Range-Aided Visual-Inertial Initialization for Micro Air Vehicles Under Insufficiently Excited Motion

Monocular Visual-Inertial Odometry (VIO) can offer precise state estimation vital for the autonomous navigation of micro aerial vehicles (MAVs). Nonetheless, state-of-art VIOs demand specific prior state information or to be executed under sufficiently excited motions for system initialization. MAVs often operate state estimation in unknown environments, with uniform or quasi-uniform motions being prevalent in their autonomous flights. When MAVs are in such insufficiently excited motion patterns, the subtle variations in acceleration can render the scale of the monocular VIO unobservable, impeding the effective establishment of an initial state. We propose the integration of a lightweight, low-cost one-dimensional Laser Range Finder (1D-LRF) to observe the initial scale, and construct a vision-inertial-range joint initialization in the sense of maximum a posteriori (MAP) problem. This initialization has been incorporated into ORB-SLAM3, through simulations and real-world experiments, we demonstrate that our proposed approach can effectively complete the initialization of the VIO in MAVs motions that are not sufficiently excited, enhancing the stability and robustness of the original system.