Mitsubishi, Sony form JV to build AI vision sensors for automated factories

Mitsubishi Electric and Sony Semiconductor Solutions will establish a joint venture to develop AI-powered vision sensors for factories seeking to automate production and reduce dependence on manual labor.

The joint venture, Advanced Vision Solutions Co., is scheduled to begin operations in October, subject to regulatory approvals, the companies said on Wednesday, 22 July.

Mitsubishi Electric will own 60% of the Yokohama-based venture, while Sony Semiconductor Solutions will hold 40%. Tamiki Kobayashi will serve as its representative.

The venture will combine Sony’s image sensors and edge AI technology with Mitsubishi Electric’s factory automation systems.

The planned sensors will analyze visual data directly on the device and pass information about product quality, machinery performance and maintenance requirements to factory control systems.

By processing images at the sensor rather than sending all the data to a separate system, the technology is intended to support faster decisions and reduce the amount of information that must be transferred or stored.

The companies said the sensors could help machinery detect changing production conditions and adjust operations with limited human intervention.

Visual information could also be combined with other factory data to detect equipment abnormalities and early warning signs that might be missed when relying on a single source.

Mitsubishi Electric plans to integrate the resulting vision data with Serendie, its digital platform, as it develops factory automation systems capable of greater autonomous control.

“The need for labor-saving, unmanned systems and autonomous optimization at production sites is greater than ever,” said Takayuki Tsuzuki, head of Mitsubishi Electric’s Factory Automation Systems group.

Sony Semiconductor Solutions President and Chief Executive Officer Shinji Sashida said image sensors could serve as the “eyes” of AI systems, allowing them to understand factory conditions and connect that information to operational decisions and controls.

The venture will target manufacturers seeking to automate production, improve quality inspection and strengthen predictive maintenance.

The companies said they also want to make the systems usable without specialist knowledge of optics or AI, which could widen adoption beyond large manufacturers with dedicated technical teams.

The partnership reflects growing interest in physical AI, in which software uses data from sensors and machines to make decisions in real-world industrial settings.

Unlike conventional factory cameras used mainly for inspection, the planned systems are intended to connect visual recognition directly with machinery controls, allowing equipment to respond to defects, changing conditions or signs of failure.

The companies did not disclose their planned investment, revenue targets or when the sensors would become commercially available.