| The anthropomorphous robotic dexterous hands is one of the key systems for hu-manoid robots. The dexterous manipulability of the robot hand depends on its manip-ulative dexterity, grasp robustness, and human operability, which greatly influences theability of humanoid robots for dexterous tasks. Consequently, the control system designand control strategies for the dexterous robot hand becomes one of the hot spots in hu-manoid robots field. Following the theme of the National High Technology Research andDevelopment Program”Research on structure design and grasping manipulation of newfive-fingered dexterous hand”, the compliant control strategies for the dexterous robothand are investigated in this thesis. Furthermore, a control system for the dexterous robothand is designed and implemented in this thesis.This thesis presents a dexterous-task-orientated flexible control platform for dexter-ous the robot hand. The flexible control platform consists of Simulink-QNX real-timecontrol layer, distributed communication modules and a virtual robot model. This controlplatform has distributed hard real-time ability, and is able to seamlessly connect real-timeand non-real-time environment. In order to fully exploit the multisensory, modular designof the HIR/DLR II, this thesis proposes a multi-layer control system for the HIT/DLRII hand, composed of a motor drive layer, embedded controller layer, and real-time con-trol layer. All the layers communicate with each other in real-time to achieve a highlyintegrated and efective hardware control system. The control platform, together withthe multi-layer control system, is modular, flexible and transparent to controller design,which fulfills the requirements of robot hands control, and hand-arm-vision control forhumanoid robots.The performance of actuator-embedded dexterous robotic hand, such as the HIT/DLRII, can be hindered by limited construction space, transmission system complexity, andjoint coupling. In order to address these issues, an impedance control strategy based onjoint torque feedback with adaptive friction compensation is proposed in this thesis. Therobot hand dynamic model is implemented with flexible joints and joint couplings, basedon which the joint and Cartesian impedance control with joint torque feedback are in-vestigated. As tasks for robot hands involve motion with shorter travel and has shorter distances between adjacent joints, the performance of robot hands is influenced by fric-tion and gravity more than robot arms. Furthermore, due to the compact design and jointcoupling, the friction model on the finger joints is difcult to explicitly identify, and jointfriction parameters could vary from finger to finger due to component tolerance. There-fore, an adaptive friction observer based on extended Kalman filter is proposed in thisthesis to adaptively observer and compensate for joint frictions. An adaptive impedancecontroller is presented based on the proposed observer. The stability of the closed-loopsystem with friction compensation is analyzed, and input-to-state stability is proved. Cri-teria for friction observer design on robot hands are concluded in the stability analysis.Experimental results on impedance control and friction compensation demonstrates theefectiveness of the proposed adaptive impedance controller.In order to accomplish dexterous manipulation tasks, multiple fingers on the robothand have to cooperatively work together. A n-fingered (n≥3)6-dimensional spatial-ly coordinated impedance control strategy is proposed based on spatial virtual springs inthis thesis, and a damping term is designed for dynamic behavior of the controller. Themodular dexterous robot hand HIT/DLR II may be configured as n-fingered robot hand.A n-fingered (n≥3) object level frame is proposed. Based on spatial virtual translationsprings and rotation springs, a6-dimensional spatially coordinated impedance controlleris presented. Additional connecting springs between the object level frame and fingertipsof the fingers are proposed to control the internal grasping force and achieve compliantgrasping. The mapping between object level frame and joint space is derived for imple-menting translational, rotational and connecting forces in finger joint space. Based on thismapping, a damping term is designed for improve the dynamic behavior of the spatiallycoordinated impedance controller. The desired positions and orientations are controlledby the spatial impedance controller. As a result, the compliant behavior of the graspedobject and internal grasping forces can be defined and controlled. Furthermore, the inerti-a of the complete hand-object system could be better defined and compensated in spatialimpedance control strategy. On the other hand, the spatial impedance controller providesa simple and intuitive interface for the high-level grasping and manipulation planning,and significantly reduces the complexity of the robot task definition and planning. Basedon energy field of the spatial connecting spring, a self-collision avoidance control strategyis presented for a multi-fingered robot hand, with damping design based on the mappingbetween spatial distance direction and joint space. Experiments are carried out on the HIT/DLR II dexterous robot hand with3,4, and5-fingered configurations. The experi-mental results show the efectiveness of the proposed6-dimensional spatially coordinat-ed impedance controller and self-collision avoidance control strategy for multi-fingeredrobot hand. |
PDF Full Text Request