Switching Linear Dynamics for Variational Bayes Filtering
This work addresses model predictive control and reinforcement learning by providing a method to better approximate nonlinear systems, though it appears incremental as it builds on existing variational and autoencoder techniques.
The paper tackled the problem of system identification for complex nonlinear systems by approximating them with switching linear dynamics, resulting in a more meaningful state space representation and improved accuracy in learned dynamics across simulated tasks.
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.