In fields like smart mobility and humanoids, the era of "Physical AI"—systems that act and make decisions autonomously based on sensor data—has begun.
Anticipating the peak of the Physical AI era, we have developed "Autonomous Distributed AI Cells" ahead of the competition.

The "Autonomous Distributed AI Cell" logicizes the very learning process of a newborn child through senses like vision and touch. We have successfully brought this "Child AI" concept to life and integrated it onto a silicon chip.

The "Autonomous Distributed AI Cell" features automatic integration of learnings from other cells, self-correction based on missing or conflicting information, and re-education capabilities for collective learning across multiple cells.
Here lies the Ultimate Physical AI: multiple "Autonomous Distributed AI Cells" learning independently, fostering mutual growth, and autonomously executing coordinated group behaviors.

The "Autonomous Distributed AI Cell" is a world-first technology enabling fully unmanned factory operations and automated facility monitoring/management.
Requiring no individual pre-configuration for sensors or environments upon factory shipment, this composite chip behaves like an artificial neural cell—autonomously sensing its surroundings the moment it is installed and continuously learning various operational elements through active deployment.
Blockchain acts as the synaptic pathways of a neural network, guaranteeing data transmission integrity and enabling distributed processing.
In addition to artificial neural cell capabilities, it is built on highly distinctive, advanced logic that incorporates autonomous audiovisual sensory cells and autonomous nervous system functions akin to the cerebellum.

Equipment installed in a factory autonomously learns local operating conditions and behaviors through visual, tactile, and other sensory inputs, establishing its own "individual identity."
Even among identical machines, slight variations occur—such as subtle discrepancies in motor RPM—so each unit collects operational data to optimize itself for its specific site.
An edge server compares and analyzes data collected from individual units to execute "identity calibration," automatically adjusting voltage and other parameters so all machines deliver uniform output.
By sharing and adjusting (re-educating) the individual variations and experiences of each machine across the entire system, this approach prevents quality inconsistencies and builds a next-generation industrial infrastructure that continues to operate with high precision.

Based on collective action instructions from a master drone (Autonomous Distributed AI Cell [Master]), multiple slave drones (Autonomous Distributed AI Cell [Slave]) execute autonomously controlled flights while coordinating with one another.
Information such as obstacle avoidance and optimal routing learned by a single drone is shared across the entire fleet (collective education) via an edge server.
As the number of operating units increases, it unleashes a powerful collective network effect where the entire drone fleet automatically becomes smarter, enabling multiple AIs to drive mutual growth while executing autonomous group actions.

"Autonomous Distributed AI Cells" can be integrated into predictive maintenance systems for individual production lines or entire factories.
For example, anomaly detection models can be individually optimized based on the unique operational variations of each robotic arm, enabling significantly more accurate failure predictions.

Currently, in many factories, data transmission between sensors, actuators, etc. and control devices is performed in an exclusive environment and via wired communication, with no external Internet connection allowed for the purpose of maintaining confidentiality.

Here, wired communication is complicated by cables, and the diameter of the cable bundles around the control unit, where the greatest number of connections are concentrated, can be several tens of centimeters in diameter. The more connections there are, the more physical space is taken up by the cables, and the more time is required for fault detection and replacement.

For this reason, there have been attempts to realize wireless communication for some time, but the number of channels for wireless communication has exceeded the allowable range due to data transmission/reception between multiple sensors and actuators and control devices, and it has been determined that a wireless connection is not feasible in reality. Furthermore, the risk of information leakage to the outside is extremely high in wireless data transmission, and together with the problem of the number of channels, this is a major issue.

Securing in-device wireless communication achieves both reduced wiring space and lower costs alongside highly reliable remote monitoring.
Sensor communications within the device are converted to wireless using Hash Chain, transmitting only validated data to the control CPU, while raw data can be saved to local servers such as distributed file systems.
Furthermore, integrating from the master to the cloud via blockchain enables tamper-proof remote monitoring and data protection from external offices.

Tamper-proof manufacturing performance data builds a robust foundation of trust secured from the physical layer.
Slaves are deployed on equipment and sensors at each manufacturing step, encrypting and chaining production and inspection data sequentially at every phase using Hash Chain.
Aggregating this data into the MES (Manufacturing Execution System) via the master enables "tamper-proof manufacturing proof," confirming that components and products were crafted through correct procedures.
This delivers robust traceability without burdening system overhead, dramatically elevating quality assurance reliability.

Wirelessly connects numerous devices simultaneously over a "single channel" while preventing takeovers and runaways caused by spoofed commands.
In ad-hoc wireless environments such as drone swarms and smart mobility where devices connect directly, Hash Chain secures communications between slaves and the master.
Even in congested radio frequency environments, our proprietary protocol connects multiple devices using just one channel to enable secure network architecture.
Fake control commands are instantly detected and rejected at the 1ms level, preventing accidents caused by remote hijacking while ensuring reliable automated multi-unit operation.