SUMMARY

Control of long-reach space and industrial manipulators are two immediate benefits of on-going research on the topic at hand. A potentially more significant benefit on the other hand, is the possibility to improve the accuracy and precision of any robotic manipulator in which actuators and sensors are non-collocated. Frequently encountered model inaccuracy on its own can severely affect the stability and consistent performance of a closed-loop control system, however the presence of non-collocation exacerbates the situation. The primary goal of this thesis is to closely examine the usefulness of the sliding mode observer, a robust non-linear observer, for application to a flexible robotic arm in the presence of non-collocation and parametric uncertainty, which is a form of model inaccuracy. Based on both simulated and empirical performance of the observer as part of a closed-loop system, this research will judge the value of the robust non-linear observer for application to control of flexible robotic arms. The experimental test-bed consists of a single flexible link actuated by a linear servo-motor and utilizes sensors including a linear encoder and an accelerometer. An intuitive and well-documented user-interface is intended to serve as the groundwork for similar future studies of model uncertainty and closed-loop observer performance. Textual and graphical programming techniques in the NI LabVIEW Real Time environment serve this purpose.