基于深度对比强化学习的无线路由优化方法及网络系统
By combining a deep contrastive reinforcement learning model with multi-dimensional routing metrics, the problem of uneven energy consumption in traditional routing algorithms and high resource consumption in deep reinforcement learning algorithms is solved, achieving efficient and adaptive routing optimization on IoT devices.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHONGQING KELANDA TECH CO LTD
- Filing Date
- 2023-12-26
- Publication Date
- 2026-07-17
AI Technical Summary
Traditional routing algorithms in wireless multi-hop networks suffer from problems such as unreasonable cluster head selection, uneven energy consumption, and uneven path load. Furthermore, routing optimization algorithms based on deep reinforcement learning consume a lot of computational and storage resources on resource-constrained devices and are difficult to adapt to dynamic environments.
A deep contrastive reinforcement learning model is adopted to improve the model's generalization ability and adaptability through contrastive learning. It combines multi-dimensional routing metrics such as the energy of candidate forwarding nodes, hop count, and buffer queue count, and adopts a centralized training and distributed interaction architecture to reduce the computational complexity of terminal nodes.
It achieves efficient and adaptive routing optimization on resource-constrained nodes, improving network lifetime and data transmission reliability, and reducing data acquisition costs and computational complexity.
Smart Images
Figure CN117749692B_ABST