USV路径跟随控制首先确定控制目标期望值,主要运用几何跟踪方法和基于运动模型的跟踪方法。几何跟踪方法包括视距导航(line of sight,LOS)、纯追踪算法和固定方位制导等,将路径跟随控制问题转化为航向跟踪问题。基于模型的跟踪方法包括模型预测控制(model predictive control,MPC)和滑模控制等,根据运动模型直接构造计划航迹的期望舵令。USV路径跟随控制中还需要获取控制对象的状态参数,包括系统状态和干扰。主要获取方式有卡尔曼滤波观测器和滑模观测器等,以及将干扰当作一种状态参数的扩张状态观测器(extended state observer,ESO)。文献[3]对观测器作了较详细的介绍。在控制器设计中,主要基于PID、滑模变结构、反步法以及最优控制理论设计控制器,实现航向或路径点的跟踪。USV路径跟随控制问题,需要同时考虑导航、观测和控制器3个部分。导航确定控制目标期望值,观测确定控制目标状态值,控制器确定控制指令输出值[4]。为提高USV路径跟随控制的精度、稳定性和实时性,国内外学者综合上述3个部分,提出了控制方法,积累了大量的研究成果。笔者将分析几种主要的控制算法对模型不确定性和风浪流干扰问题的处理方法。
针对USV路径跟随控制中模型不确定性和风浪流干扰问题以及几种常用控制方法,现有解决方法这些解决方法可归纳为以下3类:①在控制器导航部分引入自适应控制理论,使得控制器在外界干扰下,仍能获得稳定的期望艏向;②通过扩张状态观测器、神经网络以及最小二乘法等理论实时估计系统模型不确定项和干扰项;③结合2种或多种控制方法,充分利用各自的优势,获得更优的控制效果。单一的控制器在USV路径跟随控制中,难以有效应对模型不确定和风浪流干扰对系统的影响,通常需结合多种控制理论达到更好的控制效果。而其中多数方法需要建立精确数学模型来获得更优的控制效果,同时又势必增加计算量,加大工程应用难度。其中,基于PID理论的控制器,由于不需要建立复杂数学模型,设计简单,结合模糊逻辑整定PID参数,为工程应用中实现USV路径跟随控制的提供了较好的选择。而对于另一类无模型控制,即强化学习控制器,设计合理的奖励函数则是获得更优控制器性能的关键。基于此,提出无人船路径跟随控制的4种发展思路。1)当USV在干扰下艏摇频率大于某一阈值ωmax时,由于USV舵机等相关机械设备的机械特性,USV将难以通过转舵稳定在期望航向,只能通过对路径跟随过程中的偏航进行补偿实现控制。因此,对外界干扰进行分类,将干扰分为已知干扰、未知干扰,将未知干扰分为高频干扰和低频干扰,分别建立风浪流的主频干扰模型;然后通过最小二乘法,以及最小二乘支持向量机等辨识方法辨识干扰模型的主频率,根据ωmax实现低通滤波,和对高频分量的补偿,最终达到路径跟随控制,同时减少由于高频干扰的无效转舵。2)优化模糊逻辑控制器的结构和参数,包括隶属度函数和模糊规则。采用分层结构,减少规则库大小,增加对输入的敏感度,同时,优化模糊逻辑控制器结构和参数,可以有效降低去模糊化的复杂度,进一步提高其工程应用价值,见图2。例如,模糊逻辑与PID控制器结合,将一段时间序列的控制误差作为模糊推理机F1的输入;将主机转速、USV航速、转艏速度、转艏加速度以及水流速度作为F2的输入,F2输出USV航行状态,即推理USV处于转向航速下降或水流影响导致的转速变化状态;将风浪流干扰模型主频率作为F3的输入,输出干扰状态,即推理控制补偿量;F4根据航行状态修正控制补偿量;F5根据控制误差以及修正后的控制补偿量输出PID参数。文献[30]对模糊逻辑结合和参数的优化做了详细的分析。图2 分层模糊逻辑控制器 Fig.2 structures of hierarchical fuzzy systems3)结合多种控制方法的优势,分层设计控制器。如前文最优控制中所述MPC与LQG结合的分层控制,同时提高控制器控制精度和响应快速性。4)强化学习等智能控制方法也将是解决USV路径跟随问题的一个重要研究方向。将PID与强化学习结合,充分发挥强化学习处理模型不确定性和风浪流干扰影响下的控制问题的优势,实现自适应PID控制,如采用Q-learning、深度确定性策略(deep deterministic policy gradient,DDPG)整定PID参数{Kp,Ki,Kd},见图3。图3中,e(t)表示t时刻的航向误差,u(t)表示t时刻控制器输出的控制舵令,St表示t时刻USV的状态。在控制器设计中应考虑USV的时延性,采用多步时序差分求解梯度,避免小步长内船舶状态变化较小,梯度消失的问题;在奖励函数设计中充分考虑控制平滑性问题,加入不合理舵令的惩罚项,可以减少无效舵令。图3 基于DDPG的自适应PID控制器 Fig.3 adaptive PID controller based on DDPG另外,在强化学习初期中引入PID或MPC控制监督策略训练,减少训练时间,或者基于Dyna2结构等结合有模型和无模型强化学习,实现强化学习在USV上的工程应用。首先在环境干扰下设计线性MPC控制器,同时采用深层神经网络拟合该控制器,该网络即为有模型强化学习中所依据的模型,根据Dyna2结构结合有模型和无模型强化学习,将价值函数分为永久性记忆和瞬时记忆,永久性记忆利用与真实环境的交互经验来更新,瞬时记忆利用与模型的模拟交互来更新。
3 结束语
