Scan Context++: Structural Place Recognition Robust to Rotation and Lateral Variations in Urban Environments

Place recognition is a key module in robotic navigation. The existing line of studies mostly focuses on visual place to recognize previously visited places solely based their appearance. In this article, we address structural by recognizing appearance, namely from range sensors. Extending our previous work rotation invariant spatial descriptor, the proposed descriptor completes generic robust both (heading) and translation when roll–pitch motions are not severe. We introduce two subdescriptors enable topological retrieval followed 1-degree freedom semimetric localization, thereby bridging gap between metric localization. method has been evaluated thoroughly terms environmental complexity scale. source code available can easily be integrated into light detection ranging simultaneous localization mapping.