Spatio-Temporal Graph Localization Networks for Image-based Navigation

TL;DR

This paper introduces a learning-based spatio-temporal graph neural network for image-based robot localization that combines spatial and temporal cues, utilizing sim2real transfer to improve accuracy in environments with similar-looking images.

Contribution

It proposes a novel CNN-GNN architecture with semi-supervised sim2real transfer for improved localization in challenging indoor environments.

Findings

Outperforms state-of-the-art baselines in environments with similar images.

Significantly improves navigation accuracy in both simulation and real-world tests.

Effectively leverages semi-supervised learning with simulator and real images.

Abstract

Localization in topological maps is essential for image-based navigation using an RGB camera. Localization using only one camera can be challenging in medium-to-large-sized environments because similar-looking images are often observed repeatedly, especially in indoor environments. To overcome this issue, we propose a learning-based localization method that simultaneously utilizes the spatial consistency from topological maps and the temporal consistency from time-series images captured by the robot. Our method combines a convolutional neural network (CNN) to embed image features and a recurrent-type graph neural network to perform accurate localization. When training our model, it is difficult to obtain the ground truth pose of the robot when capturing images in real-world environments. Hence, we propose a sim2real transfer approach with semi-supervised learning that leverages simulator images with the ground truth pose in addition to real images. We evaluated our method quantitatively and qualitatively and compared it with several state-of-the-art baselines. The proposed method outperformed the baselines in environments where the map contained similar images. Moreover, we evaluated an image-based navigation system incorporating our localization method and confirmed that navigation accuracy significantly improved in the simulator and real environments when compared with the other baseline methods.