Introducing reComputer J601 Carrier Board: A Compact Physical AI Platform for NVIDIA® Jetson AGX Thor™

Built for embodied AI development and production, reComputer J601 Carrier Board is a compact edge AI platform designed for NVIDIA Jetson AGX Thor, combining up to 2070 TFLOPS AI computing performance with robotics-focused connectivity for humanoid robots, robotic arms, and Physical AI applications.

---

Bringing NVIDIA Jetson AGX Thor AI Computing Power to Physical AI Robotics

Robotics is entering a new era driven by Physical AI.

Unlike traditional robots that execute predefined commands, next-generation intelligent robots need to perceive their surroundings, understand complex environments, process multimodal information, and make real-time decisions. This evolution requires powerful edge AI computing platforms capable of running advanced models directly on robotic systems.

NVIDIA Jetson AGX Thor is designed for robotics, humanoid robots, and Physical AI workloads, delivering the computing performance required for real-time AI inference at the edge. Built on NVIDIA’s advanced architecture, Jetson AGX Thor delivers up to 2070 TFLOPS of AI computing performance, enabling demanding workloads including vision AI, large language models (LLMs), vision-language-action (VLA) models, and multimodal AI applications.

However, bringing this level of AI intelligence into real-world robots requires more than computing power alone. Developers need a compact and flexible hardware platform that can connect high-performance AI computing with cameras, sensors, controllers, and robotic systems.

That is where reComputer J601 Carrier Board comes in.

reComputer J601: The Compact AI Computing Platform for Humanoid and Embodied AI Robots

The reComputer J601 Carrier Board is a compact edge AI carrier board designed specifically for the NVIDIA Jetson AGX Thor module, providing the connectivity and expansion capabilities required for robotics and Physical AI development.

By combining Thor’s advanced AI computing capability with robotics-focused connectivity and expansion, J601 provides the foundation for building intelligent machines that can perceive, reason, and interact with the physical world.

With a compact 168mm × 153mm footprint, J601 is designed for space-constrained robotic platforms, making it especially suitable for humanoid robot development, robotic prototypes, and embedded AI applications.

Together with Jetson AGX Thor™, J601 provides the computing foundation for humanoid robots and embodied AI systems, supporting advanced AI workloads including:

• Vision AI Models for Robotics Perception — enabling robots to perceive and understand their surroundings through tasks such as object detection, tracking, and localization. Developers can deploy vision AI models such as YOLO for real-time object detection and LocateAnything for vision-language object localization, enabling robots to perceive and understand their surroundings.
• Large Language Models (LLMs) — enabling robots to understand natural language instructions, perform reasoning, and support intelligent task planning.
• Vision-Language-Action (VLA) Models — combining visual perception, language understanding, and robot control to enable intelligent manipulation and autonomous behaviors, such as NVIDIA® Isaac™ GR00T for humanoid robot learning, imitation learning, and autonomous manipulation.
• Multimodal AI Models — combining vision, language, and sensor information to support advanced scene understanding and embodied AI applications.
• Robotics Frameworks — supporting development ecosystems including NVIDIA Isaac, ROS 1/2, PyTorch, and Hugging Face.

Designed for Real-World Robotics Integration

Developing intelligent robots requires more than AI inference performance. A robotics computing platform must seamlessly connect multiple sensors, communication systems, and control components.

reComputer J601 is designed around these real-world requirements.

Rich Robotics Connectivity for AI-Powered Robot Systems

Modern robots rely on high-speed communication between computing modules, cameras, sensors, and controllers.

J601 integrates:

• 4× 10Gb EthernetRJ45 ports
• 4× USB 3.2 Type-A ports
• HDMI 2.1
• CAN / CAN-FD interfaces
• RS-485 and RS-232/422/485 support

These interfaces enable developers to build scalable robotic architectures for research, prototyping, and production deployment.

Advanced Vision AI With Multi-Camera Support

Perception is fundamental to embodied AI.

To help robots understand their environment, reComputer J601 supports up to 8× GMSL2 cameras through integrated MAX96712 deserialization, enabling high-bandwidth multi-camera perception systems.

Combined with Jetson AGX Thor’s AI computing capability, J601 supports advanced vision AI applications such as:

• BEV (Bird’s Eye View)
• Occupancy Grid
• SLAM
• Object detection
• Autonomous navigation

This allows developers to build robots with stronger environmental awareness and real-time decision-making capabilities.

Know More: https://wiki.seeedstudio.com/fine_tune_gr00t_n1.7_for_rebot_arm_and_deploy_on_robotics_j601/

By combining reComputer J601 with NVIDIA Isaac GR00T N1.7 and reBot Arm, developers can build vision-language-action (VLA) robotic systems that perceive their surroundings, understand instructions, and execute physical tasks.

In this workflow, camera inputs are processed by Jetson AGX Thor through J601’s high-bandwidth vision interfaces, enabling AI models to interpret visual information and generate intelligent robotic actions. This demonstrates how J601 helps bridge the gap between AI perception and real-world robot interaction, accelerating the development of next-generation embodied AI applications.

Flexible Expansion for AI and Robotics Applications

Different robotic systems require different connectivity options.

reComputer J601 provides flexible expansion through:

• M.2 Key M for high-speed storage
• M.2 Key E for WiFi/Bluetooth connectivity
• M.2 Key B for 4G/5G communication

Together with rich GPIO, I2C, I2S, and peripheral interfaces, J601 provides developers with the flexibility to customize systems for different robotics scenarios.

Built for Embodied AI Development From Simulation to Deployment

Developing Physical AI requires a complete workflow — from simulation and AI model development to real-world deployment.

reComputer J601 supports popular AI and robotics frameworks, including:

• NVIDIA Isaac
• ROS 1 / ROS 2
• PyTorch
• Hugging Face ecosystem

Developers can build complete workflows covering:

AI model development → Simulation → Edge deployment → Robot operation

By combining Jetson AGX Thor’s AI computing capabilities with J601’s robotics integration features, developers can accelerate the transition from AI models to intelligent machines.

Compact Design for Next-Generation Robot Platforms

Robotics systems often face strict size and power constraints.

With its compact 168mm × 153mm design, wide voltage input support from 19V to 48V DC, and compatibility with 48V battery systems, reComputer J601 is optimized for embedded robotics applications.

Its compact footprint makes it suitable for:

• Humanoid robots
• Robotic arms
• Autonomous mobile robots
• Edge AI devices
• Robotics research platforms

Build the Future of Physical AI With reComputer J601

The future of robotics requires AI systems that can move beyond digital intelligence and operate in the physical world.

Powered by NVIDIA Jetson AGX Thor, equipped with rich robotics interfaces, and optimized for embodied AI workloads, reComputer J601 Carrier Board provides the computing foundation for building the next generation of intelligent robots.

From humanoid robotics to advanced edge AI applications, J601 helps developers unlock the possibilities of Physical AI.

Start building with reComputer J601 Carrier Board:https://www.seeedstudio.com/reComputer-J601-Carrier-Board-for-Jetson-AGX-Thor-p-6937.html

About Author