Coordinated SLIPT and TDD Optimization for Energy-Deprived Seabed Sensors Using Meta-Learning Enabled ROV

초록

This paper considers underwater optical wireless communication between a remotely operated vehicle and multiple energy-deprived seabed sensors. The ROV first delivers control information and optical power through simultaneous lightwave information and power transfer (SLIPT), and each sensor then uploads sensing data using only the harvested energy within a time-division duplex frame. The resulting design problem jointly optimizes the time-division duplex mode-switching ratio, the SLIPT time-switching ratio, and the SLIPT power-splitting ratio under strong coupling between downlink reliability and uplink energy availability. To solve the nonconvex online control problem under dynamic underwater channels, a meta-learning-enhanced soft actor-critic (Meta-SAC) algorithm is adopted. By reusing knowledge accumulated from previously served sensors, the proposed method attains approximately 11.7 bits per second per Hertz uplink throughput, remains within 4.1 percent of optimal search, improves the rate by up to 44.4 percent over the strongest benchmark, and reduces convergence time by 62.5 percent relative to standard SAC, while requiring only 0.489 ms inference time per slot.

키워드