Learning deep dynamical models from image pixels
This addresses a challenging issue in fields such as control and robotics, where system dynamics must be inferred from noisy, high-dimensional data, representing an incremental improvement over existing methods.
The paper tackles the problem of non-linear system identification from high-dimensional observations like images, by combining deep auto-encoders with predictive transition models in low-dimensional latent space, and demonstrates that this approach enables learning good predictive models from pixel data alone.
Modeling dynamical systems is important in many disciplines, e.g., control, robotics, or neurotechnology. Commonly the state of these systems is not directly observed, but only available through noisy and potentially high-dimensional observations. In these cases, system identification, i.e., finding the measurement mapping and the transition mapping (system dynamics) in latent space can be challenging. For linear system dynamics and measurement mappings efficient solutions for system identification are available. However, in practical applications, the linearity assumptions does not hold, requiring non-linear system identification techniques. If additionally the observations are high-dimensional (e.g., images), non-linear system identification is inherently hard. To address the problem of non-linear system identification from high-dimensional observations, we combine recent advances in deep learning and system identification. In particular, we jointly learn a low-dimensional embedding of the observation by means of deep auto-encoders and a predictive transition model in this low-dimensional space. We demonstrate that our model enables learning good predictive models of dynamical systems from pixel information only.