ROBOT+AI Summit Series | Mech-Mind: The People Giving Machines "Eyes"

On automotive production lines, thousands of door inner panels need to be processed every day. Workers remove panels from material racks, manually carry them to laser welding cells, load them into welding fixtures, perform laser welding, then manually transport them to the next station for further processing.

A vehicle body's quality and cost depend heavily on its stamped components. These components are numerous and categorized by position and function into outer panels, structural parts, reinforcements, and general stampings. Among them, the door inner panel is one of the most complex body outer panels — large in size, with complex surfaces, numerous assembly holes, and high susceptibility to deformation.

Manual handling is time-consuming and error-prone. But simply replacing workers with robotic arms isn't enough. A vision system must photograph the precisely machined primary and secondary locating pin holes on the inner panel, determine their positions, and guide pin insertion for gripping. However, the pin holes are small and the workpiece surface is reflective, making precise recognition impossible for conventional vision systems.

Automotive manufacturing encompasses every known industrial category — the grand assembly of all industrial sectors — and is called the crown of civilian manufacturing. Now, this crown is being refined by ever more skilled craftsmen.

Mech-Mind's Mech-Eye PRO S features exceptional anti-reflective performance, enabling high-quality 3D imaging of reflective pin holes after precision machining. The 3D vision system guides robots to grip accurately, avoiding deformation and damage from squeezing or scraping, thereby improving door inner panel production quality.

This year, the emergent capabilities of large language models have given humanity hope of building robots that surpass ourselves. But between LLMs and the physical world lies a chasm that remains difficult to cross — from 2D images and text to the 3D world, how does an intelligent brain graft onto a mechanical body?

Researchers have done extensive work. From Google DeepMind's Robotics Transformer 2 (RT-2) to Tesla's humanoid robot Optimus, the question of how to instill will in steel, how to transcend flesh to create miracles — "seeing" the world is a critical step.

Tianlan Shao and his team are trying to give robots "eyes." To handle different scenarios, lighting conditions, and edge cases, Shao and his engineers rebuilt 3D algorithms and introduced artificial intelligence.

Mech-Mind is one of the world's largest AI + robotics innovation companies, with deep expertise in optical/mechanical/electrical core components, 3D perception, vision and robotics algorithms, robotics software, and industry application solutions. It has deployed over 10,000 cameras globally.

The industrial robotics field has seen many strategic debates, which broadly resolve into three tiers: general foundational technology, foundational products, and niche industry applications. Startups choosing to build niche products or applications can stand out if their direction is sufficiently broad and their technology sufficiently leading. China is also the fastest-growing market for machine vision applications.

Source Code Capital was the lead investor in Mech-Mind's Series B+ and Series C rounds. We recently interviewed Mech-Mind founder and CEO Tianlan Shao about the current state and future prospects of China's robotics industry. Shao believes intelligent robots have become an unmistakable trend: "A few years ago, using intelligent robots made you a rare pioneer. Today, it's a relatively mature product. In a few more years, not using intelligent robots will be like not using computers or smartphones today — you'll be left behind."

The era of intelligent robots may truly arrive soon, and a new generation of Chinese companies is gaining first-mover advantage.

1. Mech-Mind showcased its most advanced technology at this year's CIIF. Could you introduce what was presented?

At this year's CIIF, Mech-Mind demonstrated 3D rapid scanning of transparent objects. Thanks to recent research in optics and AI imaging algorithms, we now achieve excellent 3D scanning results for transparent and semi-transparent objects. We also showcased technological breakthroughs in deep-bin complex object picking, large workpiece online measurement, and many other areas. On the industry application side, we brought typical scenarios from lithium battery, photovoltaic, automotive, auto parts, home appliance, and logistics industries — adapting actual customer-site applications for the exhibition floor.

2. How is overseas market expansion progressing?

Mech-Mind Robotics is making very rapid progress in global markets. This year, overseas business is expected to account for roughly half of total company revenue. With AI development, robot intelligence is a massive trend that is already evident in China and now radiating to many other regions — Europe, America, Japan, Korea, Southeast Asia, the Middle East, and beyond. Chinese intelligent robotics companies are beginning to go global, bringing products validated in the Chinese market to the world. In 2022, Mech-Mind's shipment volume ranked first globally, exceeding the combined total of ranks 2-8 in China. Nearly half of business came from overseas, exporting to Germany, Japan, and the United States. 2023 growth is expected to exceed 100%.

3. The trend of robots being applied across industries is now very clear. Why is this happening now? What preparations have been completed in research, technology, and customer readiness?

AI, 3D vision, and other intelligent technologies have advanced extremely rapidly in recent years, which is reflected in end-user deployment volumes. In 2020, when Source Code first invested in us, total 3D vision and AI shipment volume was only a fraction of what it is today — roughly one-tenth or less. We've achieved more than tenfold growth in just a few years, thanks to core technology advances, comprehensive cost reductions, and lower barriers to use.

A few years ago, intelligent robots were still research projects at top institutions like Stanford and MIT. Today, for customers, they are standardized, mature industrial products with low barriers to adoption.

As technology advances, barriers lower, and costs decline, we believe intelligent robots will rapidly penetrate more industries — from large-scale to small and medium manufacturing, into logistics, and eventually into commercial and even household scenarios. In these scenarios, customers are smaller in scale, needs are more fragmented, intelligence requirements are higher, and sensitivity to ease of use and cost increases. We see large models, new optical imaging technologies, and vision and robot planning technologies all rapidly advancing this process.

4. In many industrial scenarios, domestic substitution is a major trend. In which areas has domestic technology now achieved leading advantages?

Traditionally, robotics and industrial automation leaned more toward mechanical and electrical control industries, where Chinese companies overall remained in a catching-up position. But with rapid advances in intelligent technology in recent years, many Chinese companies now hold leading or even distinctly superior positions in integrated optical-mechanical-electrical systems, particularly AI algorithm-related products. For example, Chinese companies now lead globally in niche areas like indoor and outdoor drone delivery and robotic guidance. I believe this stems from Chinese market's high acceptance of new products, combined with excellent talent across niche domains and strong industrial chain foundations. Thus Chinese robotics companies have excellent opportunities to globally pioneer bringing such integrated products to high maturity.

5. Looking long-term, what structural advantages does China's robotics industry have in the current wave of intelligence?

For intelligent robotics companies like ours, the technology chain and business chain are extremely long. Technologically, we span optics, electronics, imaging, AI, robot planning, software development, and much more. Business-wise, beyond core R&D, we need production, delivery, sales, overall solution design, supply chain, and more. In China, we can obtain ample support from talent and upstream-downstream ecosystem coordination. Therefore, I believe Chinese companies, particularly intelligent robotics companies, have tremendous potential in this wave of intelligence, with many opportunities to lead globally.

6. Looking ahead several years, what direction will the industry develop toward?

In coming years, I believe the robotics industry will see massive proliferation of intelligent products and applications.

Robot intelligence was niche laboratory research just a few years ago. Today it has entered many production floors. In the future, it will become standard equipment.

Consider computers: in 1995, a company using computers for office work was quite innovative, quite leading-edge. By 2005, it was utterly ordinary. Today, not using computers for office work would be extremely rare.

We believe robot intelligence technology follows the same trajectory. A few years ago, using intelligent robots made you a rare pioneer. Today, it's a relatively mature product. In a few more years, not using intelligent robots will be like not using computers or smartphones today — you'll be in a very rare, rather backward industrial position. So we believe intelligent robots will undergo rapid proliferation in coming years, quickly becoming the industry's standard solution.