| Command generation is a process by which input commands are constructed or modified such that the response of a dynamic system satisfies a set of desired output characteristics. In the past, a variety of command generation techniques, such as input shaping and inverse dynamics approaches, have been utilized to reduce vibration, improve trajectory tracking, and improve the performance of systems with respect to a variety of measures. This thesis addresses the specific problem of generating vibration-reducing commands that are robust to modeling errors for nonlinear systems. The primary type of nonlinear systems of interest in this investigation are parallel robotic manipulators. Parallel manipulators are highly nonlinear and are thus a good testing ground for advanced command generation techniques.; The approach taken in this thesis is to address five key issues in a progression towards the ultimate goal of generating robust commands for nonlinear systems. These issues are: (1) deriving the equations of motion, (2) analyzing the resulting dynamic equations of the system, (3) generating performance measures from which to design vibration-reducing commands, (4) developing techniques for generating non-robust vibration-reducing commands, and finally (5) developing techniques for generating robust commands. Through the process of addressing these issues, a variety of command generation techniques are developed for nonlinear systems that balance a variety of competing interests, such as rise-time vs. robustness, and effectiveness vs. computational expense. Ultimately, these robust and nonrobust command generation techniques are shown to be quite effective, and consequently, significantly advance the current state of the art in command generation. |
