毕业生知识库

历经多年发展，仿生机器鱼在结构设计与运动性能等方面已经取得了长足进步，然而其续航能力、作业能力等方面仍然存在较大欠缺。生物䲟鱼通过吸附功能搭乘大鱼游动的行为为解决上述问题提供了启发和灵感。通过模仿䲟鱼的吸附行为，有望为长航时机器鱼、子母式多机器鱼系统等研发和应用带来革新。基于此背景，本文围绕吸附式仿生机器鱼的系统设计和自主控制展开研究，主要内容如下：

一、针对现有机器鱼续航能力不足的问题，以生物䲟鱼为仿生对象，提出了一种新型吸附式仿生机器鱼系统设计方案，搭建了机电系统，完成了原型样机研制。实验表明，所研发机器鱼平台具有灵活的三维运动能力以及稳定可靠的吸附能力，能够模仿生物䲟鱼的搭便车行为，进而降低运动消耗。此外，综合考虑所提机器鱼丰富的胸鳍运动模态以及净浮力调节功能，基于牛顿-欧拉方法构建了多模态三维动力学模型，并开发了可视化运动仿真环境。实验表明所设计仿真环境具有良好的准确度，能够有效模拟多种机器鱼运动模态，为机器鱼控制算法的开发提供了可信的验证平台。

二、针对水下视觉定位易受折射影响的问题，分析了平面型和球面型防水罩的折射现象，提出了基于ArUco视觉码的水下双目视觉惯性定位方法和基于LED路标的水下鱼眼视觉定位方法。具体地，在折射现象分析方面，基于Snell折射原理，仿真讨论了两类防水罩的折射效应对相机成像视野以及畸变的影响，为机器鱼视觉系统的设计布局提供了指导。在双目视觉惯性定位方面，提出了一种基于平面折射校正的ArUco视觉码位姿估计算法，并利用误差卡尔曼滤波器将视觉数据与惯性数据融合，实现了稳定可靠的水下全局定位。在鱼眼视觉定位方面，提出了一种基于球面折射校正的L型LED路标位姿估计算法，包含LED图像检测、路标匹配、光路校正、位姿求解以及优化等步骤。地面和水下定位实验表明，所提算法的定位精度显著优于忽略折射的传统定位方法，为机器鱼的水下高精度作业提供了定位保障。

三、针对面向悬浮目标的机器鱼自主吸附任务，设计了基于有限状态机的自主控制策略，并提出了基于非线性模型预测的平面位置控制方法和基于胸鳍驱动与浮力调节的三维位姿控制方法。具体地，在控制策略设计方面，首先，定义了吸附任务场景；然后，基于机器鱼的运动特性，制定了包含目标趋近、位姿调节、吸附控制等六个状态在内的行为调度策略，设计了各状态对应的视觉与控制方案。在平面位置控制方面，考虑了机器鱼的欠驱动和输入耦合特性，推导了机器鱼速度约束，设计了方向-速度控制策略并证明了与位置控制的等价性，最后基于非线性模型预测算法实现了该控制策略。在三维位姿控制方面，所提方法包含独立的平面位姿控制器和深度控制器，其中平面位姿控制包括路径跟踪和精细调节两个阶段，并采用了基于胸鳍运动模态切换的控制策略；深度控制考虑了鱼鳍拍动、净浮力变化等扰动造成的影响，并利用非线性扰动观测器加以抑制。自主吸附任务的仿真和实验表明了所提控制方法的有效性。

四、针对面向移动目标的机器鱼自主吸附任务，设计了基于有限状态机的自主控制策略，并提出了基于线性矩阵不等式的鲁棒偏航控制方法和基于多模态运动的平面状态同步控制方法。具体地，在控制策略设计方面，首先，定义了吸附任务场景；然后，基于机器鱼的运动特性，制定了包含目标跟踪、同步吸附等五个状态在内的行为调度策略，设计了各状态对应的视觉与控制方案。在偏航控制方面，基于实验数据辨识得到了机器鱼的线性变参数偏航运动模型，并利用线性矩阵不等式方法设计了多目标优化控制器。在平面状态同步控制方面，所提方法包含路径跟踪控制和位置-速度控制，并采用尾鳍和胸鳍协调运动。其中路径跟踪控制负责实现航向和侧向位置同步，位置-速度控制则负责实现速度与纵向位置同步。自主吸附任务的仿真和实验表明了所提控制方法的有效性。

