以馬可夫模型對線上及離線系統之識別與應用

• Main Authors: CHAO HUI CHENG, 趙惠誠
• Other Authors: Liu, J.J.
• Format: Others
• Language: zh-TW
• Published: 1998
• Online Access: http://ndltd.ncl.edu.tw/handle/55662664908086965590

碩士 === 中正理工學院 === 兵器工程研究所 === 86 === The purpose of this thesis is to discuss three algorithms of parameter identification and apply them to some practical and simulation cases. Firstly the time-invariant system of Markov parameter model is discussed. It is to model the time-variant system. Then the comparison between recursive least-square (RLS) and fast transversal filter (FTF) identification algorithms are made in real-time system. The recursive estimation methods are implemented to the practical system. The results are shown that using the Markov parameter estimated model as the initial value of RLS is better than the RLS without initial value. In many cases it is necessary, useful, to have a model of the system available on-line while the system is operation. The model should then be based on observations up to the current time. Identification techniques that comply with this requirement will be called recursive identification methods. Since the measured input-output data are processed recursively as they become available . For a time-varying and complex systems the off-line estimation can not cope with the engineering requirement, and the errors will get. Recursive identification algorithms are instrumental for most adaptation schemes, and the on-line estimate for the real-time systems is also the powerful method for the real-time system.