UCLA

The precision and productivity requirements for modern manufacturing process drive precision systems to their physical limitations, such as the limitation of sensor resolution, actuator saturation, and traversing speed. Incorporating these restrictions, constrained optimization has been an active research area that can explicitly handle the physical constraints. This dissertation covers three important topics in productivity and precision improvements through constrained optimization: sub-count sensor estimation, constraint violation avoidance, and cycle-time

minimization.

To solve sensor quantization and measurement synchronization problem, a

model-based encoder-triggered estimation method was developed addressing state continuity. By enforcing smooth estimation in important physical states such as position and velocity, this method avoids abrupt estimation change when encoder trigger occurs. Furthermore, because continuous-time open-loop model is used to predict inter-trigger behavior, the developed sub-count estimator can be applied to asynchronous systems with untimely measurement updates. By applying this sub-count estimation method, the precision system can improve tracking performance

when only quantized and untimely sensor measurement is available.

When improving tracking performance, the precision system is often pushed to its limitation. This often causes the precision system to violate system constraints (such as control signal saturation and excessive acceleration), which results in unexpected vibration and sacrifices the tracking performance. To avoid such kind of constraint violation, the system model can be used to predict the dynamic behavior of the plant given a known trajectory and plant model. A model predictive control based optimal feed-forward tracking method is proposed to be integrable with any feedback controller.

Besides tracking performance, cycle time is also an important index for modern precision motion control. The shorter the cycle time is, the more productive the machine can be. Therefore, to reduce cycle time for precision machines, a minimal time contour tracking problem is formulated to explicitly constrained receding horizon control where the trajectory (feed) profile is determined for the specified contour under system dynamics, contour error, axial velocity, signal saturation, and monotonic feed constraints. Because minimal-time operations typically drive the system near its constraint boundary, it is vital to reject measurement noise and

compensate for modeling error in real-time. To accommodate the high sampling rate controllers in precision systems, an efficient real-time quadratic programming solver is proposed with robust warm start strategy in order to handle active constraints.