Conformalized Multimodal Uncertainty Regression and Reasoning

TL;DR

This paper presents a lightweight conformal prediction-based uncertainty estimator for multimodal regression, applied to visual odometry, which improves prediction accuracy under challenging conditions by integrating uncertainty reasoning and optical flow analysis.

Contribution

It introduces a novel conformalized uncertainty estimation method for multimodal regression and combines it with reasoning techniques to enhance visual odometry accuracy.

Findings

Uncertainty estimates adapt effectively to noise and data limitations.

The combined framework reduces prediction error by 2-3 times.

Method outperforms conventional deep learning approaches in challenging scenarios.

Abstract

This paper introduces a lightweight uncertainty estimator capable of predicting multimodal (disjoint) uncertainty bounds by integrating conformal prediction with a deep-learning regressor. We specifically discuss its application for visual odometry (VO), where environmental features such as flying domain symmetries and sensor measurements under ambiguities and occlusion can result in multimodal uncertainties. Our simulation results show that uncertainty estimates in our framework adapt sample-wise against challenging operating conditions such as pronounced noise, limited training data, and limited parametric size of the prediction model. We also develop a reasoning framework that leverages these robust uncertainty estimates and incorporates optical flow-based reasoning to improve prediction prediction accuracy. Thus, by appropriately accounting for predictive uncertainties of data-driven learning and closing their estimation loop via rule-based reasoning, our methodology consistently surpasses conventional deep learning approaches on all these challenging scenarios--pronounced noise, limited training data, and limited model size-reducing the prediction error by 2-3x.