Manufacturing
Universal Robots Accelerates Cobot Development with NVIDIA
Objective
Universal Robots is addressing the limitations of traditional robots in handling complex tasks and operating in dynamic environments with its AI Accelerator, developed with NVIDIA Isaac™ Platform. The AI Accelerator enables developers to integrate advanced vision AI technologies in their collaborative robot (cobot), allowing for real-time object recognition, adaptive motion control, and continuous learning.
Partner
Universal Robots
Use Case
Robotics
Products
NVIDIA Isaac
NVIDIA Jetson
Working with NVIDIA, Universal Robots has delivered transformative performance improvements that unlock entirely new application categories:
Breaking Through Traditional Cobot Limitations
Founded in 2005, Universal Robots revolutionized industrial automation by launching the world's first commercially viable collaborative robot in 2008—the UR5—that could safely work alongside humans without safety caging. Since then, the company has democratized industrial automation with easy-to-program cobots that work safely with humans.
Today’s high-mix, low-volume production lines require cobots that can perceive, adapt, and act with intelligence in dynamic environments – capabilities that traditional cobot programming couldn’t deliver. With labor shortages driving demand for automation and industrial AI, Universal Robots sought to build a solution to bridge the gap between current AI capabilities and evolving industrial robot deployment needs.
Building the AI Accelerator with NVIDIA
Universal Robots addressed these challenges by introducing the AI Accelerator, a platform for physical AI development that enables robots to perceive and interact with their environment and solve complex problems that were previously beyond their static capabilities. The AI Accelerator is the product of an eight-year collaboration between Universal Robots and NVIDIA that spans the evolution from early GPU computing to today's edge AI platforms.
The maturity of the NVIDIA Isaac™ robotics platform—a comprehensive suite for building, simulating, and deploying AI-powered robots—immediately captured the interest of the Universal Robots team and quickly scaled the collaboration into an active joint program. The NVIDIA Isaac platform brings together NVIDIA® CUDA®-accelerated libraries, advanced application frameworks, and pretrained AI models to provide a robust foundation for modern robotics. NVIDIA Isaac facilitates robot simulation and robot learning, enabling a new generation of intelligent automation that includes breakthroughs in humanoid robot development and adaptable industrial cobots.
Universal Robots integrated key components of the Isaac platform, such as AI models and libraries, into their AI Accelerator solution powered by NVIDIA Jetson AGX™. This created an industrial-grade, AI-enabled platform that now extends across Universal Robots’ e-Series and UR Series cobots—including the new UR15.
Technical Architecture and Ecosystem Impact
At its core, the Universal Robots AI Accelerator benefits from:
- NVIDIA Isaac Manipulator for advanced perception and motion planning, including NVIDIA cuMotion, FoundationPose, and FoundationStereo
- NVIDIA Isaac Transport for ROS (NITROS) for acceleration in ROS 2.0 environments
- NVIDIA Jetson AGX Orin™ for low-latency, high-accuracy edge AI without cloud dependencies
- PolyScope X with API-first development strategy and full ROS 2.0 support
The AI Accelerator lets cobots operate in unstructured environments and perform adaptive tasks that were previously only achievable with expensive custom systems.
The platform has attracted a growing ecosystem of AI-focused partners—including 3D Infotech, T-Robotics, AICA, Acumino, and Groundlight—who are deploying solutions today across industries ranging from automotive to food and beverage.
The AI Accelerator's technical architecture demonstrates how edge AI can be industrialized without sacrificing performance. An embedded NVIDIA Jetson Orin delivers the computational power needed for real-time AI inference, while Isaac ROS libraries provide modular, production-ready components.
The system also delivers low-latency, high-accuracy edge AI for dense industrial setups without cloud reliance—a critical advantage for manufacturers concerned about connectivity, latency, and data security.
The platform's ROS 2 foundation accelerates development and deployment, enables modular design, and improves the crucial perception-to-control handoff that determines system responsiveness. This technical approach lets you focus on application logic rather than low-level AI implementation.
Universal Robots
Bridging the Sim-to-Real Gap
Traditional robotic assembly systems struggle with adaptability in dynamic industrial environments, requiring extensive manual reprogramming for new parts or configurations. Using NVIDIA Isaac Lab and Isaac ROS, Universal Robots has delivered a sim-to-real workflow for reinforcement learning training and real-world execution. The system combines Isaac Lab’s physically accurate simulations with domain randomization to train reinforcement learning policies for gear grasping, transport, and insertion.
The robot policies are trained entirely in simulation, so they can successfully assemble multiple gears from random initial poses without real-world fine-tuning. By using synthetic data, domain randomization, and NVIDIA IndustReal algorithms and tools, Universal Robots overcomes sim-to-real gaps, enabling robot policies to adapt seamlessly to gear sequencing and spatial variations. Eliminating the need for manual reprogramming also helps Universal Robots reduce deployment costs for contact-rich applications.
Read more in the tech blog, Bridging the Sim-to-Real Gap for Industrial Robotic Assembly Applications.
Accelerating Cobot Time-to-Market
The collaboration between Universal Robots and NVIDIA demonstrates how established industrial companies can successfully integrate cutting-edge AI technologies while preserving their core value propositions. By packaging advanced AI capabilities in an accessible platform, the AI Accelerator enables companies of all sizes to deploy intelligent automation without requiring specialized AI expertise.
The solution has created new revenue streams from previously impossible applications, attracted AI-focused partners to Universal Robots' ecosystem, and consolidated the company's position as the preferred cobot platform for AI applications.
Looking Ahead
Universal Robots is developing application-specific solutions for global customers, tapping into new AI models and CUDA-accelerated libraries through Isaac ROS to engage a new generation of AI developers.
NVIDIA and Universal Robots are committed to a scalable approach for bringing advanced AI capabilities to factories worldwide. The AI Accelerator offers manufacturers a practical path to greater flexibility and efficiency while maintaining the ease of human-robot collaboration that defines collaborative robotics.
“With the AI Accelerator, we’re empowering our partners to bring AI-powered cobot solutions to market faster and with greater flexibility. The result is smarter, more adaptable automation that meets real-world needs across a wide range of applications and delivers a faster return on investment for end-customers.”
Ujjwal Kumar
Group President, Teradyne Robotics
Ready to Get Started?
Jump-start your AI robot development with NVIDIA CUDA-accelerated libraries, application frameworks, and AI models available through the NVIDIA Isaac platform for robotics.