Earth-based Simultaneous Localization and Mapping for Drones in Dynamic Environments

This paper addresses the problem of simultaneous localization, mapping, and moving object tracking (SLAMMOT) with application to unmanned aerial vehicles inside uncertain environments with both static and moving objects. Although this problem has been addressed by the community, the usual approach is to identify the moving objects and remove them from contributing to the localization and mapping algorithm. Conversely, the proposed strategy integrates the moving objects in the vehicle motions estimation, without compromising its accuracy. The proposed solution is based on the design of: i) an extended Kalman filter (EKF); ii) a modified version of the multiple hypothesis tracking method to perform data association and track moving objects; and iii) and the interacting multiple model algorithm to identify the motion models described by the environment’s objects. The performance of the devised SLAMMOT filter is studied in simulation and validated with a new public experimental dataset using an RGB-D camera on-board an instrumented quadrotor with ground-truth.