UBC Theses and Dissertations
Development of impedance sensing technology and an intelligent control system for robot-automated processing of flexible and natural objects Gu, Jianhua.
General tasks of robotic manipulation may be broadly categorized as gross/hard manipulation and flexible manipulation. In gross/hard manipulation, robotic tasks are denned by motion commands based on predefined and precise positions. In this case, the high-level task control typically adopts an open-loop approach where low-level servo controllers drive the robot with the objective of achieving a predefined joint trajectory that is consistent with the task trajectory as planned at an upper level, but the sensory information is not fed into the task level of the robot control system. In flexible manipulation, as defined in this thesis, robot motion cannot be pre-defined in terms of precise positions and motions, due to task uncertainties. Then, feedback of sensory information to upper levels of the control system, with appropriate preprocessing and abstraction, will facilitate high-level task control and improved process performance. Control systems of the existing commercial robots by and large fall into the category of gross/hard manipulation, with associated shortcomings. Focus of the present research is on fine/flexible manipulation, which is important in robotic processing of flexible and inhomogeneous, natural materials, such as fish and meat, where, the detection of material properties and transition regions in the processed object is usually important for high-level, intelligent task control. Mechanical impedance at the process interface of a robotic task, is considered to provide the information that is needed in detecting the required process characteristics, and is investigated as a significant sensing approach in flexible manipulation. The research presented in this thesis concerns technology development for flexible manipulation of robotic processes. The main goals of the present research are threefold; (1) investigate and develop an impedance-based task sensing method for flexible manipulation, (2) investigate, design, and develop a control architecture that has the capability of on-line task monitoring and interpretation, high-level feedback of information, and knowledge-based decision making, which is suitable for executing flexible manipulation of robotic processes, and (3) implement and evaluate the developed sensing and control system technologies with regard to practical applications. To achieve these objectives, the following tasks are carried out: Analytical framework for dynamics, sensing, and control. Models are formulated for robot dynamics and process impedance, with linearizing feedback and task separation for gross motions and fine manipulation. This framework is intended for facilitating on-line estimation of process impedance and implementation of knowledge-based task monitoring and control. Impedance sensing technologies. Methods of estimating mechanical impedance at the process interface of a robotic processing task are developed, which use actuator effort and the joint motion signals as the input information to the estimator. Methods of signal conditioning and estimation are developed based on the Kalman filter approach. Further interpretation and utilization of impedance information at the task level is made through knowledge-based decision making. This is intended to accommodate task uncertainties, specifically, position variables and disturbances. Control architecture. The use of mechanical impedance as intended in the present research requires feedback of sensory information to the task control level. To accommodate this and other requirements of flexible manipulation, a real-time open-architecture control system (ROACS) is designed and developed in this thesis. The control system has a hierarchical structure with the capability of multi-mode control. A systematic method is developed for intelligent task control, where tasks may be represented in a descriptive manner with unknown task variables which may be determined and assigned during operation. Design of implementation model and system prototyping. A client-server type robot control system is designed and developed that satisfies the ROACS model. The system contains three servers: servo controller, motion planner, and data analyzer; and three client applications: intelligent task control, information updating module, and a graphics user interface which includes a robot control shell and a real-time task language (RTTL) as developed in the present research. The entire system adopts a cooperative model between the real-time subsystem and the artificial-intelligence subsystem, which can efficiently carry out both low-level sensing and control tasks, and high-level process monitoring and knowledge-based decision making tasks in an integrated manner. The developed robot controller is implemented for operation with a Puma 560 robot in the Industrial Automation Laboratory. A salmon-slicing task, which produces boneless salmon steaks, is performed by using the developed system, to demonstrate the performance of ROACS and the feasibility of employing the developed technology in an industrial application. The developed technology is also implemented in an industrial CNC router machine to accurately sense the material surface of the processed object for system referencing for subsequent processing.
Item Citations and Data