APR-Transformer: Initial Pose Estimation for Localization in Complex Environments through Absolute Pose Regression

TL;DR

APR-Transformer introduces a Transformer-based framework for absolute pose regression using either camera images or LiDAR inputs to improve initialization in GNSS-denied environments. The architecture uses modality-specific preprocessing and dual Transformer branches to simultaneously regress position and orientation, with a Kendall-style loss balancing $L_p$, $L_o$, and $L_{pose}$. It achieves strong results on DeepLoc and Radar Oxford Robot-Car benchmarks and demonstrates robustness on the APR-BeIntelli dataset, including real-time validation on an autonomous vehicle and integration into Autoware.Universe. The work highlights practical applicability for initializing localization systems and suggests future multimodal fusion to further enhance accuracy and reliability. Overall, APR-Transformer provides a scalable, real-time solution for robust initial pose estimation in complex urban environments.

Abstract

Precise initialization plays a critical role in the performance of localization algorithms, especially in the context of robotics, autonomous driving, and computer vision. Poor localization accuracy is often a consequence of inaccurate initial poses, particularly noticeable in GNSS-denied environments where GPS signals are primarily relied upon for initialization. Recent advances in leveraging deep neural networks for pose regression have led to significant improvements in both accuracy and robustness, especially in estimating complex spatial relationships and orientations. In this paper, we introduce APR-Transformer, a model architecture inspired by state-of-the-art methods, which predicts absolute pose (3D position and 3D orientation) using either image or LiDAR data. We demonstrate that our proposed method achieves state-of-the-art performance on established benchmark datasets such as the Radar Oxford Robot-Car and DeepLoc datasets. Furthermore, we extend our experiments to include our custom complex APR-BeIntelli dataset. Additionally, we validate the reliability of our approach in GNSS-denied environments by deploying the model in real-time on an autonomous test vehicle. This showcases the practical feasibility and effectiveness of our approach. The source code is available at:https://github.com/GT-ARC/APR-Transformer.

Figures (10)

• Figure 1: APR-Transformer model architecture employs absolute pose regression with two Transformers separately querying features for position and orientation. This architecture utilizes a single camera image as input; features from a CNN backbone (EfficientNet variants) are extracted which are further processed by the Transformers and MLP heads regressing position and orientation.
• Figure 2: We experiment with different architecture design choices and input features by converting LiDAR data to task-specific modalities. a) 2D histogram representation of input LiDAR data; EfficientNet is used as the backbone to extract multi-resolution features further processed by the Transformer blocks performing pose regression. b) PointNet++ is used to extract features from raw 3D points of LiDAR data further processed by the Transformer blocks performing pose regression.
• Figure 3: DeepLoc dataset (0.7m, 3.35°)
• Figure 4: Robot-Car (3.65m, 0.60°)
• Figure 5: Robot-Car (4.13m, 0.63°)