Konnect-linK Co., Ltd. Fully Launches R&D of "AI that Operates On-site" (Physical AI)
Konnect-linK Co., Ltd., through its subsidiary Konnect Frontier Lab, has officially commenced full-scale research and development of "Physical AI"—AI that operates directly within IoT devices and industrial machinery. This initiative aims to address labor shortages in core industries such as manufacturing, logistics, and healthcare by enabling AI to autonomously perceive on-site conditions, make instantaneous decisions, and take direct actions in the real world.
📋 Article Processing Timeline
- 📰 Published: April 30, 2026 at 20:04
- 🔍 Collected: April 30, 2026 at 11:31
- 🤖 AI Analyzed: April 30, 2026 at 12:06 (34 min after Collected)
Konnect-linK Co., Ltd. (Headquarters: Chiyoda-ku, Tokyo; Representative Director: Kento Komoda) has officially commenced full-scale research and development of "Physical AI"—AI that operates directly within IoT devices and industrial machinery—through its subsidiary, Konnect Frontier Lab.
Traditionally, Konnect-linK has promoted AI implementation in on-premise environments and edge devices, focusing on the edge AI domain where AI perceives and makes decisions on the device side without relying on the cloud. This time, we have expanded our research and development to Physical AI and have officially launched it.
This press release signifies that our group's internal research and development, which has been ongoing for over a year, regarding the affinity between AI utilization, foundational technologies, and physical domains such as IoT sensor devices, has completed its verification and knowledge integration, and is now transitioning to the social implementation phase.
At our subsidiary, Konnect Frontier Lab, we aim to evolve AI from "AI far away" to "AI residing within objects" and then to "AI acting in physical space." This is achieved by researching and developing processes for receiving, interpreting, judging, and outputting all kinds of perceived/unperceived information, configuring appropriate original architectures, and integrating these into physical media (robots, industrial machinery, mobility, drones, sensor devices, infrastructure equipment, etc.).
This will accelerate the social implementation of "AI that truly operates on-site," where AI autonomously perceives on-site conditions, makes instantaneous decisions, and takes direct actions in the real world through robots and machinery.
Furthermore, we plan to conduct joint verification experiments with multiple companies by the next fiscal year (end of April 2027), advancing the creation of specific use cases and validation for practical operation.
Konnect-linK: From "AI far away" to "AI residing within objects," and then to "AI acting in physical space."
Background: Why "Physical AI" now?
Until now, AI primarily operated in data centers, existing to "think and answer." It read text, analyzed images, and generated optimal responses—but its scope of activity largely remained within screens.
The change happening now is a paradigm shift where AI will perceive and directly act upon the real world. This is Physical AI: AI understanding the immediate situation through cameras, reading the environment from sensors, and reflecting its judgments as movements of robots and machinery in physical space.
This change holds immense significance in areas where "what is happening on-site" has intrinsic value—Japan's core industries such as manufacturing, logistics, construction, healthcare, and nursing care. Japan is projected to face a labor shortage of approximately 7 million people by 2030 (*1), making it increasingly difficult to maintain operations solely by securing human resources. The introduction of Physical AI, where AI autonomously understands on-site conditions and takes on part of the work through machinery and robots, is becoming a realistic solution to such structural challenges.
Moreover, for AI to act upon the real world, the AI itself must operate in close proximity to the site. Cloud-based processing makes it difficult to react instantly to events occurring on-site, and there are limitations to real-time control of machinery. Physical AI, in its essence, is premised on completing perception, judgment, and control on the on-site device.
"From AI confined to screens to AI that moves reality."
Konnect Frontier Lab will engage in research on foundational technologies to implement this structure—where "AI perceives, judges, and acts upon the real world"—within all on-site machinery and devices.
Research Directions
Konnect Frontier Lab will promote research and development to complete the entire process of AI perceiving on-site data, making judgments on the spot, and taking action in real space, all on the on-site device. Specifically, we aim to realize a world where:
- Cameras constantly recognize on-site conditions and instantly execute physical actions such as line control or warnings the moment defects or abnormalities are detected.
