UAV-Borne Mapping Algorithms for Low-Altitude and High-Speed Drone Applications.

Area of Science:

• Robotics and Computer Vision
• Sensor Fusion and Perception for Autonomous Systems

Background:

• Advancements in Unmanned Aerial Vehicle (UAV) technology necessitate robust mapping solutions for low-altitude, high-speed operations.
• Accurate real-time environmental mapping is crucial for UAV navigation and mission success.

Purpose of the Study:

• To analyze state-of-the-art sensors and mapping algorithms for UAV applications.
• To evaluate the performance of Direct Sparse Odometry (DSO), Stereo DSO (SDSO), and DSO Lite (DSOL) in high-fidelity simulated environments.

Main Methods:

• Development of a novel experimental construct integrating the AirSim simulator with Google 3D maps via the Cesium Tiles plugin.
• Comparative performance evaluation of DSO, SDSO, and DSOL based on geometric accuracy and computational speed.
• Experiments conducted in a highly realistic simulated environment mimicking real-world conditions.

Main Results:

• Quantification of the trade-offs between computational performance and map accuracy for each algorithm.
• DSOL identified as the optimal choice for UAVs with limited computing resources.
• SDSO recommended for systems with payload capacity and moderate computational power.

Conclusions:

• Algorithm selection for UAV mapping requires balancing computational demands with desired map density and accuracy.
• DSO is suitable for single-camera applications demanding dense mapping.
• The study provides critical insights for researchers to select the most appropriate mapping algorithm for specific UAV applications.