On 2 November, JTEKT announced that it had reached a 97.1% success rate for picking loose parts on an automotive parts line at its Tadomisaki plant, Mynavi News TECH+ reports.
The rate was measured in difficult conditions, chosen to bring out differences in performance. With the robot's conventional control, it was 85.0%. According to the company, the average cycle time also fell by 16%.
The AI works in two stages. First, a fast model spots candidate parts in the 3D camera image. A second, more precise model then calculates the position and orientation of the part to be picked, but it runs only on 1 or 2 candidates.
The trial ran from 19 March to 31 August 2026, as part of an AWS Japan programme, with Upstage AI and Karakuri.
Next, JTEKT says it wants to add this AI to workstations that are already automated. The company is aiming for a higher success rate and zero human intervention.
Source
- マイナビニュース TECH+ (ものづくり) · news.mynavi.jp