基于多模型优化切换的海洋机器人运动控制研究 | |
Alternative Title | Optimal Laws of Multiple Models Switching Control for Unmanned Marine Vehicle Motion |
周焕银1,2 | |
Department | 水下机器人技术研究室 |
Thesis Advisor | 封锡盛 ; 刘开周 |
Classification | TP242 |
Keyword | 海洋机器人 多模型切换控制 动态状态反馈控制 动态滑模控制 控制库 |
Call Number | TP242/Z74/2011 |
Pages | 130页 |
Degree Discipline | 模式识别与智能系统 |
Degree Name | 博士 |
2011-11-25 | |
Degree Grantor | 中国科学院沈阳自动化研究所 |
Place of Conferral | 沈阳 |
Abstract | 本文主要针对两类海洋机器人(Unmaned Marine Vehicle, UMV)系统即自主水下机器人AUV(Autonomous Underwater Vehicle,AUV)与近水面机器人USV(Unmanned Surface Vehicl,USV)系统作为被控对象进行运动控制算法的研究。本文所研究的AUV系统是一种‘水下’机器人，且根据任务需要需改变自身结构具有两种控制模式。USV系统是一种‘近水面’运行的系统，根据任务需要即可在水面运行又可下潜到一定深度运行，系统下潜深度不同所受外界干扰不同，同时系统运行过程中，由于自身原因系统稳心高发生变化，故系统在运行过程中其控制模型会发生变化。 虽然比例积分微分(Proportional-Integral-Derivative，PID)控制算法由于其控制参数控制作用明确而得到众多使用者的青睐，但是如下控制弱点却不能忽视即比例环节会引起系统超调，微分环节过大会造成系统不稳定，积分环节则会引起系统调节时间变长等。由于UMV系统的强非线性与状态变量的强耦合性以及外界环境变化的影响，为了获取不同条件下鲁棒性较强的PID控制参数需要相关控制专家做大量的试验对控制参数进行分段调整。控制参数分段条件的约束，造成控制参数的离散化而引起系统运动控制性能受约束于所设的分段条件，如根据速度而分段设置的深度控制参数，在速度变化时由于控制参数的变化深度很有可能会发生抖动。 为了获取一种通用性强、控制性能好，且具有可扩充性的控制模块适合于这两类UMV系统，提出了构建模型集设置控制库的多模型控制算法。多模型控制算法实现系统底层控制能够根据控制模型变化而自动调整控制策略无需专家经验进行调整的目的。多模型控制算法本身所具有控制优势是能够根据系统控制模型的变化而自动调整控制策略且控制参数连续变化。为了实现UMV系统具有良好的控制性能，提出多控制策略如动态反馈法保证系统运动误差指数衰减从而实现系统运动控制无超调或超调小、调节时间短等控制性能，采用滑模控制解决UMV系统的强非线性受外界干扰大等问题。 本文主要针对UMV系统的航向控制与深度控制进行多模型控制算法的构建，主要研究内容如下： (1) UMV多模型集的构建。针对具有水动力参数的UMV系统，通过解耦方式将系统分为纵向速度模型集、航向控制模型集以及深度控制模型集共3个子模型集。通过对UMV系统6个动力学方程与6个运动学方程解耦，在子模型集下构建了多个子模型。针对无水动力参数的UMV系统，通过最小二乘辨识法构建了系统的航向与深度控制的临时子模型集。 (2) 多模型切换控制策略的研究。多模型切换问题是多模型控制中的关键问题，切换策略设计的优劣关系到多模型切换系统能否稳定。针对系统各控制模型特点提出了三种类型的多模型切换策略，即基于权值设置的线性多模型切换，基于能量衰减的非线性能量切换以及非完全同态的多模型控制切换策略，并通过数字仿真验证了切换控制策略的合理性。 (3) UMV运动控制策略的优化研究。优良的运动控制策略是保证UMV系统具有良好控制性能的关键。根据所构建的UMV模型集中各控制模型特点提出了基于神经网络补偿的状态反馈法、基于滑模面指数衰减的滑模控制法、以及基于控制库的控制算法等。通过MATLAB仿真与半物理仿真平台验证了这些控制算法的可行性。 (4) 基于控制参数估算的近水面UMV(USV)湖/海试现场试验研究。USV系统是一种近水面运动载体其受外界环境影响较大。在USV海试中，采用自调整PID控制系统的航向与参数自调整的动态反馈法控制系统的深度，试验数据分析表明所设计的控制算法具有很强的抗干扰能力与鲁棒性。在USV湖试中，针对系统的航向控制构建了基于辨识模型集参数估计的动态状态反馈法，针对系统的深度控制采用了动态滑模控制，试验表明系统具有良好的动态性能。 (5) 基于多模型切换的AUV湖试现场试验研究。基于多模型切换的航向动态反馈法与基于多模型的深度动态滑模控制法，多次湖泊试验证明这两种控制算法具有良好的动态性能，且对耦合项的影响具有较强的鲁棒性。 综上所述，本文主要针对UMV系统自身非线性与外界环境干扰无法预测性进行多模型控制算法的研究，针对UMV系统多模型集，分析了线性与非线性多模型的切换控制策略，针对UMV系统运动模型特点对不同类型的控制器进行优化。在不同UMV载体(USV、AUV)上进行了多模型运动控制现场试验，试验证明所构建的控制模块具有较强的抗干扰能力与通用性。 |
Other Abstract | The great devotion of Unmanned Marine Vehicles (UMVs) has been drawn a great deal of attention in many fields. With the self-governing capabilities UMVs are able to reliably perform tasks in different environments. Some technologies on UMV must be improved according to users’ demands. However, a number of complex issues of UMV make it difficult to cruise underwater or surface because of the unstructured, hazardous ocean environments. The highly nonlinear, six states coupling, uncertain and easily effected hydrodynamic coefficients and disturbances by water currents, all these factors make UMVs difficult to control. Multi-model control algorithm is advised for UMVs to improve control performances. The algorithm is developed by two kinds of UMVs which are Autonomous Underwater Vehicle (AUV) and Unmanned Surface Vehicle (USV). As two control objects of the algorithm, some different control conditions of them should be considered. AUV is one kind of UMV which travel underwater while USV can work surface as well as some certain depth underwater. AUV can change its configuration according to its tasks as well as USV is disturbed by different ocean environments. Although the PID (Proportional-Integral-Derivative) controller has been popularly used by many UMV users, its drawbacks can not be neglected such as large overshoot and long adjustment time. To get ideal control results, some authors have to adjust control parameters of PID according different conditions. To make these two kinds of UMVs work well under one control frame, the multi-model algorithm is designed including three kinds of model sets and control bank. Three kinds of models come from hydrodynamic equations and motion transform equations. The control bank is composed of some control algorithms which should