Hovering deep-sea mining provides a new direction for deep-sea nodule mining;however,accurate trajectory tracking of hovering mining vehicle remains a key challenge.This paper presents an adaptive RBF network sliding ...Hovering deep-sea mining provides a new direction for deep-sea nodule mining;however,accurate trajectory tracking of hovering mining vehicle remains a key challenge.This paper presents an adaptive RBF network sliding mode controller(ARBFNNSMC)for trajectory tracking of hovering mining vehicles.The ARBFNNSMC control loop features an RBF network to compensate for lumped external disturbance and system uncertainty,online adjustment of the RBF network parameters are achieved via adaptive laws derived based on Lyapunov function to guarantee closed-loop stability.In addition,a new type of smooth switching term is proposed on the basis of the plate-pole capacitor model for reduced chattering and adopted in ARBFNNSMC.To assess control performance and robustness,numerical simulations were performed based on three typical hovering mining trajectories.The simulation results demonstrate that the proposed controller achieved excellent robustness in all simulation cases,attaining reduced tracking error,overshoot,and settling time with improved chattering suppression in control output.Compared to conventional sliding mode controller,the mean RMS tracking error and settling time were 30.9%and 61.8%lower,respectively,while the thrust oscillation was reduced by 57.3%.展开更多
Interpretable models are essential for deploying deep-learning techniques in marine activities.However,the layered complexity of state-of-the-art deep-learning architectures hinders mechanistic insight and limits adop...Interpretable models are essential for deploying deep-learning techniques in marine activities.However,the layered complexity of state-of-the-art deep-learning architectures hinders mechanistic insight and limits adoption.Here,we introduce a three-degree-of-freedom surrogate model that renders a multivariate long term short-memory network with multiple input and output channels(MIMO-LSTM)for autonomous underwater vehicles(AUVs)transparent and tractable.The surrogate is built on a least-squares support-vector machine selected for its superior approximation and generalization capacity.Benchmark manoeuvres show that the surrogate retains predictive fidelity up to 95.8%while accelerating inference by 98.6%relative to the parent MIMO-LSTM.Interpretable parameter-dependence plots quantify the contribution of individual variables and deliver mechanistically grounded,transparent forecasts of AUV manoeuvring dynamics.展开更多
基金supported by the National Key Research and Devel-opment Program of China under grant number 2021YFC2801600the National Natural Science Foundation of China(NSFC)under grant number 42206189 and 52301328.
文摘Hovering deep-sea mining provides a new direction for deep-sea nodule mining;however,accurate trajectory tracking of hovering mining vehicle remains a key challenge.This paper presents an adaptive RBF network sliding mode controller(ARBFNNSMC)for trajectory tracking of hovering mining vehicles.The ARBFNNSMC control loop features an RBF network to compensate for lumped external disturbance and system uncertainty,online adjustment of the RBF network parameters are achieved via adaptive laws derived based on Lyapunov function to guarantee closed-loop stability.In addition,a new type of smooth switching term is proposed on the basis of the plate-pole capacitor model for reduced chattering and adopted in ARBFNNSMC.To assess control performance and robustness,numerical simulations were performed based on three typical hovering mining trajectories.The simulation results demonstrate that the proposed controller achieved excellent robustness in all simulation cases,attaining reduced tracking error,overshoot,and settling time with improved chattering suppression in control output.Compared to conventional sliding mode controller,the mean RMS tracking error and settling time were 30.9%and 61.8%lower,respectively,while the thrust oscillation was reduced by 57.3%.
基金supported by the National Natural Science Foundation of China(Grant Nos.42376187 and 52131101).
文摘Interpretable models are essential for deploying deep-learning techniques in marine activities.However,the layered complexity of state-of-the-art deep-learning architectures hinders mechanistic insight and limits adoption.Here,we introduce a three-degree-of-freedom surrogate model that renders a multivariate long term short-memory network with multiple input and output channels(MIMO-LSTM)for autonomous underwater vehicles(AUVs)transparent and tractable.The surrogate is built on a least-squares support-vector machine selected for its superior approximation and generalization capacity.Benchmark manoeuvres show that the surrogate retains predictive fidelity up to 95.8%while accelerating inference by 98.6%relative to the parent MIMO-LSTM.Interpretable parameter-dependence plots quantify the contribution of individual variables and deliver mechanistically grounded,transparent forecasts of AUV manoeuvring dynamics.
