Hard Kr Exclusive | Tsinghua's Post-2000s Team Secures Investment from F&G and Pokka Robotics to Develop Ultra-Thin Visuo-Haptic Sensors
Recently, Huiguang Innovation, a provider of robotic tactile sensors and tactile data solutions, has successfully completed seed and angel rounds of financing worth tens of millions of yuan. Fengrui Capital and a dual-currency financial investment institution led the investment, with Poke Robot, Infinite Fund SEE Fund, and Tsinghua Alumni Seed Fund following. The funds from the two rounds will mainly be used for the R & D of visual-tactile sensor products, engineering verification, the construction of data toolchains, and market expansion.
Founded in February 2026, Huiguang Innovation mainly provides high-performance visual-tactile sensors, tactile data acquisition devices, and systematic tactile perception solutions for embodied intelligence, dexterous robot manipulation, intelligent manufacturing, human-machine interaction, and scientific research scenarios.
Its entrepreneurial team is composed of post-00s. The founder and CEO, Wang Chenyuheng, graduated from the Department of Electronic Engineering at Tsinghua University with both a bachelor's and a master's degree. He has a multi-disciplinary research perspective and has been responsible for products related to dexterous hands, visual-tactile sensing, and robot data acquisition systems. He also has experience in robot product definition, engineering implementation, and team management. The CTO, Xu Jikai, studied for a master's degree in the Department of Control Science and Engineering at Zhejiang University and graduated from the School of Mechanical Science and Engineering at Huazhong University of Science and Technology for his undergraduate degree. He has long been engaged in research on visual-tactile sensors, robot perception, and manipulation, and has accumulated experience in both sensor hardware R & D and mass production. Currently, R & D technicians account for more than 70% of Huiguang Innovation's team, and the team is still expanding.
Wang Chenyuheng recalled that it was not easy for a student entrepreneurial team to enter the robotic tactile field, which has strong manufacturing characteristics. In the early days of entrepreneurship, without funds, the team paid for equipment and materials out of their own pockets and set up simple workstations in a rented house. They completed the preparation of the first-generation product in half a year. "Without a supply chain, we went to Shenzhen to visit dozens of factories door-to-door to find solutions. Without customer acquisition channels, we sold early products to developer users through a Taobao store and made hundreds of sales within a few months, earning our first pot of gold."
After accumulating production and R & D experience, the team began to acquire B2B customers, gradually obtaining orders from several customers worth hundreds of thousands to millions of yuan. They also completed seed and angel rounds of financing worth tens of millions of yuan within three months and entered the stage of systematic small-batch delivery and rapid expansion.
The competition logic in embodied intelligence is undergoing profound changes. In the past year, the industry spotlight has mainly been on large models, robot forms, and general operation demonstrations. However, when robots move from the laboratory to real physical scenarios to perform delicate tasks such as grasping, pressing, assembling, sorting, and plugging, a common bottleneck begins to emerge: Robots not only need to "see" but also need to "have a sense of touch." Key information such as whether they are in contact with an object, the magnitude of the applied force, surface slippage, and contact surface deformation is beyond the reach of visual perception.
The fundamental problem is that the current training of embodied intelligence models highly relies on visual, linguistic, and motion trajectory data, while very little tactile data is generated during real physical interactions. Without high-quality tactile data, the model has no way to learn the physical laws of "contact," and it is more difficult to generalize the robot's operation ability.
Most traditional tactile solutions, from contact switches and touchscreens to electronic skins, have difficulty balancing spatial resolution, multi-dimensional information decoupling, durability, and batch consistency. Moreover, the output signal dimensions are generally low, making it difficult to effectively connect with AI training systems. Huiguang Innovation has chosen the path of visual-tactile sensors, which collect images of the deformation of the elastomer contact surface through optical signals and reconstruct information such as contact position, force distribution, slippage, and texture through algorithmic models.
Light is an electromagnetic wave with the highest frequency and the largest bandwidth among sensing media. From the perspective of first principles, this means high precision and high resolution. Visual-tactile sensing can convert the physical contact process into image data with micron-level spatial resolution, accurately recording information such as pressure distribution and deformation on the contact surface. This type of high-fidelity data is also the core material required for physical AI training. More importantly, the image data output by visual-tactile sensing is naturally compatible with mainstream AI architectures and can be used for the training of embodied intelligence and physical AI.
However, the visual-tactile approach also has pain points such as thick volume, high computing power requirements, high price, and poor durability, which are the directions that Huiguang Innovation is committed to solving.
