UC Berkeley

Industrial robots are widely used in a variety of applications in manufacturing. Today, most industrial robots have been pushed to work near their hardware design limits. Therefore, it is essential to develop advanced control techniques to further improve the performance of industrial robots. This dissertation focuses on efficient motion planning and effective trajectory tracking control for flexible robots. The difficulties of this work arise from the facts that 1) due to the complicated nonlinear mapping between robot configuration space and workspace, the constraints applied in one space are difficult to transfer to another space, 2) due to the inherent mechanical flexibility, static and dynamic deflections between actuators and robot end-effector are frequently observed, and they degrade the overall trajectory tracking performance. In regards to these issues, this dissertation proposes several methods to improve motion control performance of industrial robots.

Regarding motion planning, an optimal control based approach is presented. Because of the inherent complexity of the motion planning problem for articulated robots with multiple joints, most existing solutions decompose motion planning as path planning and trajectory planning problems. Because of the implementation of manual or random sampling approaches in path planning, the resulting solution is in general suboptimal. This dissertation proposes to formulate motion planning as a general nonlinear optimal control problem. A practical numerical method is investigated for trajectory optimization as one solution to the underlying optimal control problem. Intelligent discretization and automatic differentiation techniques are introduced to make the proposed approach highly efficient.

Regarding trajectory tracking of robots with compliant components, two kinds of flexibility are considered. One kind of flexibility comes from the compliant transmission elements, i.e., joint flexibility. For robot with joint flexibility, back-stepping control is designed to achieve high performance of trajectory tracking. To address model uncertainties in the system, a radial basis function network is introduced for online adaptive compensation. Lyapunov stability theory is used to prove the stability of the proposed adaptive controller. A data-driven approach for the structural design of a radial basis function network is also presented to effectively reduce the computation load.

Another kind of flexibility comes from the compliant links, which is known as link flexibility. Industrial robots equipped with large articulated structures as end-effectors are good examples of robots with link flexibility. One popular and promising approach for vibration suppression involving flexible links is input shaping. However the time delay introduced by input shaping is not permissible for time stringent applications. In this dissertation two modified input shaping approaches are presented to suppress residual vibration of end-effectors without introducing time delay to the entire motion. Considering the control signal generated from input shaping does not necessarily seem to be the best for robotic application, optimal vibration suppression is also discussed based on the proposed efficient numerical method for trajectory optimization.