Pose-Graph Attentional Graph Neural Network for Lidar Place Recognition

TL;DR

This paper introduces P-GAT, a pose-graph attentional graph neural network that improves lidar place recognition by leveraging spatial-temporal relationships and attention mechanisms, outperforming current state-of-the-art methods especially in challenging environments.

Contribution

The paper presents a novel pose-graph GNN architecture that utilizes intra- and inter-attention for lidar place recognition, addressing limitations of frame-to-frame approaches.

Findings

P-GAT outperforms state-of-the-art methods on large-scale datasets.

Effective in environments with few features and different training/testing distributions.

Utilizes pose-graph SLAM concepts with attention mechanisms for improved recognition.

Abstract

This paper proposes a pose-graph attentional graph neural network, called P-GAT, which compares (key)nodes between sequential and non-sequential sub-graphs for place recognition tasks as opposed to a common frame-to-frame retrieval problem formulation currently implemented in SOTA place recognition methods. P-GAT uses the maximum spatial and temporal information between neighbour cloud descriptors -- generated by an existing encoder -- utilising the concept of pose-graph SLAM. Leveraging intra- and inter-attention and graph neural network, P-GAT relates point clouds captured in nearby locations in Euclidean space and their embeddings in feature space. Experimental results on the large-scale publically available datasets demonstrate the effectiveness of our approach in scenes lacking distinct features and when training and testing environments have different distributions (domain adaptation). Further, an exhaustive comparison with the state-of-the-art shows improvements in performance gains. Code is available at https://github.com/csiro-robotics/P-GAT.