The Atlas humanoid robot stops in front of a car and bends its waist. It grabs the wheel and tire with both arms and lifts them with force. At the workbench, it picks up parts and moves them elsewhere, and around the vehicle, it moves alongside another Atlas. Researchers standing right next to it check monitors and observe every single movement of the robot.
Atlas, which used to show off fancy tricks, is now in the middle of working on training to be deployed into the automobile production site.
According to the industry on the 1st, Boston Dynamics released a documentary titled ‘At the Forefront of Robotics Competition’, co-produced with Hyundai Motor Group and Bloomberg Media Studio on the 30th of last month.
The shooting location is the ‘Robotics Metaplant Application Center (RMAC)’ established at the Hyundai Motor Group Metaplant America (HMGMA) in Georgia, USA. Having started operation last month, this place lets Atlas learn necessary tasks for automobile manufacturing and verifies its performance before sending it to the actual production line.
Inside the RMAC shown in the video, automobile bodies, parts shelves, and workbenches are arranged similarly to a production site. Atlas constantly repeats the action of picking up and transporting objects while moving around this place. Scenes of handling heavy objects like wheels and tires and multiple units working in the same space were also captured.
Boston Dynamics selected the automobile manufacturing industry, which requires a high level of safety and reliability, as the first application field for Atlas. After verifying performance in actual production environments, it plans to expand application areas to other industries such as food and beverage and semiconductors.
Artificial intelligence (AI) plays a key role in helping Atlas learn unfamiliar tasks. It works by combining the ability to maintain balance, move arms and legs, and the ability to understand instructions to judge the next action. Moving away from the conventional method of inputting predetermined actions one by one, AI allows various tasks to be learned and responds to environmental changes.
Securing data in actual fields is also a major role of RMAC. Boston Dynamics explained that physical AI, like large language models (LLMs), improves work capacity as data increases, and Atlas must gain experience outside the lab.
Lowering costs for mass production is also underway. Boston Dynamics stated that it has reduced some component costs by 60-80% in the process of converting prototypes into mass-produced products.
Hyundai Motor Group has built a facility in the US capable of producing up to 30,000 robots annually and plans to deploy 25,000 Atlas units to global business sites in the future.
