Content: Parameter estimation is very important for identifying system static and dynamic models, because these essential parameters are known to be generally the certainty of the system's probability deterministic properties. The maximum likelihood method is based on probability statistics elements. These two lead to the formation of a maximum likelihood parameter estimation, which is considered to be one of the classical probability Bayesian methods. Maximum likelihood estimation has an attractive limiting nature of consistency, asymptotic normality and efficiency. It is widely used in aircraft dynamic parameter identification, leading to inertial instrument error coefficient estimation and traffic engineering flow monitoring.
So far, several optimization methods have been proposed to solve the maximum likelihood parameter estimation problem. It is roughly divided into three categories, conventional analytical methods, traditional numerical approximation and biological heuristic optimization methods.
Nowadays, the maximum likelihood parameter in a static system is usually estimated by traditional analytical methods. However, most of the problems estimated by the maximum likelihood parameter in dynamic systems are highly non-linear, and it is difficult to be solved by conventional analytical methods. Therefore, researchers tend to seek traditional numerical approximation techniques to overcome these difficulties.
Yongzhong Lu and his team made a study on the perspective of bioincentive optimization technology, result of which was published in the journal Autonomous Intelligence. This review attempts to provide a comprehensive view of traditional and bio-stimulus optimization techniques in maximum likelihood parameter estimation to highlight challenges and key issues and to promote further research. The focus of this paper is to study recently used traditional and bio-stimulus optimization techniques in maximum likelihood estimation.
Studies have shown that we need to further improve the traditional numerical approximation techniques in maximum likelihood parameter estimation, and we need to develop hybrid traditional and biological heuristic techniques to estimate the maximum likelihood parameters based on their merits.