Transfer Technology: equipping robotic arms with "eyes" and "brains"

Technology Author: Lina Peng Sep 19, 2023 10:57 AM (GMT+8)

During this year's "two sessions", the principle of accelerating self-reliance and self-improvement in science and technology to promote high-quality development, has been stressed. Meanwhile, according to the 2023 Government Work Report, as spending on R&D increased from 2.1% to over 2.5% of GDP, the contribution of scientific and technological progress to economic growth exceeded 60 percent in 2022.


Against the backdrop of the new era, enterprises in all fields have mushroomed and taken the leads in the global markets bolstered up by top-notch technologies and innovative spirits, showcasing the national strategy driven by innovation and technology. Among them, there are large-scale companies that rose from traditional business and keep self-evolved through these decades, as well as start-ups born to embrace innovation and globalization.

EqualOcean's new interview column themed as "Technological Innovation, Leading the World" will focus on start-ups, characterized as innovative technology and product, global competitiveness and global layout, in the advanced manufacturing, new energy, consumer technology, and digitization fields. Through close-up observation of the R&D processes and the global vision of their core management, this series aims to display the new looks and opportunities of China's technology and innovation companies.

In this article, EqualOcean conducted a dialogue with Fan Yu (Chinese:樊钰), Founder and CEO of Transfer Technology (Chinese: 迁移科技), a 3D industrial cameras and 3D visual robotic arm guidance systems supplier.

EqualOcean: Please introduce Transfer Technology to us first.

Fan Yu: Transfer Technology was founded in 2017. Our company focuses on providing 3D industrial cameras and 3D visual robotic arm guidance systems, providing industrial robots with "eyes" and "brains" to enable them to perceive, make decisions, and manipulate objects. At present, Transfer Technology has accumulated a group of top customers in industries such as automobiles, chemicals, home appliances, and heavy industry. This year, we have also achieved commercialization breakthroughs in the automotive industry in North America.

EqualOcean: Why did you choose North America as your first stop going overseas? What are the global expansion plans in the future?

Fan Yu: The 3D visual robotic arm guidance system is positively correlated with the global sales of robotic arms. To expand overseas markets more effectively, we prioritize several regions with the highest sales of robotic arms, including North America, Europe, and Japan and South Korea. The first stop going overseas in North America is both an opportunity factor and within the plan. At present, we have delivered a batch of products locally, which has boosted the confidence of the entire team to further expand overseas business.

EqualOcean: What is the background of the founding team of Transfer Technology?

Fan Yu: Members of the founding team graduated from top universities, such as Tsinghua and Zhejiang University with a major in science and engineering. Prior to the establishment of Transfer Technology, they had been engaged in research and development related to vision-based robot guidance.

AI technology was experiencing a peak in 2017. Both Google AlphaGo defeats human Go masters, Yolo Network's stunning recognition speed and accuracy have attracted attention from all walks of life. Some AI algorithms have surpassed humans in recognition accuracy on the ImageNet dataset. At the same time, a group of AI startups have also emerged in China. We initially anchored the field of entrepreneurship-to join the AI entrepreneurship trend as soon as possible.

At that time, there were two main directions for us: security and industry (AI visual defect detection and robotic arm guidance). After taking serious consideration and research, we ultimately chose the direction of robotic arm guidance. The main reasons are: first, we have been deeply involved in visual based robotic arm guidance before; second, we have determined that the threshold for 3D visual robotic arm guidance is higher, the technology chain is longer, and it is more suitable for building long-term competitive barriers.

There are two main technical routes for defect detection: one is based on traditional visual algorithms; One is deep learning, which involves collecting samples and training models. The main barriers lie in algorithms and negative sample banks. The guidance of 3D visual robotic arms first requires robust 3D camera hardware, and on this basis, 3D visual recognition algorithms are developed. At the same time, it is necessary to connect with various brands of robotic arms on the market to do motion planning and obstacle avoidance for the robotic arm. Currently, the cases of Transfer Technology implementation have covered mainstream global robotic arm brands.

After installing the "eyes" and "brain" on the robotic arm, it can complete tasks that were originally impossible for the robotic arm to complete and require higher flexibility, such as disorderly loading and unloading of components, disorderly dismantling and stacking, and high-precision assembly. In the implementation stage of product applications, we prioritize the automation upgrade for scenarios with heavy physical strength, harsh environments, and even harmful to the human body, which is also the source of our sense of achievement and value in our work. For example, the loading, unloading, and assembly of large components such as automotive engines require high physical strength. During the unpacking process of the chemical production line, harmful dust may be released, and workers need to wear gas masks for operation.

Screen Shot 2023-09-19 at 10.52.39 AM.png.png

Transfer Technology New Product Laser Camera Epic Eye Laser L

EqualOcean: What is the core sector in the entire 3D visual robotic arm guidance system?

Fan Yu: First, it should be clarified that in the industrial direction, every aspect of hardware, algorithms, and software must have no weaknesses, and any weaknesses in any aspect may cause the system to ultimately fail to be implemented on the customer's site.

