-generation service robots, RCTHB02 provides the computing foundation for intelligent, autonomous robotic systems.
Powered by NVIDIA Thor’s next-generation GPU architecture and Transformer acceleration capabilities, RCTHB02 delivers 1,000+ TOPS-class AI computing performance, providing the processing power required for demanding embodied AI workloads.
The platform supports advanced AI models and algorithms, including Large Language Models (LLMs), Vision-Language Models (VLMs), and Diffusion Policies, enabling robots to perform sophisticated perception, reasoning, and action generation directly at the edge.
By moving AI computation closer to the robot, RCTHB02 reduces dependence on cloud computing and enables fast, responsive decision-making for dynamic and unstructured environments.
RCTHB02 provides a comprehensive set of high-speed interfaces for connecting multiple sensor modalities, including:
With hardware-level time synchronization, data from vision, force, touch, and audio sensors can be precisely aligned within a unified time domain.
This enables robots to build a richer understanding of their environment and supports more accurate perception, interaction, and manipulation.
RCTHB02 is designed to support the complete robotic intelligence pipeline—from environment perception and mapping to navigation, decision-making, manipulation, and motion control.
With accelerated robotics computing capabilities and integration with the NVIDIA Isaac ecosystem, RCTHB02 can support advanced applications including:
Perception → Localization → Mapping → Planning → Decision-Making → Manipulation → Motion Control
This integrated architecture helps robotics developers build more capable and responsive intelligent machines while reducing system complexity.
RCTHB02 is designed to integrate with the modern NVIDIA robotics software ecosystem, including NVIDIA Isaac, ROS 2, TensorRT, and cuDNN, providing developers with a flexible environment for AI model deployment and robotics application development.
Developers can leverage pretrained AI models, optimize inference performance, and deploy advanced robotics algorithms directly to the robot.
Integration with NVIDIA Omniverse enables simulation and digital-twin workflows, allowing developers to validate robotic behaviors virtually before deployment to physical systems.
This helps accelerate the development cycle from AI model → simulation → deployment → real-world operation.
Despite its compact embedded form factor, RCTHB02 combines high-performance AI computing, multimodal sensor processing, and robotics software integration into a single platform.
This enables robotics manufacturers to develop intelligent robotic systems with:
RCTHB02 goes beyond a conventional embedded controller.
It serves as an AI computing and control foundation for embodied intelligence, connecting the robot’s sensors, AI models, decision-making systems, and actuators within a unified architecture.
From humanoid robots and industrial collaborative robots to inspection and service robots, RCTHB02 provides the AI “brain” that helps robots see, understand, decide, and act.
Ready to try RCTHB02 for your next robot build? Reach out to us today and let’s get you started!