Range-Visual-Inertial Odometry: No Need To Be Excited

Traveling straight at constant speed is the most efficient trajectory for most robotics applications. Unfortunately without accelerometer excitation, monocular Visual-Inertial Odometry (VIO) cannot observe scale and suffers severe error drift. This is why NASA’s Ingenuity Mars Helicopter added a 1Dlaser range finder, but it is limited to flat terrains. Our novel range-VIO approach uses facet constraints, which efficiently leverage VIO feature depths to adapt to any scene structure.Due to the small range finder footprint, range-VIO also retains the minimal size, weight, power and real-time attributes of VIO.The facet scene model can scale from a flat world assumption,to virtually no structure assumption at all based on visual feature density. An important theoretical result shows that scale, with range-VIO, is no longer in the right nullspace of the observability matrix for constant acceleration motion.The benefits are evaluated on real flight data representative of common aerial robotics scenarios. Robustness is demonstrated using indoor stress data with full-state ground truth.