PinNet: Keypoint-Aware Learned Local Descriptors with Geometric Embedding for Loop Closure in LiDAR SLAM

2026-06-26

AI Summary

This work addresses the challenge of robust loop closure detection in purely geometric LiDAR SLAM, where reliance solely on point cloud geometry often yields insufficiently discriminative local features. To overcome this limitation, the authors propose PinNet, an end-to-end neural network that jointly learns point cloud keypoints and their corresponding local descriptors. A key innovation is the introduction of a plane-based geometric self-attention module that explicitly models spatial relationships among keypoints. Notably, PinNet achieves highly discriminative local descriptions and accurate single-scan localization without requiring semantic labels or texture cues. Extensive experiments demonstrate that the method significantly improves place recognition accuracy and relative pose estimation precision across multiple large-scale, multi-session datasets.

Abstract

Loop closure is essential to reduce drift and build globally consistent maps in large-scale environments. However, reliable loop closure with only geometric information from, e.g., a LiDAR sensor, remains challenging due to the difficulty of constructing discriminative geometric features. We present PinNet, a neural network that produces local geometric descriptors from point clouds for place recognition and scanto-scan registration. PinNet incorporates a neural network that generates keypoints and their corresponding descriptors, together with a plane-based geometric self-attention module that models inter-keypoint spatial relationships to enhance descriptor discriminability for loop-closure detection and point-cloud registration. The approach is comprehensively evaluated on multiple datasets collected with different LiDAR sensors. Experimental results demonstrate strong place-recognition performance, precise relative pose estimation, and successful single-shot localization in different environments.

Problem

• loop closure
• LiDAR SLAM
• geometric descriptors
• place recognition
• point cloud registration

Innovation

• keypoint-aware descriptors
• geometric self-attention
• LiDAR SLAM
• loop closure
• plane-based embedding