Real-time visual feedback control for hand-eye coordinated robotic systems
Abstract
Because image processing and analysis are often complex and time consuming processes, visual feedback has been overlooked as a viable means for real-time control. However, as image processing equipment progresses with such features as video rate convolution and region-of-interest processing, vision will soon become a valuable source of control system feedback. This thesis discusses three coherent main topics of visual feedback control as they pertain to the control of a hand-eye coordinated robotic system. The first topic is a multi-loop resolved motion rate control structure which allows uniform sampling at the robot joint level and non-uniform sampling at the visual feedback level. Errors in image features are resolved into errors in robot joint angles through a series of Jacobian transformations. A unique feature-based trajectory generator provides smooth motion between the estimated features and the desired features while staying within the constraints of the task, image processing time, and robot dynamics. The second topic is the automatic selection of robust image features for control. Three image feature points are used to control the relative position and orientation between a camera on the robot's end-effector and a moving known workpiece. A weighted criteria function with both image recognition and control factors is used to select an optimal set of features. The third topic is the use of an adaptive self-tuning controller to fine tune the transformation from feature space to robot joint space and to predict the position of image features. An auto-regressive model is used to predict the image features' positions based on past observations and control inputs. These estimated feature positions are used to determine the optimal control input and to guide the feature extraction process. The above theory and concepts have been implemented and demonstrated on a PUMA 600-series robot arm with a CCD camera mounted on its end-effector.
Degree
Ph.D.
Advisors
Lee, Purdue University.
Subject Area
Electrical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.