Estimation technique for tracking rapid parameter changes
M. J. CHEN, John P. Norton · International Journal of Control · 1987
This paper describes a relatively simple technique that enables a recursive least-squares algorithm to track both abrupt and smooth parameter changes without being unduly affected by varying noise levels. The technique differs from methods based on a scalar forgetting factor by its use of vectors to detect parameter variation and adjust the parameter-updating gain. It also incorporates a test to determine when parameter updating should be suspended so as to avoid divergence when little new information about the parameters is arriving.