The Identification and Decision Making Research Group (IDM) is a research group at the University of West Bohemia, founded by Prof. Miroslav Šimandl. The research team foucuses on a basis and applied research in fields that are presented below.

Nonlinear filtering

We deal with the problems of estimating the unknown and immeasurable variables of nonlinear stochastic dynamic systems. The estimation appears in the problems of navigation and tracking of moving objects, econometrics, ecology, control of industrial variables, or meteorology. We solve the problems by using the filters that provide either point estimates, such as the classic extended Kalman filter, modern filters such as unscented Kalman filter, and stochastic integrated filter, or by filters that provide the estimates of the probability density functions, such as particle filter, sum filter, or the point-mass method.

Fault detection

Our attention is also focused on change and fault detection in dynamic systems that has recently received an increased attention, mainly due to increasing safety standards and needs for cost reduction. A large number of applications can be found in the aerospace, chemical and energy industry, or for example in economics. Passive fault detectors based on for example process model based methods, signal processing methods, or expert knowledge, represent a commonly used tool for fault detection in a system. The other way is to use active fault detectors that interacts with the system and through an optimally designed input signal can increase a quality of fault detection. For many years, our group has been developing a probabilistic method of optimal active fault detector and controller design provided a general design criteria.

Information fusion

Besides classically formulated problems of estimation of unknown and immeasurable variables of nonlinear stochastic dynamic systems with centrally available data, we also focus on the estimation problems with decentrally available data. The reason for the limited availability of the data is either impossibility of their transmission to a central point or too high price for their transmission. Such problems appear in a number of areas tied with distributed sensors such as aviation, logistics, defense and intelligence, intelligent transport, navigation, and tracking. We focus especially on a solution to theoretical problems of information fusion such as estimate consistency or fusion optimality.

Adaptive control

Last, but not least we focus on a design of advanced control algorithms. The overwhelming majority of real processes is connected with partial knowledge of system properties and has stochastic characteristics. One of the possibilities to achieve high control quality of such processes is to use adaptive control systems. Automatic initial controller tuning, ability to follow changes in the controlled system and automatic improvement of the process control by a suitable change (or adaptation) of controller parameters or structure are the main advantages of the adaptive approach. Our group intensively focuses on several progressive directions (largely from the theoretical point of view] of adaptive control, especially on a design of new control algorithms based on the dual and functional approach and also on the design of complex mathematical models of nonlinear stochastic systems based upon neural networks or Gaussian processes. It is evident that such areas of interest, especially nonlinear filtering and fault detection, are important building blocks of the proposed control systems.