基于高斯过程回归和深度强化学习的水下扑翼推进性能寻优方法

Processing math: 100%

杨映荷, 魏汉迪, 范迪夏, 李昂

Optimization Method of Underwater Flapping Foil Propulsion Performance Based on Gaussian Process Regression and Deep Reinforcement Learning

YANG Yinghe, WEI Handi, FAN Dixia, LI Ang

表5 传统TD3算法与GPR-TD3方法所需样本数量及习得动作

Tab.5 Number of samples and learned actions required by traditional reinforcement learning and GPR-TD3 methods

类别	传统强化学习样本数量	GPR-TD3样本数量	动作向量
推进速度局部最优	9 200	290	[ $\begin{array}{l} 0.700 & 55 & 65 & 2.860 \end{array}$ ]
推进效率局部最优	6 200	290	[ $\begin{array}{l} 0.700 & 55 & 65 & 2.837 \end{array}$ ]
推进速度100 mm/s	5 300	290	[ $\begin{array}{l} 0.501 & 55 & 65 & 3.630 \end{array}$ ]
推进速度80 mm/s	5 900	290	[ $\begin{array}{l} 0.500 & 55 & 53.615 & 4.500 \end{array}$ ]
推进速度70 mm/s	4 300	290	[ $\begin{array}{l} 0.675 & 55 & 52.178 & 2.734 \end{array}$ ]