【作者】 李岳明；

【导师】 万磊；

【作者基本信息】 哈尔滨工程大学 ， 船舶与海洋结构物设计制造， 2013， 博士

【摘要】 运动控制是自主式水下机器人进行自主作业的基础，是水下机器人的核心技术之一，因此对其进行深入研究具有重要的理论意义与工程应用价值。本文研究目的在于改善运动控制效果、提高运动控制系统的智能水平、完善控制软件的自诊断与容错处理，从而保障水下机器人具有良好的运动控制性能，能够适应复杂的海洋环境，同时保证运动控制系统具有较高的可靠性。论文研究对象为哈尔滨工程大学研制的某自主式水下机器人，该机器人搭载了多种探测传感器，具有多种探测作业功能。由于各种作业任务对水下机器人运动的需求不同，要求运动控制系统能够根据各种任务的特点，调整控制策略及控制方法，保证探测任务的顺利进行。同时要求运动控制系统能够对来自于系统外部的信息进行有效的理解，从而提高运动控制的执行效率和控制效果。另外，机器人系统可靠性要求运动控制软件具有自诊断与容错能力。因此本文重点研究了水下机器人运动S面控制方法的稳定性问题、欠驱动运动控制问题、外部信息理解技术与运动控制软件的自诊断与容错技术。针对水下机器人S面运动控制方法的稳定性问题，从位置控制和速度控制两方面进行了分析。应用李亚普诺夫稳定理论和T-无源理论，完成了S面位置控制器的稳定性证明。在位置控制器基础上推导了S面速度控制器，并对速度控制器的稳定性进行了分析。对S面控制进行自适应改进，提出了自适应S面控制方法，进行了实际海上试验验证，并给出了试验结果。针对机器人在快速航行时由于关闭槽道桨产生的横移与升沉自由度欠驱动问题，采用椭圆视线法与反步法构造了欠驱动直线路径跟踪控制器，实现了对直线路径的跟踪控制，给出了海上试验结果；利用T-S模糊融合的机器人高度信息构造了欠驱动地形跟踪控制器，实现了对水底地形的跟踪控制，并进行了试验验证；构造了基于虚拟控制量的欠驱动深度控制器，实现了带有剩余浮力的水下机器人欠驱动深度控制，并进行了仿真试验。针对规划指令与水声目标两种外部信息的理解问题分别进行了研究。借鉴Itti视觉注意力模型，建立了规划指令理解模型，并且根据规划指令的理解结果进行了再次规划，以便于规划系统期望的运动状态能够以合理的方式实现，并进行了仿真试验验证；应用带有遗忘因子的动态聚类的方法实现了水声目标的分类与甄别，并采用强跟踪卡尔曼滤波器对跟踪目标运动信息进行滤波，提高信息准确度，在水池目标跟踪试验对水声目标信息理解行了验证。针对运动控制软件的自诊断与容错问题，将基于vxWorks操作系统的运动控制软件进行分解，得到了若干个由函数组成的系统级任务与普通任务，并从系统级、任务级、函数级三个层次进行了软件故障自诊断与容错设计，保证了运动控制软件在特定故障时的可靠性。更多还原

【Abstract】 As the autonomous work basis of autonomous underwater vehicle (AUV), motioncontrol is one of the key technologies of AUV. Thus, it is significant in theory and engineeringto research on the motion control. This thesis aims at improving motion control effect,enhancing the intelligence of motion control system and modifying the fault diagnosis andtolerance ability of control software, so that better motion control performance, higherreliability and fitness for complex ocean environment can be acquired.A newly developed AUV which is taken as the research object is equipped with kinds ofdetective sensors and so can do kinds of detective works. Since there are differentrequirements for the motion performance in accordance with different working tasks, themotion control system has to adjust control strategy and control method according to thecharacteristics of different tasks. Meanwhile, the motion control system has to efficientlycomprehend the outer information in order to improve the execution efficiency and controllerperformance. Besides, the system reliability of the AUV needs a motion control system withfault diagnosis and tolerance abilities. And the stability of the motion control algorithmS-plane control algorithm, under-actuated motion control, comprehension technology of theouter information, and fault diagnosis and tolerance of motion control software are reseachedin this dissertation.Aiming at the stability of S-plane motion control method, it is analyzed in two aspects,position control and speed control. Using Lyapunov stability theory and T non-source theory,the stability of the S-plane position controller is proved. And S-plane speed controller isderived based on the position controllet, then the stability of speed controller is analyzed. Theadaptive S-plane control method is proposed by modifying the traditional S-plane control, andit is tested during real sea trial in which the experiments results are acquired.The under-actuated problems of sway and heave motion happen owing to shutting downthe tunnel thrusters while the AUV moves at a high speed, in order to solve these, the ellipsesight ray method and anti-step method are used to construct the under-actuated line path tracking controller, and the line path trackingcontrol is achieved, the ses trial results aregiven. The under-actuated terrain tracking controller is built by using the height data fused inT-S fuzzy method, and the terrain tracking under sea is achieved, the method is tested in seatrial. The under-actuated depth controller based on virtual control varible is built, so that thedepth control of the AUV with excess buoyancy is avhieved, the method is tested insimulation experiments.Researches on two kinds of outer information, comprehension of planning instructionsand acoustic targets, are discussed. The planning instruction comprehension model is builtaccording to the thoughts of Itti vision attention model, and the re-planning is done accordingto the comprehension of planning instruction, so that the expected motion state of planningsystem can be realized in a reasonable way, and the method above is tested in simulationexperiments. The classification and identification of acoustic targets are achieved usingdynamic clustering method with forgotten factor. Besides, the motion information of trackingtargets is filtered by using strong tracking kalman filter to make the information more precise,and the method is tested by target tracking trial in tank.To slove the fault diagnosis and tolerance problems of motion control system software,the motion control system software based on vxWorks OS is divided into system level tasksand normal level tasks formed by several functions. And software fault diagnosis andtolerance is designed from three levels, system level, task level and function level, thus, thereliablity of motion control system software is guaranteed when certain faults happen.更多还原