Safe deep reinforcement learning-based adaptive control for USV interception mission. (15th February 2022)

• Record Type: Journal Article
• Authors: Du, Bin; Lin, Bin; Zhang, Chenming; Dong, Botao; Zhang, Weidong
• Is Part Of: Ocean engineering. Volume 246(2022)
• Publication Date: 2022-02-15
• Subjects: Unmanned surface vessels -- Safe reinforcement learning -- Data-based learning control -- Uniformly ultimate bounded stability -- Interception mission; Ocean engineering -- Periodicals; Ocean engineering; Periodicals; 620.4162
• DOI: 10.1016/j.oceaneng.2021.110477
• Languages: English
• ISSNs: 0029-8018

Abstract

Abstract: This paper aims to develop a safe learning scheme of the USV interception mission. A safe Lyapunov boundary deep deterministic policy gradient (SLDDPG) algorithm is presented for the USV interception mission. The uniformly ultimate bounded (UUB) stability of control systems is analyzed under finite safety constraints. A single neuron proportional adaptive control (SNPAC) is applied to pre-train the deep policy network for speeding up the training process. The proposed method is evaluated by a series of simulations of the USVs interception and tracking mission. Compared with the existing results, our method can fast converge to the feasible solution subject to safety constraints and demonstrate a high performance in stability and safety by virtual-reality experiments. Highlights: A learning-based method is presented to achieve optimal automatic interception in a complicated dynamic scenario. The UUB stability of the SLDDPG is analyzed to guarantee the safety and stability of the RL control. A single neuron proportional adaptive control (SNPAC) method is given to pre-train the policy network.