Humanoid robot training in factories (recording China)
2024-08-10
한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina
The picture shows UBTECH's industrial version of humanoid robot entering the workshop for training.
Photo by Zheng Xuexiong (People's Vision)
Taking small steps, the humanoid robot came to the left side of the car body and quickly completed the quality inspection of the front and rear door locks. After stopping, it stretched its arm into the car body, rotated its wrist backward 180 degrees, pinched the seat belt with five fingers and pulled it down, and the seat belt function test was completed. These test results have been uploaded to the factory's intelligent manufacturing management system in real time.
Not long ago, UBTECH's industrial humanoid robots entered the final assembly workshop of the NIO factory in Hefei, Anhui for training, and collaborated with humans to complete quality inspections in the automobile production factory.
"Factory training means that humanoid robots step out of the laboratory and enter the real automobile manufacturing environment. By collecting scene data on the assembly line, they can understand the production needs of the enterprise and complete self-iteration and upgrade to prepare for formal 'job training'." said Jiao Jichao, vice president of UBTECH Robotics and executive director of the research institute.
Industrial robots such as industrial robotic arms, welding and painting robots are no longer uncommon in production workshops. What are the differences between humanoid robots?
In comparison, traditional industrial robots have a low level of intelligence and can only perform tasks in fixed processes. They find it difficult to effectively complete tasks in complex scenarios such as flexible assembly.
"We specialize in tasks that are difficult for traditional industrial robots to complete," said Jiao Jichao.
Lifting legs to walk, stopping and turning, visual inspection... these actions may seem simple, but the technical challenges are by no means easy.
Just like humans, humanoid robots cannot move without the assistance of joints. When the team first started developing humanoid robots, they were most worried about not being able to find suitable robot joints. "We couldn't buy them domestically, and the ones we bought abroad didn't meet our needs."
The servo drive is the joint drive of the robot and is also the key component that determines the robot's movement ability. There are as many as 50-100 high-power servo drive parts used in humanoid robots, and 40-80 low-power servo drive parts.
Through continuous technological research and development, UBTECH has completed the research and development and commercialization of small-torque micro servo drives to high-explosive large-torque servo drives, and increased the localization rate of servo drive components from 40% to more than 90%.
"The harmonic rotary servo drive has high torque density, small size, light weight and fast response speed. The robot's arms and legs are smaller, and it can walk more freely. It is more stable when carrying heavy loads such as handling and sorting." Jiao Jichao gave an example, "We take advantage of the robot's flexible disassembly feature to carry out modular design of end effectors such as 'hands' and 'feet'. For example, we design different 'hands' for the robot, such as the humanoid five-finger dexterous hand and the two-finger gripper. Through different connection methods, the end effector can be replaced at any time as needed."
When installing a car glass door seal, how can you complete the task with visual obstruction? How can you use data support to enable a humanoid robot to perceive the environment?
In addition to well-developed and dexterous "limbs", humanoid robots also need a smart and studious "intelligent brain". Talents in fields such as big data and artificial intelligence are trained to plan large models for humanoid robots in industrial manufacturing scenarios and optimize strategies through algorithms.
For example, semantic visual navigation technology is built to enhance the humanoid robot's understanding of the environment and scenes by extracting high-level environmental semantic information and establishing spatial topological relationships. "Through semantic navigation, the humanoid robot knows what it needs to do, schedules tasks autonomously, and performs different operations in different workshops," said Jiao Jichao.
Installation of window seals, assembly of engine hub motors, inspection of air conditioning condensate leaks... Different car companies have different needs for humanoid robots, so in-plant training is crucial. "Car companies provide verification and testing scenarios for humanoid robots, which can collect data, iterate algorithms, and improve performance." Jiao Jichao introduced that UBTECH cooperates with Dongfeng Liuzhou Motor, FAW-Volkswagen Qingdao Company, Geely and other automobile companies to update and iterate algorithms around real manufacturing scenarios to further improve performance. It is expected that the first batch of humanoid robots will be delivered this year and put into use in automobile manufacturing workshops.
It takes more than 7,000 parts to make a humanoid robot. In recent years, Guangdong Province has actively laid out the humanoid robot industry, a number of key technologies have made breakthroughs, and the supply chain has become increasingly complete. As of August last year, there were 39 related companies in Guangdong. In the first half of this year, the Guangdong Provincial Department of Industry and Information Technology announced an action plan to cultivate future intelligent equipment industry clusters, proposing that by 2035, Guangdong Province will be built into a global highland for the innovation and development of future intelligent equipment industries such as humanoid robots, aerospace equipment, deep-sea equipment, and deep-earth equipment. (People's Daily reporter Wang Yunna)
Source: People's Daily Online