Second, we have established research and development advantages in all aspects. In terms of hardware, we have independently developed the world's only 3D laser mechanical galvanometer camera capable of stable imaging above 120,000 lux. In terms of algorithms, our self-developed 6D pose estimation algorithm has been ranked first on the global algorithm rankings since April 2022, which is also one of the important reasons why our system can be quickly implemented on customer sites.

Software is an important factor in scaling. In terms of software, we focus on ease of use while also considering universality. First, it is necessary to ensure that integrators and even on-site workers can complete the system construction with simple operations at the system prompt; Second, with the help of the software's built-in universal algorithm engine, users can achieve more complex logic and functions through simple "drag and drop".

EqualOcean: What is the positioning of Transfer Technology in the 3D visual industry?

Fan Yu: We have a clear positioning. We are a provider of 3D visual software and hardware products. When landing or exploring new scenarios, Transfer Technology is responsible for completing the first wave of "battle", bearing the cost of innovation, and providing standardized 3D visual robotic arm guidance systems for integrators and partners to benefit from the innovative results. We will even grant the R&D experience and some patents of the first set of workstations to our integrator partners for free.

EqualOcean: In the new round of robots and AGI integration, what new opportunities does Transfer Technology see?

Fan Yu: We are very excited about this major trend. In the long run, the form of universal robots in the future will be humanoid. Residential buildings, offices, factory production lines, cars, and so on are all designed for people. If you want to design a robot that directly adapts to the environment and equipment without modifying them, so that it can perform all the work done by humans in the past, the human form must be the best.

However, the era of large-scale application of humanoid robots is still relatively far away. On the one hand, it is a consideration of cost and price. On the other hand, there are still a lot of technical challenges that have not been solved, such as obstacle avoidance and flexible movement of the feet, true perception decision-making and control ability.

At present, there is an idea in the industry that to achieve universal intelligence of robots, a fuzzy main task needs to be decomposed into specific subtasks using ChatGPT, a large language model. The subtasks are implemented through some special algorithms or proprietary patterns. Transfer Technology is currently engaged in researching proprietary models and algorithms to assist robots in performing sub tasks, such as loading and unloading, palletizing, welding, and gluing in factories. The future universal humanoid robots will also be built on the combination of subtasks.

EqualOcean: How does Transfer Technology see the opportunities and challenges in the commercial market?

Fan Yu: This issue is mainly considered from the perspective of ROI. At present, the main application scenarios for robotic arms and collaborative arms are still in industries, warehousing, logistics, etc., and their true application in consumer level scenarios is relatively rare, and it will take some time to verify. Therefore, in the next two to three years, industry will still be the main application scenario for 3D vision.

EqualOcean: Is demand mining or market education a challenge for the commercialization of 3D vision?

Fan Yu: I believe that the key customers at the current stage should not require significant persuasion. In the book 'Crossing the Chasm ', Geoffrey Moore divided users into five categories: innovators, product pioneers, early adopters, late adopters, and laggards. In the early stages of development in the 3D vision industry, we should focus on two groups that are more sensitive to new technologies: innovators and product pioneers.

In addition, in the field of 3D vision, every startup company has strong technical capabilities, but companies with strong technical capabilities are also prone to falling into the misconception of "taking over everything". The hardest part is' focusing '- deciding what to do and what not to do. Focusing on one direction is the key to precise product positioning, which provides the opportunity to create a product that exceeds customer expectations.

EqualOcean: We have also seen some cases of downstream integrators experiencing failures in reverse exploration of software and hardware products. Why does this happen?

Fan Yu: Integrators are profit oriented and focus more on how to reduce costs. But the development of 3D vision related products involves a large amount of research and development investment and upfront costs and requires long-term determination and patience to see the ideal results. At present, it seems that the window period has passed, and the current customers are not very patient in letting suppliers try and make mistakes. We may only need two or three days to deliver a 3D visual robotic arm guidance system, and the customer will not accept it if it takes one or two months to operate stably.

Yu Lu (余璐), the angel round investor of Transfer Technology and the managing partner of 01VC (零一创投), gave such a comment:"The current 3D machine vision industry is in a rapid development stage and is gradually being applied in industries such as industrial manufacturing and logistics warehousing. In the future, many hardware products will be equipped with 3D visual technology to independently collect, analyze, and utilize data to make decisions, and interact with the environment. The 3D vision system of Transfer Technology has high usability and standardization and has served multiple top customers in industries such as home appliances, new energy, automobiles, petroleum, and petrochemical."

About Transfer Technology

Transfer Technology is a leading supplier of 3D industrial cameras and 3D visual robotic arm guidance systems in the industry. Based on years of technical expertise in 3D camera hardware, algorithms, and software, Transfer Technology has created a stable, easy-to-use, and high return 3D visual system, aiming to provide technological empowerment for global industrial manufacturing and warehousing logistics automation. At present, Transfer Technology has served multiple industries such as new energy, automobiles, chemicals, home appliances, and metal manufacturing, with customers including Mercedes Benz, Tesla, CNOOC (中海油), Sinopec (中石化), Midea (美的), Haier (海尔), and other Fortune 500 companies.