【作者】 郭语；

【导师】 孙志峻；

【作者基本信息】 南京航空航天大学 ， 机械电子工程， 2013， 博士

【摘要】 为了适应日益复杂的柔性作业和远程灵巧操作的需要，具有和人手结构类似的机器人灵巧手的研究，一直是机器人研究领域的热点和难点。同时，由于人工智能理论、计算机控制技术、传感器等支撑技术发展的限制，要想使灵巧手在非结构化环境下实现完全自主的工作还是非常困难的。因此，利用主从控制技术实现灵巧手的遥操作也一直伴随着机器人灵巧手的研究。本文研究超声电机驱动的多指灵巧手及其主从控制系统，其主要研究成果如下：1、在综合分析国内外机器人灵巧手的研究现状和人手的生理结构基础上，采用超声电机作为驱动部件，设计了一种仿人灵巧手，实现了在尺寸、重量和自由度方面都接近人手的目标。2、采用机器人D-H参数建模方法建立了灵巧手的运动学模型。针对其单指具有冗余度的特点，参考人手指运动的耦合关系，提出一种解析法+数值法的快速运动学逆解算法，并进行了仿真验证。考虑到腱传动方式带来的关节运动之间的耦合问题，建立了关节空间与驱动空间之间的耦合方程，为实现灵巧手运动控制打下了基础。3、建立了人手示教方式下的多指抓取数学模型。参考人手抓取经验，提出接触稳定裕度的概念并将其引入到抓取内力规划中，实现了稳定、安全的抓取力规划。为了提高多指手的抓取性能，在关节空间建立了两个抓取性能指标，以物体位姿参数为优化变量，建立了多指手抓取物体的多目标优化抓取规划模型。利用多目标粒子群优化算法获得了规划模型的非劣解集前沿，通过三指手的仿真验证了本文所提方法的有效性。4、研制了两种不同类型的主操作手：穿戴型被动式力反馈主操作手和桌面型主动式力反馈主操作手，重点对前者进行了运动学和动力学的优化设计。基于DSP主控制器搭建了控制系统的硬件部分，并进行相关软件设计。此外，采用PID+Bang-Bang技术实现了灵巧手单指的关节空间与笛卡尔空间的轨迹跟踪控制，基于阻抗控制策略实现了手指的指尖接触力控制，并进行了频率响应实验。为了衡量两种主操作手的操作性能，分别进行了主从运动控制和力反馈的实验。5、时延以及不确定性等因素容易造成双边遥操作系统操作性能差，甚至不稳定等问题。为了解决这一问题，针对主、从端不同的任务特点，提出一种新的双边遥操作系统控制结构。由于主端在实现力反馈时易受扰动影响，设计一种非线性扰动观测器用于扰动的在线估计并进行补偿；从端采用基于滑模的阻抗控制以保证精确的位置跟踪和与环境的期望交互。利用莱威林绝对稳定准则推导出时延下保证系统稳定性的条件以及参数选取依据。最后在建立的单自由度遥操作系统实验平台上进行了实验研究，结果表明所提方法能够在保证系统鲁棒性的条件下有效提高操作性能。更多还原

【Abstract】 In order to meet the increasingly complex requirement of flexible task and dexterous teleoperation,robot multi-fingered dexterous hand, which has similar structure with human, is always an importantresearch aspect of robot. Furthemore, due to the limitations of the development of artificial intelligencetheory, computer-controlled technology, sensors and other supporting technologies, it is too difficult tomake the dexterous hand completely autonomous work in unstructured environment. Therefore, theapplication of master-slave control technology to teleoperate dexterous hand has also been researched.This thesis researches multi-fingered dexterous hand driven by ultrasonic motors (USMs) and itsmaster-slave control system.To conclude, main contributions of this dissertation are as follows.1. On the basis of analyzing existing robot dexterous hands and physiological structure of humanhand, a humanoid multi-fingered dexterous hand driven by ultrasonic motors is designed which is closeto human hand in the aspects of size, weight and degree of freedom.2. The robot D-H parameter method is applied to build kinematics model of the dexterous hand. Byreferring to the joint quadratic coupling relation of human finger, a kind of fast inverse kinematicsalgorithm, which consists of analytical and numerical methods, is proposed to solve the redundancyproblem of single finger, and the simulation is conducted. Considering the joints motion couplingproblem brought of the tendon transmission mode, mapping equations between joint space and drivingspace is established to realize motion control of dexterous finger.3. Multifingered grasping model under the teaching mode is built. The concept of contact safetymargin is proposed according to the experience of human grasping, and the nonlinear optimizationmodel of internal force with contact safety margin is also built by modifying constraint condition offriction cone, and the corresponding contact force is also obtained. In order to improve the graspingperformance, by taking object pose parameters as variables, the multi-objective optimization model ofgrasping is built based on the performance indices including position-grade and relative loadingcapability in joint space of robot hand. Then, a three-fingered robot hand grasping object is calculatedand analyzed, and multi-objective particle swarm optimization algorithm is applied to plan the objectpose while a set of non-dominated solutions is also obtained. The results demonstrate the proposedmethod can efficiently improve grasping performance under the condition of safe grasping.4. Two different types of master hand: wearable passive force feedback master hand and desktopactive force feedback master hand are developed, and the former one is conducted on the kinematics and dynamics optimization design. Then, the control system including hardware and software is establishedbased on DSP host controller. Moreover, the PID+Bang-Bang technology is applied to achievetrajectory tracking control in joint space and Cartesian space, and the fingertip contact force control isimplemented based on impedance control strategy, and frequency response test is also carried out. Inorder to measure the performance of the two developed master hands, experiments about master-slavecontrol and force feedback are respectively conducted.5. The performance and stability of bilateral teleoperation system are easily degraded by time delayand uncertainty factors. In order to solve this problem, a new control structure is proposed concerningdifferent task characteristics of master and slave sides. Because the master-side is vulnerable todisturbance in the realization of force feedback, a kind of nonlinear disturbance observer is designed foronline estimate and compensation. At the slave-side, impedance control based on sliding mode isapplied to ensure precise position tracking and desired interacting with the environment. The Liewellynabsolute stability criterion is adopted to derive stability condition and parameter selection criteria undera time delay. Finally, the proposed method is validated by experiments with a single-DOF teleoperationplatform, and the results show that the proposed method can improve operational performance whilehaving good robustness.更多还原