18606938. TECHNIQUES FOR TRAINING MACHINE LEARNING MODELS USING ROBOT SIMULATION DATA simplified abstract (NVIDIA Corporation)

TECHNIQUES FOR TRAINING MACHINE LEARNING MODELS USING ROBOT SIMULATION DATA

Organization Name

NVIDIA Corporation

Inventor(s)

Caelan Reed Garrett of Seattle WA (US)

Fabio Tozeto Ramos of Seattle WA (US)

Iretiayo Akinola of Seattle WA (US)

Alperen Degirmenci of Jersey City NJ (US)

Clemens Eppner of Seattle WA (US)

Dieter Fox of Seattle WA (US)

Tucker Ryer Hermans of Salt Lake City UT (US)

Ajay Uday Mandlekar of Cupertino CA (US)

Arsalan Mousavian of Seattle WA (US)

Yashraj Shyam Narang of Seattle WA (US)

Rowland Wilde O'flaherty of Seattle WA (US)

Balakumar Sundaralingam of Seattle WA (US)

Wei Yang of Lake Forest Park WA (US)

TECHNIQUES FOR TRAINING MACHINE LEARNING MODELS USING ROBOT SIMULATION DATA - A simplified explanation of the abstract

This abstract first appeared for US patent application 18606938 titled 'TECHNIQUES FOR TRAINING MACHINE LEARNING MODELS USING ROBOT SIMULATION DATA

The abstract describes a method for generating simulation data to train a machine learning model by creating simulation environments, tasks for a robot to perform within those environments, determining robot trajectories, and simulating the robot's movement.

• Method involves generating simulation environments based on user input
• Tasks are created for a robot to perform in each environment
• Operations are conducted to determine robot trajectories for tasks
• Simulation data is generated by simulating the robot's movement in the environment according to the trajectories

Potential Applications: - Robotics training and development - Autonomous vehicle testing - Industrial automation optimization

Problems Solved: - Providing realistic simulation data for machine learning model training - Enhancing robot performance through trajectory optimization

Benefits: - Improved accuracy and efficiency of machine learning models - Cost-effective training for robots in various environments - Enhanced safety in testing scenarios

Commercial Applications: Title: "Advanced Simulation Data Generation for Machine Learning Model Training" This technology can be utilized in industries such as robotics, autonomous vehicles, and manufacturing for training and optimizing machine learning models.

Questions about the technology: 1. How does this method improve the accuracy of machine learning models trained on simulation data? 2. What are the potential cost savings for companies implementing this technology in their robotics development processes?

Original Abstract Submitted

One embodiment of a method for generating simulation data to train a machine learning model includes generating a plurality of simulation environments based on a user input, and for each simulation environment included in the plurality of simulation environments: generating a plurality of tasks for a robot to perform within the simulation environment, performing one or more operations to determine a plurality of robot trajectories for performing the plurality of tasks, and generating simulation data for training a machine learning model by performing one or more operations to simulate the robot moving within the simulation environment according to the plurality of trajectories.