3D LiDAR-Based Place Recognition Techniques: A Review of the Past 10 Years

Accurate determination of a robot’s location, which is referred to as place recognition, is essential for achieving autonomous navigation. However, complex real-world environments pose numerous challenges for place recognition, including dynamic environmental interferences, appearance changes, and viewpoint changes. Researchers have made significant progress over the past decade in addressing these problems. In this paper, we focus on 3D LiDAR-based place recognition technology over this period and provide a comprehensive review of the methods and developments in this field. We aim to help new researchers quickly understand the current state of research and development trends in 3D LiDAR-based place recognition. We begin by providing an overview of relevant concepts and different technical approaches. We then provide a detailed review of the existing solutions for different technical approaches, the evaluation metrics, and the popular benchmark datasets. Next, we summarize the development trends of existing methods and identify the key challenges of place recognition. Finally, we discuss real-world applications of 3D LiDAR-based place recognition and outline future research directions.