Robotic mapping assisted by local magnetic field anomalies

This paper presents a method to incorporate measurement of local magnetic field anomalies into the SLAM (Simultaneous Localization And Mapping) algorithm. One of the key problems of SLAM is loop closure, which means to map the same place into the same location on the generated map when the place is revisited by the robot. It is particularly important for large area map consistency. Steel structures, furniture and equipment inside a building disturb the natural magnetic field of the earth. These local anomalies of magnetic field are usually considered as noise when using an electronic compass on a robot. But, in this paper, we utilize these data to help solving the SLAM loop closure problem, since the magnetic anomaly patterns are different across a given building and stay relatively stable over time. We used particle filter as the basic algorithm for SLAM. In the weight calculation step of the particle filter algorithm, it assigns a weight value to each particle based on its matching score to the current sensor data. This step allows us to incorporate different sensor data from both laser range finder, and magnetometer into SLAM process. To verify the effectiveness of this method, we have done experiments involving a robot runs through a building hallway. At some point, it is hijacked to an unknown place in the same hallway, and continues to run. The SLAM assisted by local magnetic field anomaly data has generated more consistent maps than those without the assistance.