Efficient Adaptive Inverse Control Algorithms

Adaptive inverse control is an automatic control-system design method which “learns” over time how to control a dynamic system (“plant”). The technique may be applied to single-input single-output (SISO) and multi-input multi-output (MIMO) plants. The plant may be linear or nonlinear. A problem with adaptive inverse control in its current state is that the learning algorithms are very slow. This is the final report for the Grigsby grant titled “Efficient Adaptive Inverse Control Algorithms.” It addresses some results concerning a new way to perform adaptive inverse control of linear MIMO systems which learns very quickly. An example is shown which demonstrates a speedup in the order of six orders of magnitude! The results from this grant have been used in part to write a proposal titled “Adaptive Inverse Control of Linear and Nonlinear Dynamic Systems” and submitted to the National Science Foundation. A technical report has also been written to document some of these results, and a paper is being submitted to the 2001 IFAC conference “Adaptation and Learning in Control and Signal Processing.”