Vision-based characterization, manipulation, and control of objects using compliant tools
Non-intrusive measurement of forces and mechanical manipulation of objects of unknown characteristics are difficult in general and especially in micro- and nano-systems. In this dissertation, a vision-based force sensing method and how it can be used for mechanical characterization, control, and manipulation of objects will be described. The premise for the method is that image processing techniques enable accurate extraction of displacements from the undeformed and deformed images of an elastic object of known geometry and material properties. By using displacements, it is straightforward to compute the deformation gradients, logarithmic strains, Cauchy's stresses, and finally the applied forces. This direct technique involves very little computation that can be done in real time, and does not assume the locations or orientations of the applied forces a priori. However, a drawback of this method is that it requires extremely high accuracy in the image capture to have reasonable error in force computation. Through analytical sensitivity analysis and numerical experiments, the error is estimated for a given image resolution. In order to make the vision-based force-sensing viable with current hardware, an alternate technique is developed wherein the locations of the forces are assumed to be known. In this method, by measuring the displacements of a finite number of points, forces are computed using suitably modified nonlinear finite element method. When the known elastic object is a compliant mechanism that is capable of sophisticated motion and force transmission tasks, it can be used to characterize, manipulate, and control any unknown object that is rigid or flexible. The visual feedback control of compliant grippers is also explored. The idea is to apply a desired force on the object with the compliant gripper using the visual feedback. The real-time visual feedback control with the compliant device is achieved. This method also can be used to calibrate the force-displacement characteristics of the unknown object. The experiments are performed both in macro and micro level. Furthermore, the compliant device is controlled with multiple actuators to increase the controllability and precision. ^
"Vision-based characterization, manipulation, and control of objects using compliant tools"
(January 1, 2002).
Dissertations available from ProQuest.