机器人视觉伺服是指利用视觉信息作为反馈，使得机器人系统对于伺服目标进行运动估计以及目标跟踪的过程。它综合了机器人学、计算机视觉、数字信号处理、数值分析、并行计算等多个学科和研究领域，被广泛认为是机器人学最重要的研究课题之一。本文在系统总结了机器人视觉伺服相关领域的研究基础之上，重点研究免标定模型无关的机器人视觉伺服问题，本文的主要工作包括：（1）比较深入和完整的总结了免标定模型无关机器人视觉伺服问题的研究进展和技术路线图，提出了免标定模型无关的机器人视觉伺服科学问题的完整数学定义，在此基础上提出了一个统一的基于状态空间模型的免标定模型无关视觉伺服问题研究框架，包括了一个比较完整的视觉伺服算法评价体系，以及多种针对算法特定性能指标的快速复杂目标机动轨迹，能够比较全面和系统的分析、比较和研究免标定模型无关视觉伺服算法的相关性能。（2）提出了基于状态空间模型的免标定模型无关视觉伺服算法鲁棒性定理，给出了视觉伺服算法渐近收敛的基本假设条件。（3）在此基础上，提出了新的免标定模型无关的基于状态空间模型的Broyden群模型更新视觉伺服分割算法（BP-SS-P），这是在本文的视觉伺服算法比较研究中，对于快速复杂机动目标轨迹最有效的跟踪算法，并且也给出了相应的鲁棒性能分析。（4）在统一的基于状态空间模型的免标定模型无关视觉伺服研究框架的基础上，对本领域内最具代表性的方法进行详细和具有说服力的比较研究，构建了基于Matlab机器人学工具箱的视觉伺服仿真实验平台以及基于6自由度Motoman UP6工业机器人的视觉伺服实验平台，不仅验证了免标定模型无关视觉伺服算法在实际工业应用中的有效性，还为后续的研究和改进相关算法打下了良好的基础。

Visual servoing is defined as the process of target motion estimation and tracking on a robotics system, utilizing visual clues as feedback. It's a cross-disciplinary research realm that combines Robotics, Computer Vision, Digital Signal Processing, Numerical Analysis and Parallel Computing and is widely recognized as one of the most promising topics in Robotics research. In this paper, uncalibrated model free visual servoing is discussed and investigated systematically. The main work of this dissertation includes the following points:(1) Based on a systematic literature review and a summary of technical road map on this topic, a complete mathematical formulation of uncalibrated model free visual servoing is defined, based on which a universal state space based framework for uncalibrated model free visual servoing is proposed, incorporating a decent performance evaluation system, different target maneuver trajectories focusing on evaluating performances in specific conditions, which enables a comprehensive and systematic analysis and comparative study.(2) A robustness theory of stated space based uncalibrated model free visual servoing algorithms is proposed to assert a series of assumptions to ensure asymptotic stability.(3) Based on that, a new state space based Broyden population based visual servoing algorithm with partition (BP-SS-P) is proposed, which is substantiated in comparative study as one of the most effective and promising selection to visually tracking target with fast and complicated maneuver, along with a robustness analysis correspondingly.(4) Then a comprehensive comparative study on uncalibrated model free visual servoing is explored, where both of a Matlab Robotics Toolbox based simulation platform as well as a six DOFs Motoman UP6 industrial robot based prototype platform are constructed, to substantiate the theoretical performances of visual servoing algorithm series via a simulation platform and verify the very applicability of uncalibrated model free visual servoing algorithms deployed in real industrial environment.