【作者】 兰天；

【导师】 刘宏；

【作者基本信息】 哈尔滨工业大学 ， 机械电子工程， 2010， 博士

【摘要】 具有多种感知功能的仿人机器人灵巧手是机器人领域中一个重要的研究方向。本文结合国家科技部(863计划)项目“新一代五指仿人灵巧手及其协调控制的研究”(课题编号:2006AA04Z255),在哈工大机器人研究所与德宇航中心联合研制的HIT/DLR I灵巧手基础上进行改进,目标是实现与人手相近的外形和尺寸(HIT/DLR I灵巧手为1.5倍人手大小,且仅有四个手指),研制达到国际领先水平的驱动内置型五指仿人灵巧手,并重点进行单手指位置控制和柔顺控制的研究以及基于该灵巧手进行多指协调操作的研究。为了能够代替人手完成复杂的任务,要求灵巧手具有与人手相似的大小和活动能力,这由执行机构、驱动控制系统和传感器系统的机械机构所决定。此外,灵巧手还须具有与人手类似的操作能力,如执行精细抓取操作时要求手指具有足够的位置精度,对易碎物体的抓取时要有足够的力控制精度,同时由于环境的不确定性,要求灵巧手具有一定的抗干扰能力,以保证对物体的稳定抓取,这些均对灵巧手控制系统的性能提出了严格的要求。为了提高灵巧手的集成度并增强控制性能,本文研制了基于DSP/FPGA控制结构的模块化嵌入式手指控制系统,成功地将手指驱动系统集成在手指内部,使得与人手尺寸和活动性相似的HIT/DLR II五指机器人灵巧手的得以实现。为实现电机相电流检测,采用一个电流传感器测量并重构了三相电流,并通过设定采样时间获得了稳定的电流信号。为解决手指DSP和FPGA之间如何稳定、高效通信的难题,结合先进先出寄存器,设计了基于串行通信接口和RS485总线的多中断差分通信系统。为了提高灵巧手的智能化水平,HIT/DLR II灵巧手集成了多个微型传感器,所有传感器的模拟输出就近转化为数字信号,以减小信号长距离传输引起的噪声干扰。针对传统巨磁阻传感器在小于1.5 mm工作距离可能引起信号失真或传感器损坏的问题,基于三维静态磁场分析技术分析了不同类型的永磁铁磁场分布,为高度集成的HIT/DLR II五指机器人灵巧手开发了工作距离为0.5 mm的微型巨磁阻传感器系统。实践结果表明经非线性误差补偿后巨磁阻传感器角度测量精度可达±1o。此外,所开发的巨磁阻传感器系统与传统的传感器系统相比,还具有结构简单、成本低、集成度高、精度高的特点,为其它微机电系统和高集成系统的微型磁传感器研究建立了良好的基础。针对HIT/DLR II五指机器人灵巧手手指结构,本文建立了驱动空间中基关节交叉耦合同步误差和关节空间中主从同步误差表达式,提出了包含同步误差和位置误差反馈项及平滑鲁棒非线性反馈补偿项的同步位置控制策略,并理论证明了该策略能够使同步误差和位置误差均收敛,且具有渐近稳定性。通过与非同步控制的PD加摩擦力补偿算法和轨迹跟踪控制算法进行对比,表明文中所设计的控制器有效地提高了指尖在自由空间中的轨迹跟踪精度。通过将关节力矩传感器的反馈信号转化为位置修正向量,本文建立了驱动空间中基关节交叉耦合同步误差和关节空间中主从同步误差表达式,提出了包含同步误差和位置误差反馈项及平滑鲁棒非线性反馈补偿项的同步阻抗控制策略。基于李雅普诺夫稳定性理论证明了所提出的控制策略能够使同步误差和位置误差均收敛,并且保证了系统的渐近稳定性。经与非同步控制的PID阻抗控制算法进行对比,验证了所提出控制策略的有效性,同时与人手的相互作用实验,可以看出设计的同步阻抗控制器能够根据感应到的力矩产生柔顺的光滑运动,表明该系统获得了理想的阻抗性能,为多指手的抓握和与人协调操作奠定了基础。针对灵巧手在不确定环境工作中有可能碰到障碍或受到外界干扰,从而发生被抓物体脱离或损坏的情况,建立了笛卡尔空间多指同步误差表达式,提出了笛卡尔空间多指同步阻抗控制算法,将受干扰手指与其它未受干扰手指的运动联系起来,保持了相对的抓取平衡状态。基于李雅普诺夫稳定性理论证明了所提出的控制策略能够保证系统的渐近稳定性。实验表明该控制策略避免了被抓取物体的脱离或损坏,从而大大提高抓取的抗干扰能力。更多还原

