An Iterative Learning Control Algorithm Based On Prescribed Input-Output Subspace

Iterative learning control (ILC for short) theory is an important branch of intelligence control field. It's a technique for improving the transient response performance of systems or processes that operate repetitively over a fixed time interval. It refines the next control input using the information such as current control input and error signals after each trial until the specified desired trajectory is followed to a high precision. It is one kind of off-line learning algorithm, needs very few information about the object to control, therefore, it has very good control effects with some complex objects that difficulty modeling. This control method suits extremely in the objects that always run repetitively, such as manipulator, numerical control engine bed, magnetic head driver, electrical machinery servo system and so on.Based on the reading of massive literatures, the paper firstly presents a review on the ILC. Then introduce a new theory of ILC based on prescribed input-output subspace, it requires neither derivatives of the error signals nor dual mapping. Based on this theory, the paper presents a new ILC algorithm, since we take the space transformation during the course of prescribing the input-output subspace and take account of the error signal of each trial, it has really reduced the affection of the measurement white noise, so the proposed ILC has the good ability of noise tolerant. The ILC proposed in this paper can be also used in the identification for a class of continuous-time systems. The paper also gives a case of the electro-hydraulic position servo control system to prove the validity of the proposed ILC.The simulation result shows that the proposed ILC can track the reference signal according to the reference signal and the input-output data; the ILC has a strong ability of noise tolerant, it can get a satisfied control result even with a big measurement noise; the ILC can accurately identify the coefficients of a class of continuous-time systems when the reference signal is followed precisely.