Robot services are becoming increasingly common in people's lives, but at present, they still need to improve in terms of force accuracy and control ability. Some robots may deform mineral water bottles when tightening their caps. To pick up fragile objects like eggs or cakes, a new part needs to be replaced.

At present, there are still many challenges in the industrialization of the core key technologies and components of embodied intelligence. At the recently held 2025 Enterprise Innovation and Integration Conference on Core components of body-type intelligent robots, Chen Xiaodong, co-founder of the Zhongguancun Rongzhi Special Robot Industry Alliance, stated that the current precision of robot perception still needs continuous improvement, the functions, response speed and flexibility of execution components need to be balanced urgently, and the computing power, energy efficiency and cost of edge computing still need to be optimized These are the key issues that stand in the way for smooth information interaction between robots and the physical world.

How can we address the problems of inflexibility, excessive force and loss of control in the interaction between robots and the physical world, and provide capabilities and services that match human expectations? The experts attending the meeting conducted professional discussions around the theme of "Tackling Core Technologies and achieving dynamic embodied intelligence".

Developing the "tactile sense" of robots

At present, robots have developed strong capabilities in vision and hearing, such as face recognition and semantic recognition, but they still have a weak point in terms of touch. Key components related to touch have become one of the important directions for innovation and entrepreneurship in the industry. Relevant experts indicate that there are currently over a dozen solutions to empower dexterous operations, ranging from embodied intelligence to dealing with very complex scenarios.

In the data-driven era, algorithms are becoming increasingly important. The ability of robots to interact needs to be acquired through continuous interaction with the environment, just like how a data-driven approach enables ChatGPT to emerge with "intelligence" in text.

In the past year, the "tactile sense" of robots has been developing. For instance, robots capable of playing the guzheng have deeply integrated the three modes of vision, hearing and touch. For instance, the Zhuoyide robot has demonstrated advanced training ranging from serving tea and water to being able to play table tennis. Meanwhile, Galaxy General has utilized data in the billions to achieve the ultimate state of robot grasping.

From grasping hard objects to controlling the force to grasp soft objects and then to being able to manipulate objects, this is the path for the continuous development of robots' "tactile sense" towards higher levels. The experts attending the conference believe that to do a good job with the "tactile" model, it is necessary for multiple disciplines and fields to conduct continuous cross-disciplinary and integrated research around common key issues.

Provide an innovative ecosystem for the advanced advancement of robots

The core of embodied intelligence lies in the agent's real-time interaction with the physical environment through advanced sensing technology and data processing algorithms, capturing information, analyzing situations, and making precise decisions.

Take agricultural robots as an example. Selective harvesting is a technical bottleneck of robots that has not yet been overcome worldwide. Li Wei, a professor from the College of Engineering at China Agricultural University, stated that the working environment in agriculture is complex. Robots need to be able to identify unstructured environments while analyzing and performing operations on various agricultural products. Force control must not only meet the operational requirements but also avoid causing irreversible damage to products such as fruits and vegetables.

Li Wei said that the team solved the above-mentioned problems by developing a rigid-flexible coupled picking end effector. The team optimized the finger design of the robot to minimize the contact parts while completing the operation tasks, facilitating the practical application of the picking robot.

Embodied intelligence is moving from the laboratory to reality, and the huge market potential it holds will bring profound innovation and transformation to all industries. At present, China has a relatively complete robot supply chain, and its domestic production capabilities in motors, sensors and AI chips are rapidly improving. In 2025, embodied intelligence and intelligent robots were written into the government work report for the first time, becoming a future industry that the country focuses on cultivating.

Industry insiders believe that the "intelligence" of robots is deeply rooted in their "bodies". Without sensitive "five senses" and strong "muscles and bones", it is difficult to achieve flexible and reliable intelligence. For embodied intelligent robots to achieve in-depth interaction with the physical world, their core components are irreplaceable.

The key to the development of the embodied intelligent robot industry lies in the independent control and breakthrough of performance of components such as motors, sensors, and AI chips. To promote scientific and technological innovation, aim at the leap in the "tactile" performance of robots, cost reduction and large-scale application, the China Association for Science and Technology actively organizes innovative explorations in aspects such as embodied intelligence technology innovation, industrial development and scene application, and promotes the deep integration of scientific and technological innovation and industrial innovation. The Qichuang Rongtong Hui Special Session on intelligent robots held this time, through links such as industrial problem argumentation, collection of outstanding solutions, and industrial connection, continuously leverages the leading role of experts and scholars at the forefront, promotes the driving effect of outstanding solutions, gathers elements such as scientific and technological achievements, industrial demands, and scene resources, collaboratively solves technical problems, and provides customized and specialized services Help robots achieve in-depth interaction with the physical world. (Reporter Zhang Jiaxing

Source: China Science and Technology Network