The closing ceremony of the 2025 World Artificial Intelligence Conference (WAIC) and High-Level Meeting on Global AI Governance was held at the Shanghai World Expo Exhibition and Convention Center on July 28. Hu Yong, Deputy District Mayor of Jing'an District, attended the closing ceremony and signed a contract with Xinzhi Embodied AI Technology Co., Ltd. (an industrialization project developed by the Institute of Trustworthy Embodied AI (TEAI), Fudan University).
The high-precision, cost-effective "Adaptive Vision-based Tactile AI Sensor", independently developed by the Xinzhi Embodied AI team from the Institute of TEAI at Fudan University, made its public debut at the 2025 WAIC.
As a representative university industrialization project, this achievement was unveiled in the central exhibition area of the special WAIC Skills Stage as one of the highlight projects at the 2025 WAIC. It serves as a critical technical foundation for developing the next generation of embodied AI large models designed for refined operations and brings humanoid robots closer to human life.
The team's self-developed vision-based tactile sensing technology enables robots to possess "tactile nerves" akin to human skin. The tactile gripper powered by this advanced technology can accurately detect subtle contact forces. As a result, it can steadily pick up fragile items such as tender tofu, potato chips, and jellies and arrange them in various shapes. With a vision-based tactile sensor, the gripper ensures gentle yet precise movements.
The essence of this technology lies in capturing the deformation of tiny particles within special materials when subjected to force through a built-in camera. AI models then convert complex tactile signals into high-dimensional visual data, enabling precise reconstruction of stress conditions in all directions.
The gripper boasts three key technical highlights: First, its flexible interface is soft and conforms seamlessly to human skin while being highly wear-resistant and adaptable to complex surfaces. Second, it can sense multi-dimensional forces, including normal pressure, tangential force, and even torsional force, enabling it to perceive every touch and differentiate between various movements. Lastly, it offers exceptionally high sensitivity, with an impressive spatial resolution of 40,000 sensing points per square centimeter, surpassing even the perceptual accuracy of human fingertips. During the on-site demonstration, the gripper gently grabbed a piece of tender tofu and a piece of jelly without causing any damage, instantly captivating the audience.
Currently, this innovative vision-based tactile sensing technology has been successfully applied to robot dexterous hands and grippers, which perform well in tasks such as high-precision sorting and refined operations. Compared with traditional electronic skin, this technology reduces costs by nearly 50-fold while significantly enhancing performance, exhibiting promising potential for widespread application and industrialization.
The newly released high-precision, cost-effective vision-based tactile sensor will further integrate the tactile multimodal dual-brain architecture, XR high-precision robot/robotic arm teleoperation software and hardware system, and world model real-time optimization learning system. This integration will provide critical technical support for developing the next generation of embodied AI large models designed for refined operations.
The Xinzhi Embodied AI was established by the Institute of TEAI at Fudan University. The Institute of TEAI is committed to advancing cutting-edge theoretical research, key technology breakthroughs, and achievement transfer and commercialization in the field of embodied AI. By integrating multidisciplinary expertise in areas such as computer vision, natural language processing, robotics, control systems, and ethics of science and technology, it develops intelligent agents with capabilities for independent exploration, continuous evolution, and alignment with human values. These efforts aim to provide a core driving force for future human-machine collaboration and the development of an intelligent society.