Decentralized Network Architecture

Ginkai operates across three distinct architectural layers to ensure scalability and security:

• The Settlement Layer (Base L2): Handles token transactions, Skill NFT minting, and governance votes.

• The Perception Layer (Off-Chain/IPFS): Stores high-bandwidth robotic sensor data and environmental logs in a decentralized manner.

• The Logic Layer (AI Oracles): Translates raw telemetry data into executable "Skills" using decentralized compute power.

Deep Dive: Universal Tensor Protocol (UTP) and Data Normalization

A critical challenge in decentralized robotics is that different manufacturers (e.g., Tesla Optimus vs. Boston Dynamics Atlas) use different kinematic chains and sensor arrays. Ginkai solves this via the Universal Tensor Protocol (UTP).

The Normalization Layer:

Ginkai Edge Adapters convert proprietary telemetry into a standardized vector space before it reaches the blockchain.

1. Input: Proprietary data (e.g., [Brand_X: Left_Arm_Angle_45, Torque_12Nm]).

2. Process: The data passes through a normalization function $f(x) \rightarrow V_{utp}$, creating a unified 768-dimensional motion vector.

3. Output: A standardized .gnk (Ginkai Motion File) capable of being read by any compatible robot.

Cross-Morphology Transfer:

This architecture enables "Cross-Morphology Learning." A task learned by a 1.60m robot can be mathematically scaled and transferred to a 1.80m robot. The system utilizes a retargeting algorithm that adjusts the kinematic trajectory while preserving the semantic goal of the movement.