Predefined-time prescribed performance second-order sliding mode path following control for underactuated marine surface vehicles using self-structuring NN

• Published in: Ocean engineering 2024-10, Vol.309, p.118333, Article 118333
• Main Authors: Liu, Haitao, Zhou, Xuecheng, Tian, Xuehong, Mai, Qingqun, Sun, Ning
• Format: Article
• Language: English
• Subjects: Path following control; Predefined-time prescribed performance; Second-order sliding mode; Self-structuring neural network; Underactuated MSV
• ISSN: 0029-8018; 1873-5258
• DOI: 10.1016/j.oceaneng.2024.118333

This paper proposes a predefined time-prescribed performance second-order sliding mode path following controller for underactuated marine surface vehicles (MSVs) with unknown external environmental disturbances based on predefined time predictor line-of-sight (PTPLOS) guidance. First, a predefined time predictor is designed to address the problem of time-varying sideslip angles, and then a guidance law is designed. Second, a prescribed performance function with predefined times is proposed, which can not only set the rate of convergence and the final convergence range but also set an upper bound on the convergence time. Third, the predefined-time controller is designed in combination with second-order sliding mode control. Then, a self-structuring neural network (SSNN) is developed to approximate external disturbances and uncertainties, and this approach can adaptively adjust the number of neurons, reducing the computational burden while maintaining high approximation accuracy. After that, the closed-loop system is proven to be predefined-time stable by using of the Lyapunov stability theory. Finally, the control method proposed in this paper is validated through numerical simulations, demonstrating its effectiveness. •The proposed predefined-time convergence algorithm allows for a simpler setting of the convergence time limit compared to finite-time and fixed-time algorithms, which require complex parameter calculations.•The proposed predefined-time controller, utilizing a performance function and second-order sliding surface, ensures robust error convergence within a set range and time. An auxiliary system addresses input saturation, preventing occurrence and enhancing controller performance.•A self-structuring neural network (SSNN) is developed to approximate external disturbances and uncertainties, and it can automatically adjust the number of neurons to achieve better approximation performance while reducing the computational burden.