The track that Tesla, Xiaopeng, and NIO are all aiming for, squeezing into traditional major manufacturers | Hardcore Analysis

Author｜Huang Nan

Editor｜Yuan Silai

All along, the successful implementation has been a life-and-death challenge for robot companies. Among many scenarios, the industrial manufacturing represented by automotive workshops is one of the most common implementation scenarios.

Compared to complex scenarios such as outdoors, households, and supermarkets, the workshop environment and processes are relatively stable and simple, with abundant real data that can provide a foundation for robot training. The technical commonality and the transferability on the supply chain side have led many automakers with inherent data advantages to focus on this new market.

Now, formidable opponents are beginning to enter this track.

Recently, GAC Group released the independently developed third-generation embodied intelligent humanoid robot GoMate. GoMate pioneered a variable wheel-foot mobile structure, integrating the two mainstream routes of wheeled and legged types. By switching between two-wheel feet and four-wheel feet, it can move on various terrains such as flat ground and slopes, with an all-terrain traffic capacity.

Unlike some enterprises that take automotive production as the preferred implementation scenario, GAC Group's humanoid robot focuses on security, health care, and automotive aftermarket service scenarios in the first stage of commercialization, and is currently in field testing.

Zhang Aimin, the head of robot research and development at GAC Group, told Yingke, "Automobile manufacturing is not the scenario that GAC prioritizes to explore. We hope to focus our energy in the early stage on directions with more urgent and direct needs. Taking security as an example, it can be clearly seen that the existing passive security will gradually be replaced by active security. Through robots actively detecting dangers and initiating responses, we believe this is a huge market."

Yingke learned that GAC Group's security humanoid robot will be launched in the first half of 2025; GoMate will also conduct global sales of self-developed components in batches in the middle of the year, and complete small-batch production and gradually expand mass production in 2026.

GAC's entry is not early. Since 2023, the popularity of embodied intelligence has been continuously rising. Including Tesla, BYD, XPeng, NIO, Chery and other automakers, they have successively released their R & D dynamics in the embodied field, recruiting engineers to form R & D teams, and launching a number of embodied robot products.

Compared to startups, automakers with money, people, and scenarios have unparalleled advantages. Their entry also adds more intensity to the embodied robot track.

Multi-chain Reuse from Software to Hardware

Automakers' layout in the embodied intelligence track seems to be able to take some shortcuts.

Firstly, at the software algorithm level, they can smoothly use the end-to-end large models in autonomous driving in recent years. Based on GAC's autonomous driving technology, Zhang Aimin's team introduced the traditional perception-decision-control route and the end-to-end route in the training of GoMate.

By integrating the perception and control modules in the system into a unified world model, after inputting the original data, the final trajectory or control instructions can be output, thereby improving the universality of the embodied intelligent system, the reliability of decision-making, and the flexibility of actions.

Among them, there are obvious differences in the application of the same technical route in autonomous driving and humanoid robots. "The control dimension of autonomous vehicles is relatively simple, and data collection is relatively easy, but the requirements for functional safety are extremely high. Conversely, the technical difficulty of humanoid robots is greater, especially in data collection and skill generalization. The advantage of safety makes it more effective in specific application scenarios and task performances," Zhang Aimin said.

GAC's Third-generation Embodied Intelligent Humanoid Robot GoMate (Image Source/Enterprise)

On the hardware side, thanks to the mature automotive industry chain, automakers including GAC have an inherent advantage in cost control.

Taking key components as an example. The components of GAC's humanoid robot are mainly composed of two parts. One is the procurement of basic components. Relying on GAC Group's supply chain system, the procurement cost can be controlled within a lower expenditure range; the other part, including the driver, motor, dexterous hand and other core components, are independently developed and produced by GAC. By reusing some automated production lines and modules, combined with controllable raw material costs, the overall cost of the humanoid robot can be effectively reduced.

