Iterative Learning and Repetitive Control --- Algorithms, Applications, Points of Commonality and Future Research

Informació ampliada

Course: Iterative Learning and Repetitive Control --- Algorithms, Applications, Points of Commonality and Future Research
Professor: Professor Eric Rogers, University of Southampton, UK
Place: IOC Seminar. (Av. Diagonal 647, 11th floor) [map]
Date: From march 26 to march 30. Every day from 11:00 to 13:00
Inscription: Free.
Contact Ramon Costa (ramon.costa@upc.edu)

An often-encountered industrial control application involves a system or machine that repeatedly performs the same task, at the end of which resetting to the starting location occurs prior to the task being repeated. An example is a gantry robot undertaking a pick and place operation where the following steps must be undertaken in synchronization with a conveyor system: collect an object from a fixed location, transfer it over a fixed, finite duration, place it on a moving conveyor, return to the original location for the next object, and then repeat the previous four actions for as many objects as required.

A controller with no learning applied will give the same tracking error on each trial and although outputs, inputs, and error signals from previous trials are available and rich in information, they are not used by a controller with no learning capability. The objective of iterative learning control (ILC) is to improve the performance from trial-to-trial by using previous trial information in the construction of the current trial input. ILC differs from other learning-type control paradigms, such as adaptive control, in modifying the control input rather than the controller. Controllers designed in the ILC setting can achieve a high performance with low transient tracking error even in the presence of large model uncertainty and repeating disturbances. Repetitive control is designed to track/reject arbitrary periodic signals of a fixed period, where tracking/disturbance rejection of periodic signals is required in many applications. At a general level, iterative learning controls are closely related but there are differences.

Since their introduction in the research literature in the 1980s, these areas have seen very significant developments with, unlike some other ideas, much experimental benchmarking and actual implementations, especially for linear model based designs, but there are still open research questions.

The goals of this course are to give a treatment of the main developments in algorithm design supported by experimental results in most cases. The latter will include very recent work where iterative learning control initially developed and validated for the industrial robotic sector has been applied to robotic-assisted upper limb stroke rehabilitation. This is an example of next generation healthcare, where this general area is now recognized in many countries as one that requires significant effort to cope with ever increasing demand.

The following is a basic outline of the structure, where the method of delivery will be a treatment of the algorithms, simulation support and in most cases experimental verification results.

An overview of open research topics will conclude the course. Supporting references to form a reading list will also be supplied.

1. Origins of iterative learning and repetitive control
2. Simple structure algorithms, performance and limitations
3. Linear model based algorithms and robustness
4. Nonlinear algorithms
5. Noise and other issues
6. Applications outside engineering
7. Open research problems