Performance Assessment of LiDAR-Visual-Inertial Fusion for GNSS-Resilient Aerial Navigation

This work proposes an innovative LiDAR-Visual-Inertial navigation algorithm specifically conceived to face the challenges offered by a flying platform in a GNSS-denied environment. An Extended Kalman Filter is employed and loose integration of exteroceptive sensors measurement is considered. LiDAR and camera measurements are coupled altogether to build a joint pose estimate which exploits the LiDAR to associate depth information to image features detected by the camera. Particular attention is paid to the integrity of the multi-sensor-based solution. To this aim, a set of gating criteria is introduced to assess the validity of each measurement to be used within the filter and thus to improve its resiliency against anomalous pose estimates. The strategy is assessed using a simulation environment based on MATLAB/Simulink and Unreal Engine, which allows to retrieve realistic visual and LiDAR data by simulating different aircraft trajectories, geometries and scenes. Results on simulated data show a maximum percentage error on the position estimate, computed with respect to the vehicle traveled distance, of 4.6%, and a maximum error on the attitude angles estimate of 1.6°. The paper also introduces the experimental setup that has been developed to collect experimental data. Flight experiments are being conducted using a customized heavy lift quadcopter equipped with high performance Inertial Measurement Unit, color camera with a narrow Field Of View, a color fisheye camera, a scanning LiDAR and a GNSS receiver.