Optimization of a Regional Marine Environment Mobile Observation Network Based on Deep Reinforcement Learning

The observation path planning of an ocean mobile network is important part the system. With aim developing a traditional algorithm to solve network, complex objective function needs be constructed, and improved deep reinforcement learning proposed. does not need establish function. agent samples marine environment information by exploring receiving feedback from environment. Focusing on real-time dynamic variability environment, our experiment shows that adding bidirectional recurrency Deep Q-network allows better estimate underlying system state. Compared with results existing algorithms, can effectively improve sampling efficiency platform. To prediction accuracy numerical system, we conduct experiments single platform, double five platforms. experimental show increasing number platforms but when exceeds 2, will accuracy, there certain degree decline. In addition, in multi-platform experiment, compared unimproved algorithm, proposed than algorithm.