Although the technological iteration of humanoid robots still requires time, two forces in the current market are trying to accelerate their transformation from0to1to achieve commercialization and implementation of products.
What are the respective advantages and disadvantages of large manufacturers, new energy vehicle companies, and specialized robot companies in developing humanoid robots? When is it expected that humanoid robots will truly enter automotive assembly workshops? If measured by100%the criteria for implementation, what level has the current technology and product reached? What challenges still exist in the current automotive manufacturing scenarios? Compared with existing mature industrial robots, what are the advantages of humanoid robots on automotive assembly lines? Will general-purpose humanoid robots replace industrial robots? What is the current level of the 'brain' of humanoid robots on the market? What are the requirements for large models to be adapted to the 'brain' of humanoid robots?
Frost & Sullivan (Frost & Sullivan, hereinafter referred to as 'Frost & Sullivan') Executive Director for Greater China, Cui Nan, was interviewed by the Financial Times magazine to discuss the above topics.
Financial Times
*Click at the end of the article Read the original article to view the complete report
Q： What are the respective advantages and disadvantages of large manufacturers, new energy vehicle companies, and specialized robot companies in developing humanoid robots?
The advantage of new energy vehicle companies lies in their control over hardware costs and their mature experience in autonomous driving applied to the development and manufacturing of humanoid robots, but there may be limitations in downstream scenarios. The advantage of specialized robot companies lies in their rich understanding of the industry, having accumulated a large number of professionals and specialized robot technologies earlier, but they also need capital support. Large manufacturers, relying on their ecological advantages and integration capabilities, can quickly conduct robot research and development and implementation, but they started relatively late and lack first-mover advantages.
Q： When is it expected that humanoid robots will truly enter automotive assembly workshops? If measured by100%the criteria for implementation, what level has the current technology and product reached?
If measured by100%the criteria for implementation, what level has the current technology and product reached?The standard is currently at30%and it is expected that it will take5-10
years to see humanoid robots widely applied in automotive assembly workshops.Q What challenges still exist in the current automotive manufacturing scenarios?
The first is the application challenge: The industrial manufacturing field has very strict requirements for the precision and efficiency of robot work, making it difficult for humanoid robots to match the already mature industrial mobile robot products; the second is the cost challenge, as advanced technology and complex engineering make the production cost of humanoid robots higher than that of mature industrial robots, which limits their popularity and practicality at this stage.
Q： Compared with existing mature industrial robots, what are the advantages of humanoid robots on automotive assembly lines? Will general-purpose humanoid robots replace industrial robots?
Many tasks on the automotive assembly line, such as installing car interiors, applying car stickers, and gluing glass windows, are difficult for ordinary industrial robots to replace humans with. This is because the structure of a car is designed entirely according to the characteristics of human movement, and ordinary industrial robots cannot bend and squat freely like humans to perform some assembly work. Relatively speaking, humanoid robots have more flexible movement capabilities and are suitable for some parts of the automotive manufacturing line. Humanoid robots will not completely replace the roles of existing industrial robots; they will more likely automate positions still operated by humans. The value of humanoid robots lies in maximizing the utilization of existing infrastructure in human society, simulating human activities to the greatest extent possible, and achieving automation and intelligent improvement.
Q： Most in the industry believethat both 3Cand automotive manufacturing are suitable for first implementation, but from the current dynamics of robot manufacturers, it seems that they place more emphasis on automotive manufacturing scenarios. So compared to3C, does automotive manufacturing have more unique advantages?
From a technical perspective,3Cscenarios have more refined parts compared to automotive, and there are higher requirements for the precision of robot execution capabilities, so manufacturers will give priority to automotive manufacturing scenarios.
Q： What is the current level of the 'brain' of humanoid robots on the market? What are the requirements for large models to be adapted to the 'brain' of humanoid robots?
The main technology of a robot's brain is artificial intelligence. Artificial intelligence is mainly divided into three stages: narrow AI, general AI, and super AI. Narrow AI, also known as weak AI, refers to AI that performs specific tasks or limited-range tasks, such as facial recognition and voice recognition. This is the most common type of AI and the level that current humanoid robot 'brains' can achieve.
As an ideal carrier of embodied intelligence, humanoid robots not only need to break down a specific task into specific subtasks at the 'task level' but also need to effectively execute these subtasks to interact with the physical world. Therefore, the execution success rate of a large model of a robot is an important indicator for measuring the degree of robot intelligence. To train the execution success rate of a large model of a robot, the robot needs to have autonomous and reliable decision-making capabilities, multimodal perception capabilities, and precise operation control capabilities. These capabilities require the robot to collect a large amount of data, train models, and conduct simulation tests to truly adapt the large model to the 'brain' of humanoid robots.
*This interview has been published in the Financial Times magazine, and the reporter is Yang Liu , with the original title ' Humanoid Robots Enter Factories for 'Internship', Do Automotive Manufacturers Still Need Workers? >
