OMRON Corporation (Head Office: Shimogyo-ku, Kyoto. President and CEO: Yoshihito Yamada) announced today that it has developed an AI engine for control equipment jointly with AISing Ltd. (Head Office: Minato-ku, Tokyo. CEO: Junichi Idesawa), which was established to develop the world’s fastest_*1 embedded AI. A leading AI venture business, AISing possesses an AI algorithm "Deep Binary Tree ("DBT")," which achieves high prediction accuracy and high-speed processing with minimal input. Through this alliance, OMRON aims to realize a production line that prevents defective products on the production floor at an "altitude of 1-10m," where everything from collection of sensing data to their feedback to controls has to be performed instantly. The two companies will utilize the AI engine created based on the DBT algorithm to develop a variety of real-time control equipment, such as "next generation AI controllers."

With the aim of achieving "zero line events," i.e., an autonomous production line that keeps learning changes in the 4Ms_*2 to maximize overall equipment effectiveness_*3 (OEE) and completely eliminate defective products. OMRON is tackling the development of innovative control equipment with embedded AI and IoT technology. To realize this goal, input devices for sensing changes in the 4Ms, logics for autonomous control, and output devices and robots that drive the production line in response to control results are essential. In addition, these devices need to be equipped with AI and IoT technology to enhance their "intelligence." Boasting the industry’s largest variety of "ILOR + S (Input, Logic, Output, Robot + Safety)" devices, OMRON will build the newly developed AI engine into a logic (controller) to bring about further evolution. Going forward, OMRON plans to add to a range of its devices, including input/output devices and then intelligent robots in order to reduce the number of line events to zero, thereby offering a solution to the shortage of skilled workers on the production floor.

Through open innovation with customers and business partners, OMRON hopes to realize "zero line events" by 2020, as we work to execute "innovative-Automation," our strategic concept for innovation in manufacturing.

• *1 Operates with the learning time of 10~120µ seconds (µ = 10_-6) and the inference time of 1~5µ seconds in a simple harmonic pendulum trajectory prediction experiment using Raspberry Pi 3
• *2 Changes in "Man," "Machine," "Material," and "Method (manufacturing process)"
• *3 An indicator of efficiency in the use of equipment, which was developed and proposed by the Japan Institute of Plant Maintenance. Calculated by Availability x Performance x Quality