- Industrial machinery constantly monitors its own status (vibration, temperature, etc.) and autonomously controls output and sends maintenance notifications when signs of malfunction are detected.
- Logistics robots understand the warehouse environment, autonomously determine the optimal route based on the situation, and execute transportation.
- Autonomous machines perceive their surroundings from multiple sensors, avoiding people and obstacles while safely continuing operations.
- On-site equipment understands worker's speech and instructions, providing appropriate responses and controlling related equipment.
Traditionally, Konnect-linK has promoted AI implementation in on-premise environments and edge devices, focusing on the edge AI domain where AI perceives and makes decisions on the device side without relying on the cloud. This time, we have expanded our research and development to Physical AI and have officially launched it.
This press release signifies that our group's internal research and development, which has been ongoing for over a year, regarding the affinity between AI utilization, foundational technologies, and physical domains such as IoT sensor devices, has completed its verification and knowledge integration, and is now transitioning to the social implementation phase.
At our subsidiary, Konnect Frontier Lab, we aim to evolve AI from "AI far away" to "AI residing within objects" and then to "AI acting in physical space." This is achieved by researching and developing processes for receiving, interpreting, judging, and outputting all kinds of perceived/unperceived information, configuring appropriate original architectures, and integrating these into physical media (robots, industrial machinery, mobility, drones, sensor devices, infrastructure equipment, etc.).
This will accelerate the social implementation of "AI that truly operates on-site," where AI autonomously perceives on-site conditions, makes instantaneous decisions, and takes direct actions in the real world through robots and machinery.
Furthermore, we plan to conduct joint verification experiments with multiple companies by the next fiscal year (end of April 2027), advancing the creation of specific use cases and validation for practical operation.
Konnect-linK: From "AI far away" to "AI residing within objects," and then to "AI acting in physical space."
Background: Why "Physical AI" now?
Until now, AI primarily operated in data centers, existing to "think and answer." It read text, analyzed images, and generated optimal responses—but its scope of activity largely remained within screens.
The change happening now is a paradigm shift where AI will perceive and directly act upon the real world. This is Physical AI: AI understanding the immediate situation through cameras, reading the environment from sensors, and reflecting its judgments as movements of robots and machinery in physical space.
This change holds immense significance in areas where "what is happening on-site" has intrinsic value—Japan's core industries such as manufacturing, logistics, construction, healthcare, and nursing care. Japan is projected to face a labor shortage of approximately 7 million people by 2030 (*1), making it increasingly difficult to maintain operations solely by securing human resources. The introduction of Physical AI, where AI autonomously understands on-site conditions and takes on part of the work through machinery and robots, is becoming a realistic solution to such structural challenges.
Moreover, for AI to act upon the real world, the AI itself must operate in close proximity to the site. Cloud-based processing makes it difficult to react instantly to events occurring on-site, and there are limitations to real-time control of machinery. Physical AI, in its essence, is premised on completing perception, judgment, and control on the on-site device.
"From AI confined to screens to AI that moves reality."
Konnect Frontier Lab will engage in research on foundational technologies to implement this structure—where "AI perceives, judges, and acts upon the real world"—within all on-site machinery and devices.
Research Directions
Konnect Frontier Lab will promote research and development to complete the entire process of AI perceiving on-site data, making judgments on the spot, and taking action in real space, all on the on-site device. Specifically, we aim to realize a world where:
- Cameras constantly recognize on-site conditions and instantly execute physical actions such as line control or warnings the moment defects or abnormalities are detected.
- Industrial machinery constantly monitors its own status (vibration, temperature, etc.) and autonomously controls output and sends maintenance notifications when signs of malfunction are detected.
- Logistics robots understand the warehouse environment, autonomously determine the optimal route based on the situation, and execute transportation.
- Autonomous machines perceive their surroundings from multiple sensors, avoiding people and obstacles while safely continuing operations.
- On-site equipment understands worker's speech and instructions, providing appropriate responses and controlling related equipment.