UC Berkeley

This dissertation focuses on the design of feedback and feedforward controllers for direct application to industrial manipulators. In this dissertation, an iterative online controller tuning algorithm based on nonlinear programming concepts and extremum seeking control is introduced and applied to a 6 degree of freedom FANUC M16$i$B industrial robot. Details regarding stepsize selection and gradient estimation for the proposed controller tuning is also discussed. Experimental results show that the proposed controller tuning method is able to improve robot performance by successively reducing a cost function. Additionally, a controller tuning framework based off the disturbance observer is also introduced for stable controller tuning. The framework is shown to be robustly stable under most practical tuning applications. The assumptions and constraints of the proposed framework is also detailed. The framework is experimentally verified by sweeping through a variety of controller gains that satisfy the framework conditions. Finally, an input shaping technique for application to industrial robots with elastic joints is also proposed in this dissertation. The approach is simple to apply and can be easily integrated to existing trajectory generation techniques for industrial robots. Experimental results show substantial joint vibration suppression during transient motions.