User detection, recognition and tracking is at the heart of Human–Robot Interaction, and yet, to date, no universal
robust method exists for being aware of the people in a robot’s surroundings. The present paper imports into existing social robotic platforms different techniques, some of them classical, and other novel, for detecting, recognizing and tracking human users. The outputs from the parallel execution of these algorithms are then merged, creating a modular, expandable and fast architecture. This results in a local user mapping through fusion of multiple user recognition techniques. The different people detectors comply with a common interface called PeoplePoseList Publisher, while the people recognition algorithms meet an interface called PeoplePoseList Matcher. The fusion of all these different modules is based on the Unscented Kalman Filtering technique. Extensive benchmarks of the sub-components and of the whole architecture demonstrate the validity and interest of all levels of the architecture. In addition, all the software and datasets generated in this work are freely available.