In terms of the technical route, Huiguang Innovation adopts the gray-scale reconstruction route, which can reduce the complexity of the internal optical structure and algorithmic link. This brings three core product advantages: thinner structure, lower computing power requirements, and better cost control. Based on this, Huiguang Innovation's self-developed ultra-thin visual-tactile sensor LIGHT TILE has its first version with a thickness controlled at 3 - 4mm and has entered the productization stage. It is being deeply adapted for end-effectors such as robot fingertips and grippers.
The thin product definition also comes from AI-native thinking: The source of intelligence is data, and near-field tactile interaction data comes from contact. Thinner tactile sensors are more conducive to large-scale real-machine data collection. At the same time, visual-tactile data has the largest amount of information and the most redundancy, and the collected data is more likely to be reduced to different types of data for cross-ontology generalization. In addition, in the design of real robot bodies, the thickness of the sensor directly determines whether it can be embedded in compact structures such as dexterous hands. Thinner means a leap from laboratory devices to basic robot components.
In addition to sensor hardware, Huiguang Innovation will also cooperate with partners in the future to provide tactile data services. Tactile data can be divided into two aspects: one is high-fidelity real-machine data, which is collected through the ultra-thin visual-tactile sensor series in cooperation with partners; the other is simulation data, which is used to feed back the construction of the simulation platform through real-machine data.
The commercialization path will also be advanced in two stages: In the early stage, the LIGHT TILE series of hardware will be used to enter the market, and customers and scenarios will be acquired through direct sales. In the later stage, as the deployment scale expands, tactile data and perception algorithms will be productized to form continuous revenue such as dataset licensing and algorithm APIs. Currently, the fingertip and gripper models have signed contracts with leading customers for small-batch delivery.
After the completion of this round of financing, the company will continue to promote the iteration of core sensors, engineering verification, supply chain construction, and pilot cooperation, improve the tactile data collection and processing system, and promote the application of visual-tactile sensors and data infrastructure in the embodied intelligence and consumer electronics industries.
The following is an excerpt from an interview with Wang Chenyuheng, the founder of Huiguang Innovation, by Yingke:
Yingke: The founding team members are all post-00s and relatively young. Why did you choose to start a business at that time?
Wang Chenyuheng: The core driving force is that we all love playing with robots. During our scientific research, we personally felt two gaps: one is the lack of truly useful and implementable robotic tactile hardware on the market, and the other is the extreme shortage of large amounts of high-quality tactile data in the entire industry. At that time, we judged that for embodied intelligence to enter the real physical world, tactile perception must be a key and unavoidable link, and this direction matches our technical accumulation and interests. So we decided to start our own business, hoping to make up for this shortcoming in the industry.
Yingke: For a young team, how will you handle the industrial chain, supply chain, and engineering delivery capabilities?
Wang Chenyuheng: We fully understand this kind of doubt. After all, our average age is less than 30, and we are working on hardware. At the beginning, we really had a lot to explore in production and manufacturing. However, our team has strong execution and learning abilities. At first, to verify market demand, we opened a Taobao store and made hundreds of sales in a few months. In this process, we ran through the entire chain of order receiving, production, delivery, and after-sales service, proving that there is a demand for our products and getting a clear understanding of the basic process. Later, we actively visited factories in Shenzhen, Dongguan, and other places, talked to them one by one, actually inspected the production lines, and communicated repeatedly with engineers and workers in the upstream and downstream. It was this experience of "manual production + door-to-door promotion" that gave us real customer feedback and production experience before the financing.
Yingke: Where are the technical barriers and core advantages of your products mainly reflected?
Wang Chenyuheng: It can be explained from two aspects: hardware and software. In terms of hardware, we have significantly reduced the thickness and volume of the product. The first-generation product has a thickness of 3 - 4mm, which is relatively advanced in the field of visual-tactile sensing. In terms of software, we have a complete self-developed algorithm and data processing process, which can accurately and quickly reconstruct high-dimensional tactile information such as three-dimensional morphology, pressure distribution, and texture from optical images. In summary, our sensors have three prominent features: first, they are very friendly to AI training and deployment, with low computing power requirements and a data format that is more suitable for mainstream deep learning frameworks; second, the products are thin and suitable for tactile data collection; third, they are cost-effective. To achieve these, in addition to choosing the right technical route, we also attach great importance to supply chain integration to control costs, consistency, and durability from the source.
Investor's view:
Li Gang, Vice President of Fengrui Capital: The lack of tactile data has become a common consensus and challenge in the current field of embodied intelligence. Against this background, Huiguang, with AI Native as the core design concept, has overcome multiple problems such as tactile accuracy, durability, and physical volume, achieving a technological breakthrough with reliable hardware and efficient algorithms. It is particularly noteworthy that Huiguang's team, with post-00s as the core force, shows an industry vision and execution ability beyond their age. They have actively cooperated with several dexterous hand, robot body, and world model companies, promoting technology implementation and industrial collaboration with the creativity of the new generation.