CONTROLLED FILTER WITH ACTIVE PERCEPTION FOR SIMULTANEOUS LOCALIZATION AND MAPPING (SLAM)

Resumo

This research presents a new method, called Controlled Filter with Active Perception (CFAP), to approach the problem of Simultaneous Localization and Mapping (SLAM). SLAM aims to map an unknown environment while estimating the trajectory of a mobile agent moving within that environment. CFAP combines the two fundamental pillars of SLAM, which are scan alignment and loop closure, into a single process. The method is inspired by human perception of locating oneself and for this, it uses a mental map to guide itself. CFAP uses a Gaussian distribution to estimate possible poses and performs the alignment process in cycles, where each cycle is influenced by the results of the previous one. The Active Perception mechanism is used in each cycle to determine the quality of each of the possible poses, allowing for more accurate simultaneous localization and mapping.