Visual SLAM for Unmanned Aerial Vehicles: Localization and Perception

Area of Science:

• Robotics and Artificial Intelligence
• Computer Vision
• Autonomous Systems

Background:

• Localization and perception are fundamental for autonomous Unmanned Aerial Vehicle (UAV) applications.
• Simultaneous Localization And Mapping (SLAM) is a key technology for these tasks, evolving with hardware, multi-sensor, and AI advancements.

Purpose of the Study:

• To survey the development of visual SLAM and its application in UAVs.
• To review state-of-the-art algorithms addressing challenges in visual SLAM for UAVs.
• To outline research progression and future directions in UAV localization and perception.

Main Methods:

• Review of recent and state-of-the-art visual SLAM algorithms.
• Analysis of solutions for real-time performance, texture-less, and dynamic environments.
• Discussion of visual-inertial fusion and learning-based enhancements for UAVs.

Main Results:

• Identified critical problems and solutions in visual SLAM for UAVs.
• Highlighted the role of visual-inertial fusion and learning-based methods.
• Provided comprehensive preliminaries including algorithm components, camera configuration, and data processing.

Conclusions:

• Visual SLAM is crucial for autonomous UAVs, with ongoing advancements.
• Future trends point towards enhanced real-time performance, robustness in challenging environments, and AI integration.
• The survey offers a decade-long perspective on visual SLAM for UAVs, identifying research gaps and future opportunities.