Franka Research 3 AI Inference powered by NVIDIA Isaac GR00T N1.5 robot foundation model at automatica 2025

At automatica 2025 in Munich, a new application of embodied AI will be publicly demonstrated for the first time. From June 24 to 27, visitors will be able to observe AI inference using the NVIDIA Isaac GR00T N1.5 robot foundation model operating on the Franka Research 3 (FR3) in a live, dual-arm setup FR3 Duo.

This demonstration highlights the capabilities of the Franka’s dual arm systems powered by GR00T N1.5, a NVIDIA foundation model for robotic control. By integrating this advanced action model, the system is able interpret visual input, understand task context, and autonomously perform complex manipulation—without the need for task-specific programming or hardcoded logic.

Overview of the Demonstration

The system consists of two FR3 robotic arms, coordinated as a dual-arm platform. It is controlled by an inference using NVIDIA Isaac GR00T N1.5.

Key elements of the demo:

• No scripted behavior: The robot does not follow pre-programmed instructions or custom logic trees.
• Real-time adaptation: Based on camera input and contextual cues, the system infers goals and executes suitable actions.
• General-purpose reasoning: The model generalizes across tasks rather than being optimized for a single fixed scenario.

The demonstration highlights how foundation models can reduce the need for manual integration and task engineering in robotics.

About NVIDIA Isaac GR00T N1

NVIDIA Isaac GR00T N1 is an open foundation model for generalized humanoid robot reasoning and skills. It is designed to handle embodied intelligence tasks by integrating multiple sensory and contextual inputs, such as:

• Visual perception
• Language-based goal specification
• Robot state and proprioception
• Action planning
  
  

GR00T N1 empowers robotic systems to interpret instructions and act autonomously, adapting to new tasks and environmental changes—without retraining or reprogramming.
In the FR3 Duo demonstration, GR00T N1.5 runs inference directly on sensor input, generating real-time actions. This marks a shift away from traditional robotics pipelines based on handcrafted code and rigid task logic.

Why Use Franka Research 3?

The Franka Research 3 (FR3) platform offers technical advantages that support the deployment of foundation models like GR00T N1.5:

• Torque sensor at every joint: Enables compliant and sensitive motion, suitable for tasks involving physical contacts.
• Low-latency control: Through Franka Control Interface (FCI), developers (and AIs) can access real-time joint-level control and full state feedback.
• Modular and accessible hardware: The FR3 system is compact and lightweight, allowing for flexible setup and integration.
• Human-robot interaction ready: FR3 is designed with built-in safety features and responsive motion control to enable safe collaboration.

These features make FR3 well-suited for AI-driven control systems that rely on continuous feedback and adaptive motion generation.

Technical Impact

Running GR00T N1.5 inference on the FR3 Duo demonstrates several key capabilities of modern robotic AI systems:

• Task generalization: Ability to perform a variety of tasks without task-specific programming.
• Robustness: System can adapt behavior to changing conditions or layouts.
• Dual-arm coordination: High-level planning that synchronizes two arms for shared tasks.
• Reduced development effort: Minimal manual configuration required for new scenarios.
  
  

This approach significantly reduces the effort needed to deploy new robotic tasks, making advanced robotics more accessible to integrators and researchers.

Explore other Nvidia-powered robotics innovations: https://blogs.nvidia.com/blog/automatica-robotics-2025/