GPU-NPU-based ai system and method for dynamic update and distributed processing
The dynamic update system addresses performance degradation and network costs by integrating cloud-based GPU training with edge-based NPU inference, ensuring adaptive and efficient AI model updates in resource-constrained environments.
Patent Information
- Application Number
- PCT/KR2025/019017
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2025-06-19
- Filing Date
- 2025-11-18
- Publication Date
- 2026-05-21
AI Technical Summary
Conventional AI systems in edge computing environments face challenges in flexibly responding to new situations and suffer from performance degradation due to fixed AI models, with separate GPUs and NPUs operating independently, leading to excessive network costs and energy consumption.
A dynamic update system is implemented, where a cloud server trains AI models using GPUs and edge devices perform real-time inference with low-power NPUs, enabling adaptive learning and model updates based on situational awareness, with model management and feedback mechanisms to ensure continuous performance.
The system achieves high-performance AI inference and adaptability by autonomously updating models, optimizing for resource constraints, reducing network costs, and improving model accuracy through distributed processing and quantization.