Skip to yearly menu bar Skip to main content


Switching Linear Dynamics for Variational Bayes Filtering

Philip Becker-Ehmck · Jan Peters · Patrick van der Smagt

Pacific Ballroom #256

Keywords: [ Time Series and Sequence Models ] [ Robotics ] [ Representation Learning ] [ Deep Generative Models ] [ Approximate Inference ]


System identification of complex and nonlinear systems is a central problem for model predictive control and model-based reinforcement learning. Despite their complexity, such systems can often be approximated well by a set of linear dynamical systems if broken into appropriate subsequences. This mechanism not only helps us find good approximations of dynamics, but also gives us deeper insight into the underlying system. Leveraging Bayesian inference, Variational Autoencoders and Concrete relaxations, we show how to learn a richer and more meaningful state space, e.g. encoding joint constraints and collisions with walls in a maze, from partial and high-dimensional observations. This representation translates into a gain of accuracy of learned dynamics showcased on various simulated tasks.

Live content is unavailable. Log in and register to view live content