【Abstract】 Dexterous robot hand which has the function of multiform perception is main aspects of robot research. Based on national high technique program (863)“Research on new genetation five-fingerd anthropopathic dexterous robot hand and its cooperative control”, it is the improvement of HIT/DLR I robot hand, and the aim is to accomplish the five-finger HIT/DLR II dexterous robot hand that resembles the human hand regarding size and movability (HIT/DLR I robot hand about 1.5 the size of human hand with only four fingers). The dissertation emphasizes on position control and compliance control of a finger and coordinated grasping manipulative strategy of multiple fingers.To replace human hand, the robot has to fulfill the following requisites: first the robot hand must resemble the human hand regarding size and movability. This determines the hardware, i.e. the mechanical structure, actuator and sensor system. Second the robot hand must have skills and a sleight of hand comparable to the human hand. For example, the hand must be both have enough position accuracy in fine grasping manipulation, and enough force accuracy when grasp fragile object. Moreover, it requires the ability of anti-interference in uncertain environment. This determines the software, i.e. the desired performance of the control system.In order to develop five-finger dexterous robot hand with high integration, a DSP&FPGA-based multilevel control system is developed for multi-sensory and multi-DOF dexterous robot hand, in which the drive, control and sensor systems are integrated in the finger. So the five-finger HIT/DLR II dexterous robot hand that resembles the human hand regarding size and movability is realized. To accomplish the phase current detection of the motor, a single direct current link current sensor is used to measure and reconstruct the three phase currents, and stable current signal is measured by optimizing sample instant. To solve the problem of efficient and reliable communication between finger DSP and FPGA, a kind of multi-interrupt differential communication system based on serial communications interface (SCI) and RS485 bus using first in first out (FIFO) registers is designed. To enhance the levels of intelligence manipulation, HIT/DLR II robot hand is equipped with multiple sensors, and all the analog signals are converted in-situ into digital signals to minish noise interference.To solve the problem that the working distance of traditional giant magnetoresistance (GMR) sensor must be more than 1.5 mm, or else it maycause severe disorientation of the angular measurement or permanently damage the sensor element. Based on 3D static magnetic analysis technique, a method for the modeling of ferromagnetic component of the sensor system is presented, and the ultra-miniature giant magnetoresistance sensor system with only 0.5 mm working distance is newly developed for highly integrated HIT/DLR II five-finger dexterous robot hand. The experimental results show an angular accuracy of less than±1°with only the residual offset compensation of ultra-miniature GMR sensor system is obtained. It also has the qualities of simple structure, low cost, ultra-miniature size, and high accuracy, all of which makes up for the shortcomings of traditional measurement sensors.According to the finger structure, a cross-couple synchronized error expression in drive space and master-slave synchronized error expression in joint space are built for HIT/DLR II robot finger. Then a synchronized position control approach is present, including feedbacks of synchronous errors and position errors, and a smooth robust nonlinear feedback compensator. According to Lyapunov stability analysis, it is proved that the proposed method can guarantee both synchronization and position errors converge and asymptotical stability of the system. Compared with conventional non-synchronized PD friction compensation and trajectory tracking control, experimental results demonstrate the proposed control strategy improve the trajectory tracking precision of finger tip in free space.By translate the information of joint torque sensor into position vector, and a cross-couple synchronized error expression in drive space and master-slave synchronized error expression in joint space are built for HIT/DLR II robot finger. Then a synchronized impedance control approach is present, including feedbacks of synchronous errors and position errors, and a smooth robust nonlinear feedback compensator. According to Lyapunov stability analysis, it is proved that the proposed method can guarantee both synchronization and position errors converge and asymptotical stability of the system. Compared with conventional non-synchronized PID impedance control, experimental results demonstrate the validity of the proposed control strategy. And in interaction of robot finger with a person, it generates compliance and smooth movement according to the joint torque, which demonstrates that the ideal joint impedance performance is successfully achieved. All this will provide a good base of multi-finger grasping manipulation and human-robot coordinating manipulation.When the dexterous hand works in uncertain environment, the finger may encounter obstacle or disturbance, which may cause release or damage the grasped object. To solve this problem, the synchronized error expression between multiple fingers is built in Cartesian space, and then a synchronized impedance control approach in Cartesian space is present. The main idea is cooperates the undisturbed fingers with the disturbed finger, which keep relative balance of grasping manipulation. According to Lyapunov stability analysis, it is proved that the proposed method can guarantee asymptotical stability of the system. The experimental results demonstrate the proposed control strategy can avoid damaging or releasing the grasping object, and enhance the anti-interference ability of the dexterous robot hand.更多还原