华中科技大学学报（自然科学版）2026，Vol.54Issue(4)：14-21,8.DOI:10.13245/j.hust.250484
融合CNN洋流预测的长航程AUV节能控制方法
Energy-efficient control method for long-endurance AUVs integrated with CNN-based ocean current prediction
摘要
Abstract
For the energy-saving control problem of long-range autonomous underwater vehicles(AUVs)in ocean-current environments,a hierarchical model predictive control(MPC)method integrating ocean-current prediction by a convolutional neural network(CNN)was proposed.First,a CNN-based ocean-current predictor was established,by which the ocean-current field in the large-scale spatiotemporal region involved in AUV operations was forecasted.Then,an energy-consumption objective function was constructed based on the predicted ocean-current state variables,and the energy-optimal dynamic desired speed and heading angle were obtained through online solution.Finally,an MPC controller satisfying the AUV dynamics and boundary constraints was designed,in which the quadratic objective function was minimized with respect to the AUV desired-state deviations and the control inputs,and a further reduction in navigation energy consumption was achieved.Theoretical analysis and simulation results show that good robustness is possessed by the proposed scheme,and the total navigation energy consumption is reduced by 12.2％compared with the comparative method.关键词
神经网络/视线法制导/最优控制/欠驱动自主水下航行器/模型预测控制Key words
neural network/line-of-sight guidance/optimal control/underactuated AUV/model predictive control分类
信息技术与安全科学引用本文复制引用
孙玉山,舒国洋,张英浩,林宇涵..融合CNN洋流预测的长航程AUV节能控制方法[J].华中科技大学学报（自然科学版）,2026,54(4):14-21,8.基金项目
国家自然科学基金资助项目(52071104) （52071104）
黑龙江省自然科学基金资助项目(ZD2020E005). （ZD2020E005）