笔者对近年几种主要的USV路径跟随控制方法分别进行了分析,提出了各方法可能的发展方向,并分析了强化学习等智能控制方法在该问题上的应用前景,其中基于PID和模糊逻辑的控制方法设计简单,经济性好,工程应用广泛;随着工控机等的计算能力逐渐提高,非线性MPC的计算需求得到满足,工程应用价值也将逐步提升;在应对路径跟随控制系统日益智能化、复杂化的问题上强化学习等智能控制具有明显优势。基于上述分析,提出了USV路径跟随控制几种可能的发展方向,为相关研究提供参考。参考文献References[1] AZZERI M N,ADNAN F A,Md.Zain M Z.Review of course keeping control system for unmanned surface vehicle[J].Journal Teknologi,2015,74(5):11-20.[2] 郭 晨,汪 洋,孙富春,等.欠驱动水面船舶运动控制研究综述[J].控制与决策,2009,24(3):321-329.GUO Chen,WANG Yang,SUN Fuchun,et al.Survey for motion control of underactuated surface vessels[J].Control and Decision,2009,24(3):321-329.(in Chinese)[3] CHEN Wenhua,YANG Jun,GUO Lei et al.Disturbance-observer-based control and related methods:An overview[J].IEEE Transactions on Industrial Electronics,2016,63(2):1083-1095.[4] MCCUE L.Handbook of marine craft hydrodynamics and motion control[J].IEEE Control Systems,2016,36(1):78-79.[5] LIU Sheng,FANG Liang,GE Yaming,et al.GA-PID adaptive control research for ship course-keeping system[J].Journal of System Simulation,2007,19(1):3783-3786.[6] 欧阳子路,余文曌,贺宏伟,等.基于改进遗传算法的船舶航向PID控制器[J].中国航海,2017,40(1):13-15.OUYANG Zilu,YU Wenzhao,HE Hongwei,et al.PID control with improved genetic algorithm for ship steering[J].Navigation of China,2017,40(1):13-15.(in Chinese)[7] LIU Sheng,XINGBowen,ZHU Wanlu.A fusion fuzzy PID controller with real-time implementation on a ship course control system[C].2015 23thMediterranean Conference on Control and Automation(MED),Torremolinos:IEEE,2015.[8] FAN Yunsheng,LI Changfei,WANG Guofeng,et al.Design and validation of course tracking controller for unmanned surface vehicle[J].Journal of Dalian Maritime University,2017,43(1):1-7.[9] WANG Ru,SHEN Zhipeng.Fuzzy adaptive iterative sliding mode control for sail-assisted ship trajectory tracking[C].2017 4thInternational Conference on Information,Cybernetics and Computational Social Systems(ICCSS),Dalian:IEEE,2017.[10] WANG Renqiang,ZHAO Yuelin,SUN Jianming.Application of optimized RBF neural network in ship's autopilot design[C].Advanced Information Management,Communicates,Electronic&Automation Control Conference.Xi'an:IEEE,2017.[11] WANG Xianfeng,ZHAO Xiren,PENG Xiuyan,et al.Statistic model for LQG control of ship longitudinal motion[C].International Symposium on Systems&Control in Aerospace&Astronautics.Harbin:IEEE,2006.[12] WAHL A,GILLES E D.Model predictive versus linear quadratic control for the tracking problem of automatic river navigation[C].Control Conference,Karlsruhe:IEEE,2015.[13] 柳晨光.基于预测控制的无人船运动控制方法研究[D].武汉:武汉理工大学,2017.LIU Chenguang.Motion Control of Unmanned Surface Vehicles[D].Wuhan:Wuhan University of Technology,2017.(in Chinese)[14] 郭宝珠.非线性系统的自抗扰控制引论[J].数学建模及其应用,2017,6(1):13-22.GUO Baozhu.An introduction to active disturbance rejection control for nonlinear systems[J].Mathematical Modeling and its Applications,2017,6(1):13-22.(in Chinese)[15] LIU Cheng,ZOU Zaojian,YIN Jianchuan.Trajectory tracking of underactuated surface vessels based on neural network and hierarchical sliding mode[J].Journal of Marine Science and Technology,2015,20(2):322-330.