Autonomous robotic SLAM-based indoor navigation for high resolution sampling with complete coverage

Recent work has shown the feasibility of pedestrian and robotic indoor localization based only on maps of the magnetic field. To obtain a complete representation of the magnetic field without initial knowledge of the environment or any existing infrastructure, we consider an autonomous robotic platform to reduce limitations of economic or operational feasibility. Therefore, we present a novel robotic system that autonomously samples any measurable physical processes at high spatial resolution in buildings without any prior knowledge of the buildings' structure. In particular we focus on adaptable robotic shapes, kinematics and sensor placements to both achieve complete coverage in hardly accessible areas and not be limited to round shaped robots. We propose a grid based representation of the robot's configuration space and graph search algorithms, such as Best-First-Search and an adaption of Dijkstra's algorithm, to guarantee complete path coverage. In combination with an optical simultaneous localization and mapping (SLAM) algorithm, we present experimental results by sampling the magnetic field in an a priori unknown office with a robotic platform autonomously and completely.