Online Identification Framework for AUV Based on PD-like Event-Triggered Mechanism and Data Filtering Window

The development of the Autonomous Underwater Vehicle (AUV) has garnered significant attention. Establishing an accurate mathematical model is playing a crucial role in AUV control. In this paper, the least squares support vector machine (LS-SVM) algorithm is applied to identify the vertical model of AUV for long-term operation. Additionally, a PD-like (proportional and derivative-like) event-triggered condition is proposed to determine the appropriate time of model updates. In addition, an online identification framework for AUV models is established by adopting a filtering window structure. The stochastic excitation experiments are conducted on simulated data to demonstrate that the proposed framework is highly adaptable and capable of maintaining a balance between model accuracy and computation time during long-term AUV operation.