Showing 1–7 of 7 results for author: Kouw, W M

Online Bayesian system identification in multivariate autoregressive models via message passing

Authors: T. N. Nisslbeck, Wouter M. Kouw

Abstract: We propose a recursive Bayesian estimation procedure for multivariate autoregressive models with exogenous inputs based on message passing in a factor graph. Unlike recursive least-squares, our method produces full posterior distributions for both the autoregressive coefficients and noise precision. The uncertainties regarding these estimates propagate into the uncertainties on predictions for fut… ▽ More We propose a recursive Bayesian estimation procedure for multivariate autoregressive models with exogenous inputs based on message passing in a factor graph. Unlike recursive least-squares, our method produces full posterior distributions for both the autoregressive coefficients and noise precision. The uncertainties regarding these estimates propagate into the uncertainties on predictions for future system outputs, and support online model evidence calculations. We demonstrate convergence empirically on a synthetic autoregressive system and competitive performance on a double mass-spring-damper system. △ Less

Submitted 3 June, 2025; originally announced June 2025.

Comments: 6 pages, 1 figure, conference: ECC2025

Coupled autoregressive active inference agents for control of multi-joint dynamical systems

Authors: Tim N. Nisslbeck, Wouter M. Kouw

Abstract: We propose an active inference agent to identify and control a mechanical system with multiple bodies connected by joints. This agent is constructed from multiple scalar autoregressive model-based agents, coupled together by virtue of sharing memories. Each subagent infers parameters through Bayesian filtering and controls by minimizing expected free energy over a finite time horizon. We demonstra… ▽ More We propose an active inference agent to identify and control a mechanical system with multiple bodies connected by joints. This agent is constructed from multiple scalar autoregressive model-based agents, coupled together by virtue of sharing memories. Each subagent infers parameters through Bayesian filtering and controls by minimizing expected free energy over a finite time horizon. We demonstrate that a coupled agent of this kind is able to learn the dynamics of a double mass-spring-damper system, and drive it to a desired position through a balance of explorative and exploitative actions. It outperforms the uncoupled subagents in terms of surprise and goal alignment. △ Less

Submitted 14 October, 2024; originally announced October 2024.

Comments: 14 pages, 3 figures, accepted to the International Workshop on Active Inference 2024

Planning to avoid ambiguous states through Gaussian approximations to non-linear sensors in active inference agents

Authors: Wouter M. Kouw

Abstract: In nature, active inference agents must learn how observations of the world represent the state of the agent. In engineering, the physics behind sensors is often known reasonably accurately and measurement functions can be incorporated into generative models. When a measurement function is non-linear, the transformed variable is typically approximated with a Gaussian distribution to ensure tractab… ▽ More In nature, active inference agents must learn how observations of the world represent the state of the agent. In engineering, the physics behind sensors is often known reasonably accurately and measurement functions can be incorporated into generative models. When a measurement function is non-linear, the transformed variable is typically approximated with a Gaussian distribution to ensure tractable inference. We show that Gaussian approximations that are sensitive to the curvature of the measurement function, such as a second-order Taylor approximation, produce a state-dependent ambiguity term. This induces a preference over states, based on how accurately the state can be inferred from the observation. We demonstrate this preference with a robot navigation experiment where agents plan trajectories. △ Less

Submitted 18 September, 2024; v1 submitted 3 September, 2024; originally announced September 2024.

Comments: 13 pages, 3 figures. Accepted to the International Workshop on Active Inference 2024

Bayesian grey-box identification of nonlinear convection effects in heat transfer dynamics

Authors: Wouter M. Kouw, Caspar Gruijthuijsen, Lennart Blanken, Enzo Evers, Timothy Rogers

