3-D Collaborative Occupancy Grid-Based Inertial SLAM Based on Switching Floor Event

The miniature inertial measurement unit-based pedestrian navigation system (MIMU-PNS) has been widely discussed and researched in academia and industry, and there is a growing focus on the 3-D relative position information among multipedestrians. Nonetheless, position errors tend to accumulate over time in the MIMU-PNS, the conventional range-based collaborative navigation methods are subject to nonlinear line-of-sight (NLOS) in complex indoor environments, and depend on accurate knowledge of the pedestrian’s initial relative position and heading. To address the challenges associated with 3-D position error divergence, instability of ranging information, and poor access to initial relative states in MIMU-PNS, this article proposes a 3-D collaborative occupancy grid-based inertial simultaneous localization and mapping system (3D-COGI-SLAM) based on switching floor event (SFE) matching. Our proposed system does not require ranging information and any a priori information about the building and can effectively adapt to elevator, escalator, and stair activities. The results of the three-person experiments conducted in the two building scenarios demonstrated that our system has improved performance over the conventional range-based collaborative navigation system. When the initial relative state is unknown, our system achieves similar accuracy to the known case, and the average horizontal and height positioning errors are 2.87–0.13 m, respectively.