Optimized Trajectory Tracking for ROVs Using DNN + ENMPC Strategy

Autor: Guanghao Yang, Weidong Liu, Le Li, Jingming Xu, Liwei Guo, Kang Zhang
Jazyk: angličtina
Rok vydání: 2024
Předmět:
Zdroj: Journal of Marine Science and Engineering, Vol 12, Iss 10, p 1827 (2024)
Druh dokumentu: article
ISSN: 2077-1312
DOI: 10.3390/jmse12101827
Popis: This study introduces an innovative double closed-loop 3D trajectory tracking approach, integrating deep neural networks (DNN) with event-triggered nonlinear model predictive control (ENMPC), specifically designed for remotely operated vehicles (ROVs) under external disturbance conditions. In contrast to single-loop model predictive control, the proposed double closed-loop control system operates in two distinct phases: (1) The outer loop controller uses a DNN controller to replace the LMPC controller, overcoming the uncertainties in the kinematic model while reducing the computational burden. (2) The inner loop velocity controller is designed using a nonlinear model predictive control (NMPC) algorithm with its closed-loop stability proven. A DNN + ENMPC 3D trajectory tracking method is proposed, integrating a velocity threshold-triggered mechanism into the inner-loop NMPC controller to reduce computational iterations while sacrificing only a small amount of tracking control performance. Finally, simulation results indicate that compared with the ENMPC algorithm, NMPC + ENMPC can better track the desired trajectory, reduce thruster oscillations, and further minimize the computational load.
Databáze: Directory of Open Access Journals