[16] 邱 峰.船舶航向与航迹积分滑模自抗扰保持控制[D].大连:大连海事大学,2018.QIU Feng.Ship course and track integral sliding mode keeping control based on auto disturbance rejection control[D].Dalian:Dalian Maritime University,2018.(in Chinese)[17]ASHRAFIUON H,MUSKE K R,MCNINCH L C,et al.Sliding-Mode Tracking Control of Surface Vessels[J].IEEE Transactions on Industrial Electronics,2008,55(11):4004-4012.[18] 张 凯,李铁山,秦 奇,等.神经网络和二阶滑模算法在船舶航向控制中的应用[J].中国水运(下半月),2016,16(11):24-26.ZHANG Kai,LI Tieshan,QIN Qi,et al.The application of neural network and second order sliding mode on course control of ships[J].Chinese Water Transport,2016,16(11):24-26.(in Chinese)[19] LI Chen,WANG Yongji,LIU Lei.Path following control of ship based on time-varying threshold prescribed convergence law sliding mode[C].2018 37thChinese Control Conference(CCC),Wuhan:IEEE,2018.[20]TZENG C Y,GOODWIN G.C,CRISAFULLI S.Feedback linearization design of a ship steering autopilot with saturating and slew rate limiting actuator[J].International Journal of Adaptive Control&Signal Processing,2015,13(1):23-30.[21] ZWIERZEWICZ Z.On the ship course-keeping control system design by using robust and adaptive control[C].International Conference on Methods&Models in Automation&Robotics.Miedzyzdroje:IEEE,2014.[22] 张松涛,任 光.基于反馈线性化的船舶航向保持模糊自适应控制[J].交通运输工程学报,2005,5(4):72-76.ZHANG Songtao,REN Guang.Fuzzy adaptive control for ship steering autopilot based on backstepping technique[J].Journal of Traffic and Transportation Engineering,2005,5(4):72-76.(in Chinese)[23] DO K.D.Global path-following control of underactuated ships under deterministic and stochastic sea loads[J].Robotica,2016,34(11):26.[24] WITKOWSKA A,TOMERA M,ROMAN Mierzchalski.A backstepping approach to ship course control[J].International Journal of Applied Mathematics and Computer Science,2007,17(1):73-85.[25] LIU C,CHEN C L P,ZOU Z,et al.Adaptive NN-DSC control design for path following of underactuated surface vessels with input saturation[J].Neurocomputing,2017,267(1):466-474.[26] 沈智鹏,代昌盛.欠驱动船舶路径跟踪的强化学习迭代滑模控制[J].哈尔滨工程大学学报,2017,38(05):697-704.SHNE Zhipeng,DAI Changsheng.Iterative sliding mode control based on reinforced learning and used for path tracking of under-actuated ship[J].Journal of Harbin Engineering University,2017,38(05):697-704.[27] CARLUCHO I,DE Paula M,VILLAR S A,et al.Incremental Q-learning strategy for adaptive PID control of mobile robots[J].Expert Systems with Applications,2017,80(1):183-199.[28] LILLICRAP T P,HUNT J J,PRITZEL A,et al.Continuous control with deep reinforcement learning[J].Computer Science,2015,8(6):187-201.[29] 王文飒,梁 军,陈 龙,等.基于深度强化学习的协同式自适应巡航控制[J].