Midterm Colloquium Laurens Espada

Title: Improving LiDAR-Inertial Odometry for Drones in Aircraft Inspections

Supervisor: Manuel Mazo Espinosa

Abstract: Mainblades is a company which uses drones to perform autonomous inspections of aircraft to automatically detect and notify aircraft operators of various types of damage. To do this, the drone needs to be able to navigate around the aircraft in high-traffic environments and localize with respect to the aircraft with centimeter-level accuracy, often under the stress of wind disturbances and low-visibility conditions.

The common LiDAR-inertial sensor-fusion methods for determining the drone’s pose and velocity can either result in jittery signals with large amounts of drift, be computationally expensive, fail to deal with the dynamic nature of aircraft hangars, or perform very poorly under low-visibility conditions. I discuss the trade-offs one has to make when designing a state-estimation system for this application and propose a method to address them.