GoMate Integrated Joint Motor (Image Source/Enterprise)

Currently, many automakers are accelerating the layout pace in the embodied intelligence track. Dongfeng Liuzhou Automobile, BYD, NIO and other automakers have cooperated with UBTECH. UBTECH's Walker S series humanoid robots have entered multiple factories for practical training; BYD has released relevant recruitment information to form an embodied intelligence research team; and BAIC Group's BAIC Capital has also joined the battlefield by investing in multiple embodied enterprises such as Galaxy General and Pacinian Perception Technology.

But a reality is that the market's imagination of the landing space for humanoid robots is relatively limited. Before moving to the next stage, it is crucial to expand the boundary of the landing scenarios as much as possible and conduct integration, polishing and iteration based on real tasks.

From Passive Security to Active Security

According to Goldman Sachs' estimation, it is expected that in 2025, a small number of well-funded humanoid robot companies are expected to enter commercial pilots. Among many landing directions, GAC Group focuses the early application scenarios of GoMate on security, health care, and the automotive aftermarket, with particular attention to the security field.

Zhang Aimin told Yingke that in recent years, important changes are taking place in the security field, and the iterative demand for the previous generation of security equipment has begun to emerge. Compared to the passive security with blind spots and post-event responses, active security will become the mainstream mode.

By integrating sensors, AI large models and multi-layer defense systems, the GoMate security robot can conduct spontaneous patrols in the managed area. Based on real-time monitoring and environmental data analysis at the algorithm end, it can actively identify abnormal behaviors or potential threats and trigger corresponding security measures.

For example, in Community A, there are several stores that store and sell flammable and explosive items. When the weather is dry, the GoMate security robot can spontaneously increase the inspection frequency of fire-fighting equipment such as fire hydrants in this area. Based on environmental factors such as temperature and humidity, it can predict the probability of fire occurrence and prevent the occurrence of potential threat events.

In addition, the GoMate security robot can also achieve active intelligent multi-level defense, quickly authorize smart wearable devices, and be compatible with multiple interfaces such as AI + monitoring systems, smart gates, and drones. For example, replacing the dexterous hand of the GoMate security robot with a shield, baton, etc., to ensure the work efficiency and personal safety of security personnel.

The Application of GoMate in the Security Scenario (Image Source/Enterprise)

The energy consumption problem not only relates to the operating cost and maintenance cost of the robot, but also directly affects the sustainability of its work. Currently, GoMate can achieve a continuous working time of more than 6 hours. When the device stays on a flat road surface and switches to the four-wheel mode, in the standby state, other motors of GoMate will automatically power off, saving more than 80% of energy compared to similar products.

Flush data shows that the annual output value of China's intelligent security industry in 2023 is approximately 1,010 billion yuan, with an increase of 6.8%; among them, the market scale of intelligent security software and hardware is expected to increase to 91.3 billion yuan in 2025.

"Security is a rigid demand. As users pay more and more attention to safety and privacy, the development space of active security is accelerating. From a commercial perspective, the demand for maintaining stability and security in most parks and buildings is constant, which also brings a certain market order guarantee for the early landing of the industry," Zhang Aimin said.

Currently, GAC Group is cooperating with the leading enterprises in the security industry in Guangzhou to upgrade the passive security system to an active security defense system.

The Application of GoMate in the Industrial Scenario (Image Source/Enterprise)

Zhang Aimin told Yingke that the GoMate security robot has a low center of gravity and strong stability. The four-wheel-foot mode is mostly used to deal with complex road surfaces and extreme working conditions in unexpected situations, and it is not easy to tip over even in the case of personnel pushing, making it more suitable for large outdoor parks. The two-wheel-foot mode is flexible and efficient in actions, and can complete more daily tasks.

This robot is currently undergoing practical training in GAC's own park. After the product stability is further improved, it is considered to expand the external deployment, and it is planned to be released in the first half of this year. In addition, GAC Group also plans to launch specific service robots for different scenarios such as communities with a higher population density and medium-sized communities.

In other implementations, GAC Group will not focus on the layout of automobile main engine plant production and manufacturing in the short term. In addition to security, it will focus on health care and the automotive aftermarket. For example, GoMate will be introduced in 4S stores and exhibition halls for guidance and service work. In the future, more scene applications such as logistics and education will also be explored.