Bionic robotic fish has been developed for many years, where great progress has been made in mechanical design and motion performance optimization. However, the endurance and autonomous operation capability of robotic fish are still unable to fulfill the demand of the practical application. The hitch-hiking behavior of the suckerfish brings a brand new thought for addressing the above problems. By imitating the attaching behavior of suckerfish, it is promising to revolutionize the study of the long endurance robotic fish and the underwater mother-son multi-robot system. Based on such background, this dissertation mainly concerns the prototype design and autonomous control of the bionic robotic fish with attaching ability. The technical contributions are summarized as follows.

Firstly, in order to improve the endurance of the robotic fish, a novel design of bionic robotic fish with attaching ability is proposed. By constructing its mechanical, hardware, and software systems, the prototype is developed. The experimental results validate the strong maneuverability and reliable attaching capability. Further, considering the rich motion modes of pectoral fins and the buoyancy adjusting function, a multi-mode dynamic model based on the Newton-Euler method as well as a three-dimensional (3D) visual simulation environment are constructed for the proposed robotic fish. The experiment verifies the accuracy of the simulation environment, which can effectively simulate the various motion modes of the robot and provide a credible verification platform for the control algorithm.

Secondly, aiming at achieving accurate underwater visual localization, the refraction phenomenon of the planar and spherical waterproof housing is analyzed. Moreover, an ArUco-based stereo visual-inertial localization method and an LED-based visual localization method with the fisheye camera are proposed. In the aspect of refraction analysis, the visual field change and imaging distortion caused by refraction are discussed based on Snell's law, which can guide the future vision system layout of robotic fish. In the stereo-based method, a pose estimation algorithm with planar refraction correction is designed for the ArUco marker, and an error-state Kalman filter is applied to fuse the visual and inertial data. In the fisheye-based method, visual detection and pose estimation algorithm with spherical refraction correction is proposed for an L-shape LED marker. The experiments demonstrate the effectiveness of the proposed methods, which lays a solid foundation for the high-precision underwater operation of the robotic fish.

Thirdly, an autonomous control strategy based on a finite state machine (FSM) is designed for the hitch-hiking task aiming at a floating host. Besides, a nonlinear model predictive planar position control method, as well as a pectoral-fin-actuated and buoyancy-driven 3D pose control method, are proposed. In terms of control strategy design, the task scenario of hitch-hiking is first defined. Then, the FSM-based behavior strategy, as well as the control and vision schemes corresponding to every state, are designed. In the planar position control, an orientation-velocity control strategy is developed and proved to be equivalent with position control, and then a nonlinear model predictive controller is designed to implement this strategy. In the 3D pose control, the whole controller consists of the independent planar pose controller and depth controller. The planar pose control is divided into path following and fine adjustment stages, which apply the pectoral-fin-mode-switch-based control method. The depth controller utilizes the nonlinear disturbance observer to overcome the interference caused by fin flapping and buoyancy changing. The simulation and experiment of the autonomous hitch-hiking task demonstrate the effectiveness of the proposed control method.

Fourthly, an autonomous control strategy based on FSM is designed for the hitch-hiking task aiming at a moving host. Besides, a robust yaw controller based on linear matrix inequality (LMI) and a multi-mode planar state synchronization controller are proposed. In terms of control strategy design, the task scenario of hitch-hiking is first defined. Then, the FSM-based behavior strategy, as well as the control and vision schemes corresponding to every state, are designed. In the yaw control, a linear parameter varying yaw model is identified by the measured data, and then a multi-objective optimization controller applied LMI techniques is designed. In the synchronization control, the whole controller consists of the path following controller and position-velocity controller, which is driven by the coordination of the pectoral and caudal fins. The path following controller guarantees the synchronization of lateral position and orientation, and the position-velocity controller ensures the synchronization of longitudinal position and velocity. The simulation and experiment of the autonomous hitch-hiking task demonstrate the effectiveness of the proposed control method.