Tech

Humanoid Robots: Poised for Takeoff?

I
1. There are three main problems that have long plagued the upgrading and development of intelligent control of humanoid robots, and breakthroughs have been made in all of them. With the advancement of technology and market development, the manufacturing cost of humanoid robots is expected to drop from 6.7 million yuan to less than 200,000 yuan. The turning point of commercialization has been reached.

2. With the slowdown of global population growth and the intensification of population aging, the development of humanoid robots has ushered in new opportunities. Manufacturing factories with low levels of automation, such as automobile assembly workshops, have become ideal entry points for the commercial application of humanoid robots. Therefore, it can be seen that UBTECH and Tesla’s humanoid robots are the first to be piloted in these scenarios.

3. Two types of companies are expected to benefit the most in the entire humanoid robot industry chain. One is companies that make finished humanoid robots, such as UBTECH, Tesla, etc. This depends on whether the company’s product design can meet market demand. Can it be commercialized quickly to seize market share? Another type of enterprise is core component suppliers. This type of enterprise focuses on production capacity reserves and cost control capabilities.

II. Large model empowerment and cost control are the keys to commercialization
Since 1927, the development of humanoid robots has gone through four main stages: concept inception, technology exploration, technology growth, and technology integration. During the technology growth stage, the Pepper robot launched by SoftBank sparked a craze in the industry, but was eventually discontinued due to a lack of suitable commercial application scenarios. However, with the debut of Tesla Optimus, the market once again saw the commercial potential of humanoid robots, such as replacing some workers in Tesla factories. Elon Musk even claimed that Tesla’s future value will mainly rely on humanoid robots, causing the field’s attention to continue to rise.

The logic behind designing robots to be humanoid lies in their efficiency and economy. The original intention of robots was to replace human labor. Today, as the level of automation continues to increase around the world, many workplaces and supporting infrastructure are designed based on human habits. Therefore, robots designed to be humanoid can integrate into these environments more seamlessly, thereby achieving higher economic benefits and application value.

There are many humanoid robots on the market, and the market is focusing on general humanoid robots. This type of robot has the highest level of artificial intelligence and human-like structure. Breakthroughs in core technology and the assistance of artificial intelligence are crucial to realizing the realization of robots. Crucial. There are three main problems that have long plagued the upgrading and development of intelligent control of humanoid robots, and breakthroughs have been made in all of them:

(1) Imitation learning ability: This requires humanoid robots to have strong AI generalization capabilities and be able to achieve autonomous decision-making and learning upgrades by imitating human actions, thereby improving the integrity and consistency of task execution. Stanford’s Mobile ALOHA facilitates human-machine collaborative training and achieves major breakthroughs in imitation learning. Not only can it operate autonomously, but it can also support full-body remote control to complete complex tasks such as laundry, watering plants, and making coffee. It can also learn daily manual operations through neural networks. Behavior, outstanding potential in the field of housekeeping.

(2) End execution capability: This emphasizes the operational accuracy of the robot’s hand and requires that under the control of the central processor (i.e. “brain”), the robot’s action output must be precise to reduce errors and improve the correctness and accuracy of tasks. . Some dexterous hands at home and abroad have been mass-produced, and domestic products have significant price-performance advantages. For example, the domestic Yinshi robot dexterous hand has been mass-produced and its price has dropped to the level of 10,000 yuan. It can basically complete more than 70% of the grasping actions of the human hand.

(3) Perception-based motion control capabilities: This involves the robot’s all-terrain mobility capability. Similar to autonomous driving technology, humanoid robots need to be able to control their own movement based on the perception of the surrounding environment to enhance their all-round operational capabilities. Improve the timeliness of task completion. The virtual environment algorithm provided by NVIDIA accelerates the improvement of robot perception capabilities. Isaac Sim creates “lifelike digital twins” to facilitate the design of AMR robots. Isaac AMR accelerates the deployment of fully autonomous mobile robots and reduces costs.

