Abstract:Quasi-Bayes algorithm, combined with stabilized forgetting, provides a tool for efficient recursive estimation of dynamic probabilistic mixture models. They can be interpreted either as models of closedloop with switching modes and controllers or as a universal approximation of a wide class of non-linear control loops.Fully probabilistic control design extended to mixture models makes basis of a powerful class of adaptive controllers based on the receding-horizon certainty equivalence strategy.Paper summarizes… Show more
“…Compared with the results presented in (Kárný et al, 2003) this article has three distinct features. Firstly, based on using the well developed probabilistic based MDN methods (Herzallah, 2012), the involved pdfs in the FPD method are estimated such that their parameters are dependent on the input values in the way reflecting the uncertainty of the network dynamics.…”
mentioning
confidence: 99%
“…Depending on the problem being handled, a number of modeling techniques are employed to represent the local models including standard neural networks (Park and Sandberg, 1991), statistical mixture models (Titterington et al, 1985;Smídl et al, 2005), and regression type models (Wang et al, 2013) among others. The multiple model approach has been recently exploited in the FPD method for deriving a randomised controller for systems that operate in different operation modes (Kárný et al, 2003). The method proposed in (Kárný et al, 2003), however, is constrained by its high computational complexity.…”
mentioning
confidence: 99%
“…The multiple model approach has been recently exploited in the FPD method for deriving a randomised controller for systems that operate in different operation modes (Kárný et al, 2003). The method proposed in (Kárný et al, 2003), however, is constrained by its high computational complexity. In principle, the optimal control law can be obtained by solving a recurrence equation analogical to the dynamic programming solution.…”
mentioning
confidence: 99%
“…The implementation and exploitation of the available analytical results of FPD become even more arduous for the majority of practical systems, which are non-linear and non-Gaussian in nature. Most important, the parameters of the probabilistic models involved in the FPD are assumed in (Kárný et al, 2003) to be input and state independent, limiting the resulting control strategies to deterministic certainty-equivalent systems.…”
Robust controllers for nonlinear stochastic systems with functional uncertainties can be consistently designed using probabilistic control methods. In this paper a generalised probabilistic controller design for the minimisation of the Kullback-Leibler divergence between the actual joint probability density function (pdf) of the closed loop control system, and an ideal joint pdf is presented emphasising how the uncertainty can be systematically incorporated in the absence of reliable systems models. To achieve this objective all probabilistic models of the system are estimated from process data using mixture density networks (MDNs) where all the parameters of the estimated pdfs are taken to be state and control input dependent. Based on this dependency of the density parameters on the input values, explicit formulations to the construction of optimal generalised probabilistic controllers are obtained through the techniques of dynamic programming and adaptive critic methods. Using the proposed generalised probabilistic controller, the conditional joint pdfs can be made to follow the ideal ones. A simulation example is used to demonstrate the implementation of the algorithm and encouraging results are obtained.
“…Compared with the results presented in (Kárný et al, 2003) this article has three distinct features. Firstly, based on using the well developed probabilistic based MDN methods (Herzallah, 2012), the involved pdfs in the FPD method are estimated such that their parameters are dependent on the input values in the way reflecting the uncertainty of the network dynamics.…”
mentioning
confidence: 99%
“…Depending on the problem being handled, a number of modeling techniques are employed to represent the local models including standard neural networks (Park and Sandberg, 1991), statistical mixture models (Titterington et al, 1985;Smídl et al, 2005), and regression type models (Wang et al, 2013) among others. The multiple model approach has been recently exploited in the FPD method for deriving a randomised controller for systems that operate in different operation modes (Kárný et al, 2003). The method proposed in (Kárný et al, 2003), however, is constrained by its high computational complexity.…”
mentioning
confidence: 99%
“…The multiple model approach has been recently exploited in the FPD method for deriving a randomised controller for systems that operate in different operation modes (Kárný et al, 2003). The method proposed in (Kárný et al, 2003), however, is constrained by its high computational complexity. In principle, the optimal control law can be obtained by solving a recurrence equation analogical to the dynamic programming solution.…”
mentioning
confidence: 99%
“…The implementation and exploitation of the available analytical results of FPD become even more arduous for the majority of practical systems, which are non-linear and non-Gaussian in nature. Most important, the parameters of the probabilistic models involved in the FPD are assumed in (Kárný et al, 2003) to be input and state independent, limiting the resulting control strategies to deterministic certainty-equivalent systems.…”
Robust controllers for nonlinear stochastic systems with functional uncertainties can be consistently designed using probabilistic control methods. In this paper a generalised probabilistic controller design for the minimisation of the Kullback-Leibler divergence between the actual joint probability density function (pdf) of the closed loop control system, and an ideal joint pdf is presented emphasising how the uncertainty can be systematically incorporated in the absence of reliable systems models. To achieve this objective all probabilistic models of the system are estimated from process data using mixture density networks (MDNs) where all the parameters of the estimated pdfs are taken to be state and control input dependent. Based on this dependency of the density parameters on the input values, explicit formulations to the construction of optimal generalised probabilistic controllers are obtained through the techniques of dynamic programming and adaptive critic methods. Using the proposed generalised probabilistic controller, the conditional joint pdfs can be made to follow the ideal ones. A simulation example is used to demonstrate the implementation of the algorithm and encouraging results are obtained.
“…A general characterization of SF is given by [6]. SF is discussed in the context of control by [7], and SF is justified in the context of recursive estimation by [5], [8]. In general terms, recursive Bayesian estimation distinguishes two stages: a time step that predicts a new state belief given previous measurements, and a data step that updates the predicted state belief with information from a new measurement, see e.g.…”
Abstract-State-space modeling of non-stationary natural signals is a notoriously difficult task. As a result of context switches, the memory depth of the model should ideally be adapted online. Stabilized linear forgetting (SLF) has been proposed as an elegant method for state-space tracking in context-switching environments. In practice, SLF leads to state and parameter estimation tasks for which no analytical solutions exist. In the literature, a few approximate solutions have been derived, making use of specific model simplifications. This paper proposes an alternative approach, in which SLF is described as an inference task on a generative probabilistic model. SLF is then executed by a variational message passing algorithm on a factor graph representation of the generative model. This approach enjoys a number of advantages relative to previous work. First, variational message passing (VMP) is an automatable procedure that adapts appropriately under changing model assumptions. This eases the search process for the best model. Secondly, VMP easily extends to estimate model parameters. Thirdly, the modular make-up of the factor graph framework allows SLF to be used as a click-on feature in a large variety of complex models. The functionality of the proposed method is verified by simulating an SLF state-space model in a context-switching data environment.
Real-world, multidimensional, dynamic, non-linear processes typically exhibit many distinct modes of operation. Mixtures of dynamic models improve greatly on traditional one-component linear models in this context. Improved prediction then points the way to effective adaptive control design. This paper presents the experience gained under the EU Project, ProDaCTool, in designing and implementing advisory systems, based on dynamic mixtures, in diverse domains: urban traffic regulation, therapy recommendations in nuclear medicine, and operator support for metal-strip rolling mills. Efficient, recursive estimation of the dynamic mixtures from archive data is accomplished using the quasi-Bayes (QB) algorithm, implemented with dedicated software developed within ProDaCTool. The advisory systems are designed using the probabilistic control design technique presented in the previous paper. Highly encouraging prediction and performance enhancements are reported for the applications considered.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.