be improved according to these models. The multi-model algorithm serves for the depth control and heading control of UMVs motions. The major contents are organized as follows. Firstly, to make every thing work well, multiple models sets of UMVs should be conceived. Fixed model sets are composed of surge speed control model subset, heading control model subset and depth control model subset, as well as every subset are made up of some sub-models. These three subsets are decomposed from six hydrodynamic equations and six motions transform equations. Meanwhile, some identification model sets are considered for USV which has not any hydrodynamic coefficients. The least square method is introduced to the identification method to obtain heading control temporal models subsets and depth control temporal models subsets. Secondly, some switching control laws are advised for model subsets. Switching control laws are the key factor which influences the stability of the sub-model switching moment. Some multi-model switching control laws are proposed for different sub-models such as linear sub-models switching control laws, nonlinear sub-models switching control laws and different states switching control laws. All these laws are validated their smoothing during switching moment by some simulations. Thirdly, some improved control algorithms are introduced to sub-models of UMVs. It is well known that one robust control is important for motions of UMVs. Neural network control compensator with dynamic state feedback control is designed for some depth sub-models. Sliding mode control algorithm with exponential decay sliding surface serves as these three subsets. One control bank which is included by some control algorithms is recommended for all UMVs fixed subsets. All these mentioned algorithms are validated their robust by MATLAB simulations and virtual environment simulations. Fourthly, some control algorithms for USV are test under different sea conditions and under lake conditions. As we know USV is one kind of UMVs which can cruise surface and travel underwater also. The around environments of USV have an immediate effect on the vehicle so these control algorithms for USV should not be sensitive to disturbance. The self-adjusted PID control algorithm and self-adjusted dynamic control algorithm are presented for heading control and depth control separately. The motions of USV are strongly robust under different sea conditions all that denote these algorithms are robust to disturbances. In lake tests, some control algorithms are designed base on these identification models. These algorithm shows that they have good control performances. Finally, some lake tests are arranged for AUV system to verify the control performances of multiple models switching control algorithms. Multi-model dynamic state feedback control algorithm works for the heading control models meanwhile the multi-model dynamic sliding mode control algorithm is introduced to the depth sub-models. Some lake tests denoted that all these algorithms make their corresponding states beautiful dynamic control performances as well as they are insensitive to coupled stated. Above all, this thesis was wholehearted for multi-model control algorithm based on USVs which have different models and are influence diverse environments. Some switching control laws are considered to different kinds of model sets such as linear model sets, nonlinear model set and different state model sets. At last some tests on different kinds of UMVs admitted that the multi-model control algorithm were robust and well control performances. |
Language | 中文 |
Contribution Rank | 1 |
Document Type | 学位论文 |
Identifier | http://ir.sia.cn/handle/173321/9272 |
Collection | 水下机器人研究室 |
Affiliation | 1.中国科学院沈阳自动化研究所 2.中国科学院研究生院 |
Recommended Citation GB/T 7714 | 周焕银. 基于多模型优化切换的海洋机器人运动控制研究[D]. 沈阳. 中国科学院沈阳自动化研究所,2011. |
Files in This Item: | ||||||
File Name/Size | DocType | Version | Access | License | ||
基于多模型优化切换的海洋机器人运动控制研（1646KB） | 开放获取 | License | Application Full Text |
Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.
Edit Comment