交通信息与安全,2019,37(03):93-100.WANG Wensa,LIANG Jun,CHEN Long,et al.Collaborative adaptive cruise control based on deep reinforcement learning[J].Journal of Transport Information and Safety,2019,37(03):93-100.(in Chinese)[30]KONDRATENKOYP,SIMOND.Structuralandparametricoptimizationoffuzzycontrolanddecisionmakingsystems[J].StudiesinFuzzinessandSoftComputing,2018,361(1):273-289.An Overview of Path Following Control Methods for Unmanned Surface VehiclesZHANG Xuanwu1,2,3 XIE Lei1,2 CHU Xiumin1,2,4▲ XIE Shuo1,2,3LIU Chenguang1,2ZHANG Daiyong1,2,3(1.Intelligent Transportation Systems Research Center,Wuhan University of Technology,Wuhan 430063,China;2.National Engineering Research Center for Water Transport Safety,Wuhan University of Technology,Wuhan 430063,China;3.School of Energy and Power Engineering,Wuhan University of technology,Wuhan 430063,China;4.College of Physics and Electronic Information Engineering,Minjiang University,Fuzhou 350108,China)Abstract:The problems of non-linearity,time delay,system model uncertainty,and interference of wind,wave,and current are present in path following of unmanned surface vehicle(USV),which make it difficult to precisely control an USV in real-time.Therefore,several major methods of path following are analyzed.PID,feedback linearization,and backstepping are difficult to meet requirements of control precision when the motion of an USV is highly non-linear.Approaches that can suppress chattering of sliding mode control can be further optimized.For model predictive control,it is difficult to obtain high precision in real time.To improve control precision of fuzzy logic control,it is usually requires an increase in fuzzy rule base,resulting in complicated calculation.Intelligent control algorithms such as reinforcement learning have great application prospects,but the control performance needs to be improved and lack of relevant tests.Based on these,possible methods to improve control precision of PID,feedback linearization,and backstepping are summarized,and hierarchical control theory is introduced to solve the complicated calculation problem of fuzzy control and model predictive control.The possible directions for development of intelligent control such as reinforcement learning in USV path following control are shown.Key words:intelligent ship;USV;path following;control algorithm中图分类号:U675.91文献标识码:Adoi:10.3963/j.jssn.1674-4861.2020.01.003收稿日期:2019-06-28*国家自然科学基金项目(51709220)、国家重点研发计划(2018YFB1600400)、工信部高技术船舶研发专项(MC-201920-X01)、福州市科技计划项目(2018-G-92)资助第一作者简介:张旋武(1993—),硕士研究生.研究方向:智能船舶.E-mail:zhangxuanwu@whut.edu.cn▲通信作者:初秀民(1969—),博士,研究员.研究方向:智能船舶、交通安全与信息化.E-mail:chuxm@whut.edu.cn往期推荐阅读往期热文(点击文章标题即可直接阅读):