Fast and Robust Initialization for Visual-Inertial SLAM

TL;DR

This paper presents a fast, robust initialization method for visual-inertial SLAM that significantly improves accuracy and reliability, enabling quick and precise system startup in complex environments.

Contribution

It introduces a generalized, efficient initialization approach with novel observability and consensus tests, enhancing accuracy and robustness over previous methods.

Findings

Achieves less than 2 seconds initialization time on EuRoC dataset.

Reduces scale errors from up to 156% to around 5%.

Further decreases errors to below 1% with additional bundle adjustment.

Abstract

Visual-inertial SLAM (VI-SLAM) requires a good initial estimation of the initial velocity, orientation with respect to gravity and gyroscope and accelerometer biases. In this paper we build on the initialization method proposed by Martinelli and extended by Kaiser et al. , modifying it to be more general and efficient. We improve accuracy with several rounds of visual-inertial bundle adjustment, and robustify the method with novel observability and consensus tests, that discard erroneous solutions. Our results on the EuRoC dataset show that, while the original method produces scale errors up to 156%, our method is able to consistently initialize in less than two seconds with scale errors around 5%, which can be further reduced to less than 1% performing visual-inertial bundle adjustment after ten seconds.