LG Electronics Deepens Robot Training Data Push With Nvidia, Targets 12 Years' Worth by Year-End

LG Electronics announced on the 18th that it is deepening its collaboration with Nvidia to expand the training data needed for robot development. The two companies plan to broaden the buildout of training data for physical AI-based robot development and use data collected from real manufacturing and everyday environments.

They also plan to add amplified and synthetic virtual data, bringing the total volume of training data to 100,000 hours by the end of the year. The company plans to use the secured data to advance its Robot Foundation Model (RFM).

To that end, LG Electronics invited Nvidia officials to the 'LG Data Factory' it is building in Seocho-gu, Seoul, to discuss the direction of collaboration in robotics and possible commercialization plans. On the 18th at LG Data Factory in Seocho-gu, Seoul, Cho Joo-wan, CEO of LG Electronics, and Madison Huang, senior director at Nvidia, reviewed data augmentation and synthetic solutions.

The meeting was attended by LG Group executives including Cho Joo-wan, CEO of LG Electronics, Hyun Shin-kyun, CEO of LG CNS, and Jung Soo-hyun, head of LG Sciencepark. LG and Nvidia signed a memorandum of understanding (MOU) on the 13th, local time, at Nvidia headquarters in Santa Clara, California, on future strategic business cooperation.

LG Electronics and Nvidia plan to combine data gathered from manufacturing and logistics sites with Nvidia robotics technology for use in robot training. The two companies are also building a system to augment and synthesize data collected in real-world environments, establish an expansion framework for training data, and feed the expanded data into robot learning.

The key base for this work is the LG Data Factory under construction in Yangjae, Seoul. The LG Data Factory spans four floors, from one basement level to three above-ground floors, with a total floor area of 10,000 square meters (about 3,025 pyeong), and LG Electronics is aiming for full operation by year-end.

LG Electronics is using its in-house home robot, LG CLOiD, at the Data Factory to generate and collect data. In a space that replicates a home environment, CLOiD performs cleaning tasks, while in a manufacturing environment modeled after a washing machine plant in Tennessee, the robot carries out part transport, stacking, and assembly tasks. In addition, the company is securing robot input data in the LG CNS logistics automation solution area and in LG Innotek's robot hand training space.

LG Electronics is using Nvidia physical AI technology to expand collected data. The technologies used are Nvidia Omniverse libraries, Cosmos open-world models, and Isaac open robotics development platforms. The company plans to use these for data augmentation and synthesis, as well as robot training, to scale up the volume of training data.

LG Electronics plans to increase the number of robots deployed at the Data Factory to the hundreds by year-end. It also plans to combine data collected directly with amplified and synthetic virtual data. Through this, the total volume of training data expected to be secured by year-end is 100,000 hours.

Large-scale training data buildout will draw on task data accumulated at LG Electronics' manufacturing and logistics sites, as well as the working methods of skilled workers. LG Electronics has work data accumulated over decades. By combining Nvidia robotics technology with its own data and work methods, LG Electronics plans to build a data virtuous cycle in which data collection, training, and performance improvement are repeated.

The secured data will be used to advance the Robot Foundation Model (RFM), which determines the performance of humanoid robots.

LG Electronics is expanding its robotics business organization. In June, it announced the launch of a Robotics Business Center directly under the CEO, and the organization went into operation on July 1 to oversee the company's robot business. LG Electronics is positioning itself to expand its business area from industrial and commercial robots to home robots, and it is working to strengthen its robotics capabilities by linking actuators and other core robot components with the Data Factory.

Cho Joo-wan, CEO of LG Electronics, said the company intends to secure competitiveness in physical AI through One LG synergy that brings together the group's core capabilities and through strategic collaboration with global partners. He also set out a goal of transforming into a total robotics solution provider.

Source: IT DAILY · Kim Byung-joo
Original: https://www.itdaily.kr/news/articleView.html?idxno=241047

References

• https://www.itdaily.kr/news/articleView.html?idxno=241047