When Will We See Robot "Crown Jewel" Reach Human-Level Dexterity?

A Securities Times reporter, Wang Yiming

Why do people find it easy to grasp chopsticks and tie shoelaces, but difficult for machines to achieve?

From a construction perspective, the human hand is an outstanding achievement of evolutionary process, with a control region occupying about one-third of the motor cortex and sensory area of the brain. Beneath the skin, the human hand consists of 27 bones, 27 joints, 34 muscles, 123 ligaments, three main peripheral nerves, and numerous blood vessels and soft tissues, allowing it to produce approximately 24 degrees of freedom in motion.

For a dexterous hand robot, achieving the ability to execute high-level human-like operations poses multiple challenges. According to developers interviewed, on one hand, hardware design needs to ensure that the structure is robust enough to accommodate suitable components and generate sufficient flexibility, with overall difficulty comparable to any humanoid product. On the other hand, in software algorithms, it is crucial to extract effective information from massive high-dimensional sensor data and generate accurate action instructions in real-time for adaptive and dexterous interaction. For example, in multi-finger coordinated operations, dynamic planning of joint torques is required to avoid self-collision and adapt to different object shapes and weights, which also poses extremely high requirements on computing resources.

In fact, developing robot dexterity technology comparable to the human hand is a leading research direction at the intersection of intelligent robots, intelligent manufacturing, and artificial intelligence, which has been recognized as one of the most challenging scientific problems by Science. Therefore, the dexterous hand is also referred to as the "crown jewel" of robot technology.

From an industrial development perspective, in the 1980s, Utah University and MIT cooperated on the development of the Utah/MIT Hand, which made a breakthrough in joint design and sensing technologies, laying the foundation for dexterous hand design. By the end of the 20th century, with the development of embedded hardware technology, multi-fingered dexterous hands began to integrate more sensors and higher-level control systems. In recent years, AI has injected new momentum into dexterity technology, allowing these algorithms to train on big data and enable autonomous learning and adaptation to complex tasks; at the same time, dexterity hand design tends towards simplifying system structure to reduce costs and improve reliability.

Looking ahead, industry insiders generally believe that while hardware still needs continuous optimization and cost reduction, especially in mass production, most functions have already been achieved.

Over the past half century, scientists have been pursuing the creation of machines with hands as flexible as those of humans. Today, with AI developing at a rapid pace, many breakthroughs have been witnessed. Perhaps, the goal of achieving robot "crown jewel" technology will be realized in the not-too-distant future.