A Comparison of LiDAR-based SLAM Systems for Control of Unmanned Aerial Vehicles

TL;DR

This study evaluates LiDAR-based SLAM systems for UAV control, comparing their pose estimation accuracy and control performance in flight scenarios, to determine their suitability for autonomous drone navigation.

Contribution

It provides a comparative analysis of Cartographer, LOAM, and HDL graph SLAM for UAV pose feedback, including integration into control systems and performance evaluation.

Findings

Cartographer, LOAM, and HDL SLAM differ in pose accuracy and control performance.

SLAM algorithms can be effectively integrated into UAV control systems.

Performance varies with flight scenarios and trajectory complexity.

Abstract

This paper investigates the use of LiDAR SLAM as a pose feedback for autonomous flight. Cartographer, LOAM and HDL graph SLAM are first introduced on a conceptual level and later tested for this role. They are first compared offline on a series of datasets to see if they are capable of producing high-quality pose estimates in agile and long-range flight scenarios. The second stage of testing consists of integrating the SLAM algorithms into a cascade PID UAV control system and comparing the control system performance on step excitation signals and helical trajectories. The comparison is based on step response characteristics and several time integral performancecriteria as well as the RMS error between planned and executed trajectory.