Robust Locally-Linear Controllable Embedding

TL;DR

This paper introduces a robust locally-linear controllable embedding (RCE) model that improves upon previous methods by directly modeling predictive densities and incorporating structures for noise robustness, enhancing control in noisy environments.

Contribution

The paper proposes a new RCE model with a structured generative process and a variational approximation that accounts for future observations, improving robustness over existing methods.

Findings

RCE outperforms E2C in noisy dynamics scenarios.

The model provides more accurate embeddings under noise.

Experimental results demonstrate significant improvements.

Abstract

Embed-to-control (E2C) is a model for solving high-dimensional optimal control problems by combining variational auto-encoders with locally-optimal controllers. However, the E2C model suffers from two major drawbacks: 1) its objective function does not correspond to the likelihood of the data sequence and 2) the variational encoder used for embedding typically has large variational approximation error, especially when there is noise in the system dynamics. In this paper, we present a new model for learning robust locally-linear controllable embedding (RCE). Our model directly estimates the predictive conditional density of the future observation given the current one, while introducing the bottleneck between the current and future observations. Although the bottleneck provides a natural embedding candidate for control, our RCE model introduces additional specific structures in the generative graphical model so that the model dynamics can be robustly linearized. We also propose a principled variational approximation of the embedding posterior that takes the future observation into account, and thus, makes the variational approximation more robust against the noise. Experimental results show that RCE outperforms the E2C model, and does so significantly when the underlying dynamics is noisy.