【作者】 武星；

【导师】 楼佩煌；

【作者基本信息】 南京航空航天大学 ， 机械电子工程， 2010， 博士

【摘要】 自动导引车AGV(Automated Guided Vehicle)作为一种轮式移动机器人,广泛用于机械、汽车、电子、造纸、烟草、制药和食品等众多行业的自动化物流输送,在国内外市场上具有很大的应用需求,研发具有自主知识产权的高性能AGV具有重要的理论意义和工程应用价值。基于视觉的标线跟踪导航只需识别人工设置的导引标线,可达到很高的导航精度和实时性。两轮差速驱动能实现AGV的零半径转向,表现出良好的机动性。本文针对视觉导航和差速转向的AGV,从理论研究与技术开发相结合的角度努力提高其运动控制性能。在理论研究方面,围绕着如何通过调节速度差控制量消除AGV位姿偏差和如何通过调节电压控制量消除驱动轮速度误差两条技术主线,深入研究基于有限纠偏能力的路径跟踪算法和保证跟踪输出的伺服控制算法。首先比较运动学、动力学和控制受限的运动学三类移动机器人运动控制方法,提出了一种包含路径跟踪和伺服控制的混合运动控制模型,通过速度和加速度约束以及速度差控制量匹配路径跟踪算法的位姿纠偏能力与伺服控制算法的速度纠偏能力。其次研究AGV的路径跟踪问题,分析线性二次型调节器LQR和预测控制在优化目标选取、控制量超限和控制步数设置等方面的难点。对小偏差情况提出了一种基于多步运动预测的LQR最优控制算法,通过纠偏协调性最优的多步运动控制同步消除两种位姿偏差,通过最小化速度和加速度约束下的控制步数保证可实现的最快跟踪。对大偏差情况提出了一种基于视野状态分析的智能预测迭代控制算法,以最优偏差状态转化策略描述控制目标,取代二次型加权和形式的目标函数,通过同步控制算法协调消除理想纠偏状态的两种位姿偏差。再次研究驱动系统的伺服控制问题,将系统模型辨识和PID参数整定描述为多目标优化问题。提出了一类基于精英导向机制的Pareto型多目标遗传算法,通过精英导向、多样性保持和多种群进化机制,快速有效地定向搜索满足决策偏好(路径跟踪需求)的Pareto最优解。在技术开发方面,研究AGV运动控制的多智能体功能建模和嵌入式技术实现,提出了一种基于多智能体结构的嵌入式系统设计方法,建立了控制器智能体向智能体结构和任务的转化模型,为有效实现路径跟踪和伺服控制算法提供了一种高性能嵌入式控制器。本文先通过计算机数值仿真验证所提出理论的可行性,再利用嵌入式技术将理论研究成果转化为高性能AGV车载控制器,并成功应用于自行开发的视觉导航AGV系统(NHAGV)。经过大量系统运行测试与路径跟踪实验,实验结果充分验证了本文所提出的控制技术的有效性和所开发的AGV车载控制器的先进性,这为研发具有自主知识产权的高性能AGV车载控制器奠定了坚实的技术基础。更多还原

【Abstract】 As a kind of wheeled mobile robot, Automated Guided Vehicle (AGV) has been used widely in manufacturing, automotive, electronics, paper, tobacco, pharmaceutical, food and other industries for automated material transportation, and there is a great demand for AGV product in the domestic and international markets. Developing a high-performance AGV with independent intellectual property rights has a great theoretical significance and engineering value.Vision-based tracking navigation only needs to identify guiding lines preset manually, which can achieve a very high accuracy and real-time. Two-wheel differential driving can make AGV turnaround at a zero radius, which shows a good maneuverability. This paper attempts to improve the motion control performance for an AGV with vision navigation and differential turnaround in the perspective of integrating theoretical research with technology development.At the level of theoretical research, two main technical guidelines on how to correct pose errors of AGV by adjusting speed difference output and how to eliminate speed error of driving wheel by regulating voltage output are followed. This paper has made an intensive research on path tracking algorithm with the limited control capability and servo control algorithm able to implement tracking output. Firstly, three kinds of motion control approaches for mobile robots, including kinematics, dynamics, and kinematics with control constraints are compared. A hybrid motion control model containing path tracking and servo control is presented. Speed and acceleration constraints and speed difference output are used to match the capability of path tracking algorithm to correct pose errors with that of servo control algorithm to eliminate speed error.Secondly, path tracking of AGV is investigated. The difficulties in optimized objective selection, output beyond constraints, and control step setting are analyzed for linear quadratic regulator (LQR) and predictive control. A LQR optimal control algorithm based on multi-step motion prediction is proposed for the small error state, which uses multi-step motion control with the best coordination to reduce two pose errors synchronously, and minimizes control steps under speed and acceleration constraints to keep an achievable fastest tracking. An intelligent predictive iterative control algorithm based on state analysis of visual field is proposed for the large error state, which replaces a quadratic weighted sum function with an optimal conversion strategy of error states as a description of control objectives, and uses a synchronous control algorithm to coordinate a correction process of two pose errors for the ideal rectification state.Thirdly, servo control of driving system is discussed, in which system model identification and PID parameter tuning are both expressed as the multi-objective optimization. A kind of Pareto-type multi-objective genetic algorithm based on elitist guidance mechanism is proposed, which makes a fast, effective and directional search for Pareto optimal solutions by using these mechanisms of elitist guidance, diversity preservation and multi-population evolution, in order to meet decision-making preferences that are required by path tracking.At the level of technology development, functional modeling by using a multi-agent system view and implementation by using embedded technology is presented for motion control of AGV. An embedded system design method based on a multi-agent structure is presented, and a transformation model from a controller agent to an agent structure and task is constructed, which can provide a high -performance embedded controller for path tracking and servo control algorithms effectively.Computer numerical simulation is used to verify the feasibility of the theories proposed in this paper, and then theoretical research results are converted into a high-performance AGV vehicular controller based on embedded technology, which has been applied successfully to an AGV (named NHAGV) with vision navigation we develop independently. Many system operation tests and path tracking experiments are carried out. Experiment results have sufficiently verified the effectiveness of the control techniques and the advantages of AGV vehicular controller presented in this paper, which provides a solid technical foundation for developing a high-performance AGV vehicular controller with independent intellectual property rights.更多还原