A method, device and medium for determining routing of an underwater acoustic sensor network
By combining Q-learning and Dijkstra's algorithms in an underwater acoustic sensor network, routing paths are optimized, solving the data transmission problem under dynamic underwater environment changes. This results in higher packet delivery rate, throughput, and lower latency, improving the network's adaptability and robustness.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHEJIANG UNIV
- Filing Date
- 2026-06-23
- Publication Date
- 2026-07-21
AI Technical Summary
In underwater acoustic sensor networks, existing routing algorithms struggle to achieve reliable, low-latency, and high-efficiency data routing in dynamically changing underwater environments. Classical algorithms tend to select unstable links, leading to a decline in overall network performance, while online learning algorithms consume significant resources and compromise plug-and-play functionality.
The Q-learning algorithm is used to perform offline iterative training on the initial Q-table to construct an offline link cost matrix. The routing path is optimized by combining the Dijkstra algorithm, integrating historical experience and real-time status, and determining the optimal path by reshaping the online initial link cost matrix.
It improves the data transmission adaptability and robustness of underwater acoustic sensor networks, increasing packet delivery rate by 13.7%, throughput by 23.4%, and reducing end-to-end average latency by 11.2%, significantly optimizing communication performance without changing the existing protocol stack.
Smart Images

Figure CN122437808A_ABST