Towards Dependable Autonomous Systems Based on Bayesian Deep Learning Components

TL;DR

This paper presents a method for propagating uncertainty in Bayesian Deep Learning components within autonomous systems, improving overall system confidence and performance in aerial navigation tasks.

Contribution

It introduces a novel approach to consider uncertainty propagation between BDL components, enhancing system dependability and performance.

Findings

Uncertainty propagation improves system confidence.

The approach slightly enhances navigation performance.

It enables better uncertainty estimation in autonomous systems.

Abstract

As autonomous systems increasingly rely on Deep Neural Networks (DNN) to implement the navigation pipeline functions, uncertainty estimation methods have become paramount for estimating confidence in DNN predictions. Bayesian Deep Learning (BDL) offers a principled approach to model uncertainties in DNNs. However, in DNN-based systems, not all the components use uncertainty estimation methods and typically ignore the uncertainty propagation between them. This paper provides a method that considers the uncertainty and the interaction between BDL components to capture the overall system uncertainty. We study the effect of uncertainty propagation in a BDL-based system for autonomous aerial navigation. Experiments show that our approach allows us to capture useful uncertainty estimates while slightly improving the system's performance in its final task. In addition, we discuss the benefits, challenges, and implications of adopting BDL to build dependable autonomous systems.