(Dr. Nader Sadegh, advisor)
"Auto-Tuning Control Using Optimization Techniques"
The goal of this research is a “plug and play” controller. This controller is placed in-line with any system needing control and set with limits on parameters and performance. It then executes a series of tests that it learns from and auto-tunes the control parameters for optimal performance. The tuning algorithm is driven mainly by the data it receives from the system controlled and is thus a software program with hardware in the loop.
This method attempts to design the controller with minimal knowledge
of the system it is controlling. There is no accurate model that
helps determine the appropriate control parameters. Instead the end
user selects a desired performance parameter (i.e. settling time, following
error) and the controller tunes the control parameters itself to get the
required performance. Intelligent software was written that uses
optimization methods to learn how to produce the desired response.
The parameters are optimized within limits specified by the end user and
within bounds posed by physical constraints, such as non-linear and inequality
type constraints associated with saturation in actuators. Initially
the software was designed in MATLAB with simulations of various plants.
Then the software was tested using equipment provided by CAMotion, Inc.