In terms of cost, with the advancement of technology and market development, the manufacturing cost of humanoid robots is expected to be reduced from 6 to 7 million yuan to less than 200,000 yuan. The turning point of commercialization has been reached. Measures to reduce the cost of humanoid robots mainly come from two major aspects: First, mass-produced humanoid robots reduce the cost of a single robot through scale effects. The second is the core components of domestically produced humanoid robots, such as joints, drives, reducers, etc. According to research, when the production volume of humanoid robots reaches 10,000 units, the cost of Ubiselect’s humanoid robots is expected to be reduced to 200,000. Kepler Robotics mentioned that it plans to control the price of humanoid robots to about US$20,000-30,000 (equivalent to RMB 140,000-210,000 yuan). Zhiyuan Robot estimates that the manufacturing cost of its first humanoid robot, the Expedition A1, will be controlled within 200,000 yuan in the future. The cost of the star company Tesla’s “Optimus Prime” humanoid robot is expected to be US$20,000 (approximately RMB 140,000). Yuan) or so.

III. The market for humanoid robots is vast, and commercialization is expected to start in the manufacturing industry.
With the slowdown in global population growth and the intensification of population aging, the development of humanoid robots has ushered in new opportunities. The global population generally shows a downward trend. Except for the immigration wave caused by the war in Germany, the population growth rate of other major countries is generally lower than the global average. Especially in Japan, South Korea and the United Kingdom, the negative population growth rates in these countries are increasing year by year, and China will also begin to experience negative population growth in 2022. At the same time, the proportion of the world’s elderly population aged 65 and above continues to rise. The degree of aging in all major countries is higher than the global average. Among them, Japan has the most serious aging problem, with the proportion of the elderly population approaching 30%, followed closely by the United States. The aging ratio exceeds 20%. Expected future application scenarios for humanoid robots include:

(1) Slowing population growth and aging trends have intensified the tightness of social labor supply, leading to rising labor costs for enterprises, thereby increasing the demand for robots to replace human labor.

(2) With the increasing demand for rehabilitation and elderly care brought about by aging, the market demand in the fields of medical health and elderly care is growing. According to forecasts from the Organization for Economic Cooperation and Development (OECD), by 2030, the average annual growth rate of the global population aged 65 and over will reach approximately 3%, and the gap in elderly care workers will further expand.

Manufacturing is the first scene where humanoid robots were put into practice. The manufacturing industry requires humanoid robots to perform repetitive and precise operations, while the home service field places greater emphasis on the functional diversity, high degree of autonomy, and smooth interaction with humans of humanoid robots. Therefore, the service industry requires humanoid robots to perform more complex operations with human users. and direct interaction. In view of the current technical limitations of humanoid robots in terms of interaction capabilities and collaborative work capabilities, market penetration strategies should start from scenarios where the environment is relatively closed and the processes are relatively simple and standardized. Manufacturing factories with low levels of automation, such as automobile assembly workshops, have become ideal entry points for the commercial application of humanoid robots. Therefore, it can be seen that UBTECH and Tesla’s humanoid robots are the first to be piloted in these scenarios.

However, with technological breakthroughs and the performance of humanoid robots gradually improving, commercial implementation scenarios will increase. Therefore, the market penetration of humanoid robots can be divided into three stages:

(1) Taking manufacturing factories as the entry point, it mainly replaces simple and repetitive labor;

(2) With product improvements and cost reductions, humanoid robots have begun to enter more commercial scenarios such as restaurants and shopping malls, and their application in factory scenarios has also been strengthened;

(3) Realize the combination of general artificial intelligence (AGI) and humanoid robots to reach the level of embodied intelligence. People’s acceptance of humanoid robots has reached its peak, and robots have begun to widely enter the home environment.

IV. Which directions in the industrial chain will benefit most?
Two types of companies are expected to benefit the most in the entire humanoid robot industry chain. One is companies that make finished humanoid robots, such as UBTECH, Tesla, etc. This depends on whether the company’s product design can meet market demand and whether it can Quickly implement commercialization to seize market share. Another type of enterprise is core component suppliers. This type of enterprise focuses on production capacity reserves and cost control capabilities.

Based on the design of the first-generation Optimus Prime scheme and taking into account the mass production scale effect, Great Wall Securities analyzed the long-term cost of each component, among which the joint assembly (rotary actuator + linear actuator + hand joint assembly) accounted for 61 %, intelligent systems are 5%, others are 13%, manufacturing costs are 14%, and software costs are 7%. In terms of cost proportion of each core component, six-dimensional torque/torque sensor (17%), lead screw (14%), frameless motor (11%), harmonic/planetary reducer (9%), coreless motor (7%), chip (3%), IMU&encoder (2%).

error: Content is protected !!