Nvidia, Intrinsic Collaborate to Enhance Robot Task Management

Alphabet spinout Intrinsic has partnered with Nvidia to use generative AI for improving robotic programming tasks.

The two companies announced the partnership at the recent Automate 2024 trade show, where the pair integrated Nvidia’s robotics solutions into Intrinsic’s platform to power a robotic arm.

The robot arm was entirely trained for its task on synthetic data and was able to perform a variety of grasping tasks

Using Nvidia’s Isaac Manipulator foundation models in Intrinsic’s platform, the robot seamlessly picked up and grouped objects.

Intrinsic CEO Wendy Tan White wrote in a blog post that synthetic data generated from a foundation model could make future robot deployments easier, enabling units to be more adaptive to a variety of tasks while also reducing development costs.

“Instead of hard-coding specific grippers to grasp specific objects in a certain way, efficient code for a particular gripper and object is auto-generated to complete the task using the foundation model and synthetic training data,” Tan White said.

With the latest AI foundation models, companies can program a variety of robot configurations that are able to generalize and interact with diverse objects inside real-world environments,” said Deepu Talla, Nvidia’s vice president for robotics and edge computing. “As the collaboration between Intrinsic and NVIDIA deepens we will be able to help companies scale and automate their industrial manufacturing operations.”

Related:Nvidia, Teradyne Robotics Launch AI-Powered Industrial Robots

Launched in 2021 as part of Alphabet’s X moonshot program, Intrinsic was formed as a software-first robotics company. Its work focuses on developing more sophisticated robotics solutions for industrial settings including manufacturing.

The company previously partnered with Siemens to “democratize access to robotics.”

Intrinsic is also working with Google DeepMind to develop tools to help robotics programmers input information faster.

At Automate 2024, the pair of Alphabet-owned companies jointly created an AI-based robot motion planner, which allows engineers to program multiple robots sharing the same workspace at once.

Using a model trained on synthetic data, Intrinsic and DeepMind were able to generate motion paths and trajectories which can be easily taught to robots to help them plan their paths more effectively.

The pair were able to use the synthetic data approach to instruct four robots working on welding a car. The model automatically generated the motion plans and trajectories for each robot.

Intrinsic said the machine learning-based model enabled the robots to perform their roles 25% better than traditional programming methods.

Related:Microsoft, Sanctuary AI Team to Advance General Purpose Robots

DeepMind and Intrinsic also worked together to improve robot dexterity. The companies came up with a method for improving a robotic unit’s ability to handle and manipulate objects.

Intrinsic data was combined with DeepMind’s model knowledge to create a method of training a robot on human input using remote devices, which allowed for human skills to be translated and taught to a robot.