A heterogeneous machine learning joint relay selection and power allocation decision method for underwater acoustic sensor networks
By combining a heterogeneous machine learning framework with UCB1 and DDPG algorithms, the coupling problem of relay selection and power allocation in underwater acoustic sensor networks was solved, achieving efficient energy and rate balance in dynamic environments and improving network performance.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- YANSHAN UNIV
- Filing Date
- 2026-03-24
- Publication Date
- 2026-07-17
AI Technical Summary
In existing underwater acoustic sensor networks, the optimization methods for joint relay selection and power allocation rely on accurate channel state information and static energy models, which makes it impossible to effectively balance energy and rate in dynamic underwater environments. Furthermore, existing machine learning methods cannot effectively handle the coupling of strategies, causing the optimization results to deviate from the global optimum.
A heterogeneous machine learning framework is adopted, combining the UCB1 algorithm for discrete relay selection and the DDPG algorithm for continuous power allocation. A cyclic interactive structure is used to realize bidirectional information transmission, and an energy-sensing reward function is designed to balance energy utilization.
It reduces computational complexity, improves the coordination and robustness of relay selection and power allocation, enhances network performance in dynamic environments, and achieves a dynamic balance between energy and rate.
Smart Images
Figure CN122419633A_ABST