Three key technologies and challenges of humanoid robots

Author: Liu Wei

On November 2, 2023, the Ministry of Industry and Information Technology published the "Guiding Opinions on the Development of Humanoid Innovation" on its official website, proposing that advanced technologies such as humanoid robot integration, high-end manufacturing, and new materials are expected to become disruptive after computers, robots, and automobiles, and will profoundly change human production and lifestyles and reshape the global industrial development pattern. China plans to initially establish a humanoid robot innovation system by 2025, achieve breakthroughs in a number of key technologies such as "brain, cerebellum, and limbs", and ensure the safe and effective supply of core components. The whole machine product reaches the international advanced level and realizes mass production, and is demonstrated and applied in special, manufacturing, and people's livelihood services, exploring and forming effective governance mechanisms and means.

1. A brief history of humanoid robots

Humanoid robots are robots that look like humans. They can interact and communicate with humans by imitating human appearance, movements and expressions. Humanoid robots usually have multiple sensory abilities such as vision, hearing, speech, touch, and can adapt to different environments and tasks through autonomous decision-making and learning. The development history of humanoid robots can be traced back to ancient times, but the real development began in the 20th century.

The following are some important milestones in the development of humanoid robots: (1) 1920s: The first humanoid robots were introduced. In 1921, the word "robot" was first mentioned in the play "Rossum's Universal Robots" by Czech writer Karel Capek. The word "Robota" in the title was used to describe a cyborg assembled from biological parts - a slave serving humans. The word later evolved into Robot, becoming a synonym for artificial humans and robots. The play aroused people's attention to the concept of robots. (2) 1950s: The rise of ergonomics research. With the advancement of technology, people began to study how to design robots to imitate human appearance and movements. Robots of this period were mainly used for simple tasks on industrial production lines.

(3) 1960s: The rise of Japan’s robotics industry. Japan became one of the leading countries in the development of humanoid robots. During this period, robots that could imitate human walking and lifting objects began to appear.

(4) 1980s: Rapid development of robotics. With the advancement of computer technology, the level of robots has been improved. During this period, some robots with artificial intelligence and functions appeared.

(5) 2000s: Development of special-purpose robots. In addition to industrial and scientific research, humanoid robots have also begun to be used in other fields, such as medical care, education, and entertainment. These robots are able to interact with humans in a more complex way.

(6) Since the 2010s: Humanoid robots have become a research hotspot. The research on humanoid robots in the fields of scientific research and technology has received widespread attention. Some humanoid robots with highly complex motion control and emotional communication capabilities have begun to appear. At present, the development of humanoid robots is constantly moving forward. With the continuous progress of artificial intelligence and other fields, the intelligence and realism of humanoid robots will continue to improve, bringing more convenience and innovation to human life. In ancient China, there was no concept of robots like the modern one.

However, ancient China had excellent craftsmanship and technology, and there were some automated machines that could be modeled after human form, including dancing robots during the reign of King Mu of Zhou in the Western Zhou Dynasty and wine pouring robots during the Tang Dynasty. Although these ancient mechanical machines are not called robots, they demonstrate the interest and creativity of ancient Chinese people in automated machines. These machines are not just works of art or entertainment tools, but also the crystallization of ancient science and technology and culture. Recently, some domestic companies such as Xiaomi, UBTECH, Dahua, Fourier, Dreame, and Zhiyuan are actively deploying humanoid robots. Junpu and ByteDance are also planning to enter the robot market to explore the use of large model capabilities in robots.

2. Three key technologies and challenges of humanoid robots

(1) Servo control. Humanoid robots need to have precise motion and posture control capabilities to mimic human movements and actions. Servo control technology can achieve fine control of the joints and body parts of humanoid robots, enabling them to perform various complex movements such as walking, running, jumping, etc.

(2) Artificial intelligence. Humanoid robots need to have artificial intelligence capabilities to be able to perceive and understand the surrounding environment and make corresponding responses and decisions. By using machine learning and other technologies, humanoid robots can learn and adapt to different environments and tasks, thereby achieving more intelligent behavior.

(3) Motion control. The motion control technology of humanoid robots is the key to achieving their complex movements and flexible mobility. Motion control technology can ensure the stability and balance of humanoid robots, enabling them to maintain balance in different terrains and environments and perform various actions and movements, such as avoiding obstacles and climbing. In order to achieve efficient motion control, and are usually used to monitor and adjust the robot's motion state. The above three major issues can be further broken down into the six main challenges faced by humanoid robot research and development:

(1) Power and energy management. Humanoid robots usually require a lot of energy to drive various electric joints and sensors, so how to achieve efficient energy management is an important challenge.

(2) Mechanical design and motion control. Humanoid robots need to have complex mechanical structures and precise motion control capabilities to mimic human movements and actions. Solving these problems requires solving complex problems such as dynamics, kinematics, and control theory.

(3) Perception and sensory processing. Humanoid robots need to be able to perceive their surroundings and their own status, for example, using visual sensors, tactile sensors, and sound sensors. At the same time, they also need to process this sensory data to make appropriate decisions and behaviors.

(4) Intelligent decision-making and planning. Humanoid robots need to have intelligent decision-making and planning capabilities and be able to make correct choices and behaviors in different environments. This involves research in areas such as machine learning, artificial intelligence, and motion planning.

(5) Human-machine interaction and safety. Humanoid robots are usually designed to interact and cooperate with humans, so they need to be able to understand human language and behavior and be able to interact with humans safely. Ensuring the safety and user experience of humanoid robots is an important challenge.

(6) Cost and availability. The research and development costs of humanoid robots are generally high, and there is currently no popular consumer market. Addressing these challenges requires reducing costs, improving availability, and promoting the application and commercialization of humanoid robot technology.

3. Bottlenecks in humanoid robot perception, understanding, and decision-making

The bottlenecks of humanoid robots' perception, understanding, and decision-making mainly include limitations in perception, language understanding, decision-making, and social interaction. To this end, humanoid robots need to establish accurate environmental models and be able to analyze and understand the scene. This requires the use of machine learning and reasoning technology, so that robots can predict and speculate on the current environmental state and possible dynamic changes based on historical data and prior knowledge. The bottlenecks of humanoid robots' perception, understanding, and decision-making are specifically manifested as:

(1) Limited perception capabilities. Humanoid robots need to be able to perceive various aspects of their surroundings, including vision, hearing, touch, etc. However, the current perception capabilities of robots are far inferior to those of humans, and they are unable to obtain accurate perception information from complex scenes like humans.

(2) Understanding language and context. Humanoid robots need to be able to understand human language and contextual information and extract key information from it. However, natural language understanding still has difficulties in processing complex semantic and pragmatic information, and robots often cannot accurately understand human intentions and needs.

(3) Insufficient decision-making ability. Humanoid robots need to be able to make reasonable decisions based on the information they perceive and understand. However, when faced with complex and uncertain environments, robots often find it difficult to make accurate and adaptable decisions. This involves improving the robot's reasoning and planning capabilities.

(4) Limited social interaction capabilities. Humanoid robots need to have good social interaction capabilities, be able to understand human emotions and intentions, and communicate with humans in an appropriate manner. At present, robots' capabilities in this regard are still limited, and they are unable to communicate emotions and interact socially like humans. In summary, improving the situational awareness capabilities of humanoid robots requires combining multiple key technologies such as sensor technology, algorithms, data fusion and processing, environmental modeling and scenario analysis, and real-time decision-making and planning. The continuous development and innovation of these technologies will provide greater room for improvement in the situational awareness capabilities of humanoid robots.