Unmanned surface vehicle escape strategy based on hybrid sampling deep Q-network

Table 1. Parameters of neural network
View table
View in Article
Table 1. Parameters of neural network
输入输出
输入层 21×1 −
隐藏层1 128×1 256×1
隐藏层2 256×1 256×1
隐藏层3 256×1 256×1
隐藏层4 256×1 128×1
隐藏层5 128×1 64×1
隐藏层6 64×1 4×1
输出层 − 4×1

Table 2. Parameters setting for HS-DQN algorithm
View table
View in Article
Table 2. Parameters setting for HS-DQN algorithm
参数数值
学习率α 0.001
折扣因子γ 0.9
ε-greedy探索策略概率 0.1
网络学习频率 10
经验池大小 10000
样本池大小NS 100
混合池大小$ {M_{\text{S}}} $ 64

Table 3. Comparison of evaluation indicators
View table
View in Article
Table 3. Comparison of evaluation indicators
DQN HS-DQN
平均值 −6.68 −4.41
方差 883.46 773.82

Table 4. Initial position and speed of pursuit ship
View table
View in Article
Table 4. Initial position and speed of pursuit ship
围捕船编号位置/m 速度/(m·s⁻¹)
1 [265.9, 346.1] [−11.498, 3.432]
2 [−449.5 146.8] [−11.655, 2.858]
3 [−197.6, −381.0] [2.992, −11.621]
4 [320.3, −252.4] [5.512, −10.659]

Tools

Get Citation

Copy Citation Text

Yuanpeng YANG, Lifei SONG, Jiaqi MAO, Yi LI, Houjing CHEN. Unmanned surface vehicle escape strategy based on hybrid sampling deep Q-network[J]. Chinese Journal of Ship Research, 2024, 19(1): 256

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites