【Zrek Corporation】Realizing Autonomous Factories Where Machines Self-Think and Improve! Robotics DX Creating the "Brain" of Autonomous Robots Combining Image Recognition and Generative AI

Machines Self-Thinking and Improving. The Roadmap to Building the Most Efficient Autonomous Factory on Earth

Thank you very much. My name is Yuki Imamura, Representative Director of Zrek Corporation.

We are a company that creates the \"brain\" of robots. Our company name, \"Zrek,\" is a combination of the letter \"Z,\" which represents the ultimate destination, and \"TREK,\" which means a journey. We have our headquarters in Shibuya and our research and development base at \"KBIC,\" a facility in Kawasaki City. I have been deeply interested in programming and space since my childhood, and majored in mechanical and aerospace engineering at university. After experiencing student entrepreneurship and other activities, I became absorbed in research on simulation and AI triggered by the COVID-19 pandemic, and founded this company to apply that technology to solve real-world issues.

Currently, a serious labor shortage and population decline are progressing in manufacturing sites worldwide, including Japan. Many manufacturing companies are experiencing a loss of business opportunities (business issues) where they cannot expand their business simply because there are no people.

Of course, Japan has long had highly precise and excellent \"industrial robots\" that play an active role in the automotive industry. However, they are basically designed to repeat pre-determined simple actions without a millimeter of error. Robots that can safely cooperate with humans or autonomously adapt to irregular environmental changes have not yet been put into practical use.

Even more fatal is that existing industrial robots require massive programming (system integration) for each individual factory or task. Since the products and layouts handled in each factory are all different, a SIer must enter and build individually for each company. The time and communication costs associated with this are enormous, making it virtually impossible to roll out to multiple locations. We aim to fundamentally eliminate this barrier to introducing robotics using software and AI, and ultimately build \"autonomous factories where machines self-think and improve their own processes, which is the most efficient on Earth.\"

Equipping Cheap Hardware with AI Brain. Accomplishments in Autonomous Automation in Kawasaki City's Demonstration and Chemical Fields

To achieve the ultimate goal of establishing autonomous factories, we are currently taking several clear development phases.

As our first approach, we are working on the \"autonomous automation of a part of operations.\" Instead of developing expensive proprietary hardware, we procure cheap cooperative robot arms from existing robot manufacturers and equip them with our proprietary \"AI brain model\" that combines image recognition and generative AI to deliver to factories.

Our project to autonomously automate machine tools in precision sheet metal processing has been selected as a demonstration model in Kawasaki City this fiscal year, receiving high evaluation. In manufacturing sites, there is a realistic, gritty issue where metal powder called \"chips (kiriko)\" generated during metal processing adheres to sensors and machines, causing defective products. We take an approach to solve this physical trouble using software, enabling the robot arm to autonomously recognize the situation with camera images and proceed with work while handling it.

Furthermore, validation of our autonomous robots is also progressing at sites such as chemical product prototype development. In the chemical field, a large number of simple repetitive tasks dependent on human researchers' manual work occur, such as weighing reagents accurately in milligrams, stirring liquids, and defoaming to remove bubbles. We have also succeeded in demonstrating how these working processes are replaced by robot arms, letting them autonomously perform tasks through \"imitation learning\" of human example actions from camera videos.

In this way, we have established a technology package that leads areas that previously required huge programming costs to automation in a short period and at low cost by letting AI learn.

Becoming Infrastructure While Being a Player. Group Army Strategy Utilizing the Wisdom of Japan as a Family Business Superpower

Our business model and roadmap do not end with simply \"selling robots\" or \"contract development.\"

We equip cheap robots with AI, introduce them to factories, and run a cycle of constantly updating software by collecting real-world data in the field. While packaging and horizontally expanding this partial automation technology, we plan to proceed with M&A (mergers and acquisitions) and alliances of manufacturing companies in the mid-to-long term, directly establishing and operating \"autonomous factories\" within our own group.

In Japan, there are mountains of small and medium-sized family businesses that have highly advanced manufacturing wisdom and craftsmanship but are in danger of survival due to a lack of successors and labor shortages. We group these companies and introduce our self-developed autonomous automation robots internally to build an overwhelmingly efficient, fabless contract manufacturing group. In other words, we generate profits as a manufacturing player ourselves while polishing the usefulness of robots in our own factories, and ultimately sell \"the autonomous factory system itself\" to other manufacturing industries.

We ask for help from alliance partners and manufacturing companies for our challenge to maximize Japan's manufacturing geographical superiority and historical wisdom with cutting-edge software like image recognition and generative AI. Thank you very much.

Q&A and Feedback

Commentator (Mr. Mori): Thank you, Mr. Imamura, for the presentation in a field with great dreams and expectations. The approach of solving robotics challenges by specializing in software is very rational.

My question is, what kind of hardware do you plan to equip with your AI brain in the future? Do you plan to specialize in rigid industrial robots like Fanuc, or do you target versatile robots like Tesla's humanoid proposed by Elon Musk or cheap human-like robots from China? Please tell us about your future positioning and vision.

Mr. Imamura: Thank you for the question. To put it simply, we ultimately want to create \"robots with high versatility specialized in specific tasks (task-specific general-purpose AI).\"

A complete general-purpose humanoid robot that perfectly does housework is still in its infancy in terms of both technology and cost. However, if we narrow down the business domain to \"repetitive tasks and setup changes in factories,\" we can provide practical versatility by combining current generative AI and image recognition technologies.

The reason why this is necessary is that conventional industrial robots required engineers to spend dozens of hours rewriting programming every time they changed a single task, which was the biggest bottleneck (wall of impossibility) for multi-location rollout. By equipping them with our AI brain, we aim to popularize arm robots that autonomously judge by looking at the situation with cameras and switch tasks without programming—providing \"versatility within operations.\"

Mr. Mori: I see. Instead of a complete jack-of-all-trades, you secure versatility within the boundary of factory operations using AI.

Another point: as a business development form, will it be an \"Apple type that vertically integrates both hardware and software like Tesla,\" or will it be an \"Android type that specializes in software and app platforms and procures hardware from various manufacturers?\" Which image is closer?

Mr. Imamura: In that regard, our position is an image of taking the advantages of both the US style of \"procuring foreign hardware with software as the axis\" and the Chinese style of \"procuring cheap hardware and software.\"

The US is strong in software but weak in hardware, and China enters from cheap hardware and places compatible software later. We currently procure cheap overseas robots and integrate them as a fabless company, but Japan has an overwhelming accumulation of \"manufacturing factories and craftsmanship wisdom\" developed over a long history. To leverage this asset (strength), we do not take the Apple style of making everything by ourselves. Instead, we group small and medium-sized factories suffering from successor shortages through M&A and alliances, and internally introduce common AI software to build an autonomous manufacturing platform. We aim for a hybrid rollout, functioning as a player while refining the package and ultimately selling the system like Android.

Mr. Mori: Leveraging Japan's context as a family business superpower, grouping them, and pouring in software. Starting from a SIer-like movement and ultimately selling and platforming common packages has high reality and potential. I look forward to the demonstration in Kawasaki City and expansion to other industries.