Abstract: We propose a computational procedure for identifying convection in heat transfer dynamics. The procedure is based on a Gaussian process latent force model, consisting of a white-box component (i.e., known physics) for the conduction and linear convection effects and a Gaussian process that acts as a black-box component for the nonlinear convection effects. States are inferred through Bayesian smoo… ▽ More We propose a computational procedure for identifying convection in heat transfer dynamics. The procedure is based on a Gaussian process latent force model, consisting of a white-box component (i.e., known physics) for the conduction and linear convection effects and a Gaussian process that acts as a black-box component for the nonlinear convection effects. States are inferred through Bayesian smoothing and we obtain approximate posterior distributions for the kernel covariance function's hyperparameters using Laplace's method. The nonlinear convection function is recovered from the Gaussian process states using a Bayesian regression model. We validate the procedure by simulation error using the identified nonlinear convection function, on both data from a simulated system and measurements from a physical assembly. △ Less

Submitted 9 July, 2024; v1 submitted 1 July, 2024; originally announced July 2024.

Comments: 6 pages, 4 figures. Published in the proceedings of the IEEE Conference on Control Technology and Applications 2024

Information-seeking polynomial NARX model-predictive control through expected free energy minimization

Authors: Wouter M. Kouw

Abstract: We propose an adaptive model-predictive controller that balances driving the system to a goal state and seeking system observations that are informative with respect to the parameters of a nonlinear autoregressive exogenous model. The controller's objective function is derived from an expected free energy functional and contains information-theoretic terms expressing uncertainty over model paramet… ▽ More We propose an adaptive model-predictive controller that balances driving the system to a goal state and seeking system observations that are informative with respect to the parameters of a nonlinear autoregressive exogenous model. The controller's objective function is derived from an expected free energy functional and contains information-theoretic terms expressing uncertainty over model parameters and output predictions. Experiments illustrate how parameter uncertainty affects the control objective and evaluate the proposed controller for a pendulum swing-up task. △ Less

Submitted 22 December, 2023; originally announced December 2023.

Comments: 6 pages, 3 figures

Variational Bayes for robust radar single object tracking

Authors: Alp Sarı, Tak Kaneko, Lense H. M. Swaenen, Wouter M. Kouw

Abstract: We address object tracking by radar and the robustness of the current state-of-the-art methods to process outliers. The standard tracking algorithms extract detections from radar image space to use it in the filtering stage. Filtering is performed by a Kalman filter, which assumes Gaussian distributed noise. However, this assumption does not account for large modeling errors and results in poor tr… ▽ More We address object tracking by radar and the robustness of the current state-of-the-art methods to process outliers. The standard tracking algorithms extract detections from radar image space to use it in the filtering stage. Filtering is performed by a Kalman filter, which assumes Gaussian distributed noise. However, this assumption does not account for large modeling errors and results in poor tracking performance during abrupt motions. We take the Gaussian Sum Filter (single-object variant of the Multi Hypothesis Tracker) as our baseline and propose a modification by modelling process noise with a distribution that has heavier tails than a Gaussian. Variational Bayes provides a fast, computationally cheap inference algorithm. Our simulations show that - in the presence of process outliers - the robust tracker outperforms the Gaussian Sum filter when tracking single objects. △ Less

Submitted 28 September, 2022; originally announced September 2022.

Comments: 6 pages, 8 figures. Published as part of the proceedings of the IEEE International Workshop on Signal Processing Systems 2022

Online system identification in a Duffing oscillator by free energy minimisation

Authors: Wouter M Kouw

Abstract: Online system identification is the estimation of parameters of a dynamical system, such as mass or friction coefficients, for each measurement of the input and output signals. Here, the nonlinear stochastic differential equation of a Duffing oscillator is cast to a generative model and dynamical parameters are inferred using variational message passing on a factor graph of the model. The approach… ▽ More Online system identification is the estimation of parameters of a dynamical system, such as mass or friction coefficients, for each measurement of the input and output signals. Here, the nonlinear stochastic differential equation of a Duffing oscillator is cast to a generative model and dynamical parameters are inferred using variational message passing on a factor graph of the model. The approach is validated with an experiment on data from an electronic implementation of a Duffing oscillator. The proposed inference procedure performs as well as offline prediction error minimisation in a state-of-the-art nonlinear model. △ Less

Submitted 2 September, 2020; originally announced September 2020.

Comments: 10 pages, 5 figures. Accepted to the International Workshop on Active Inference (final author version)