Introduction
The field of aerial robotics has undergone a period of breakneck expansion over the last two decades. By 2016, premier robotics conferences such as ICRA and IROS were featuring over 60 UAV-related papers annually. This "information explosion" has made it increasingly difficult for researchers to identify long-term trends, compare disparate works, or establish a unified taxonomy for the staggering variety of vehicle types. This paper addresses these challenges by providing a systematic survey of 1,318 high-quality papers published between 2001 and 2016, offering a curated roadmap for the evolving landscape of Unmanned Aerial Vehicles (UAVs).
Systematic Methodology
To ensure a reproducible and high-quality overview, the authors employed a rigorous three-step identification process. First, an automated screening script collected over 30,000 titles and abstracts from eight top-tier robotics journals and conferences. Second, a keyword filtering phase narrowed this pool using 35 specific terms such as "quadcopter," "MAV," and "VTOL."
Finally, the authors conducted a manual verification against five strict criteria. For instance, papers were required to include flight-related results (simulated or experimental) and, for computer vision topics, utilize data collected from an onboard camera rather than a handheld or external source. The final corpus of 1,318 papers was organized using TagSpaces, an open-source file tagging software, and has been made publicly available to the research community.
Evolution of the Field: Growth and Platforms
A primary insight from the survey is the accelerating pace of research. Since 2001, the volume of high-quality publications has followed a clear exponential trajectory, reflecting the field's transition from a niche interest to a central pillar of robotics research.
Beyond sheer volume, the nature of research platforms has shifted significantly. While helicopters were the early favorites, quadcopters have become the dominant research platform due to their mechanical simplicity, low cost, and ease of control. More recently, hexacopters and octocopters have gained traction because they offer redundant actuation—allowing for safe landings even in the event of motor failure—and the ability to carry heavier sensor payloads like 3D LiDAR.
Shifting Research Priorities
The survey reveals a maturation of research objectives. Early work was heavily weighted toward "how to fly," focusing on hardware development and low-level modeling and control. While these areas still comprise over half of the research output, the focus is increasingly shifting toward "what to do"—encompassing task planning, autonomous navigation, and human-robot interaction (HRI).
As shown in the topic distribution, hardware development (30.6%) and modeling/control (27.8%) remain foundational. However, task planning (17.1%) and localization/mapping (12.4%) represent significant and rapidly growing sectors. This shift is accompanied by new taxonomies that categorize drones not just by physical constraints like mass or flight time, but by functional dimensions such as autonomy and sociability.
Conclusion and Future Outlook
The data highlights a maturing field with clear geographic leaders; the USA accounts for 35.1% of the paper volume, followed by Switzerland (10.5%) and France (8.7%). There has also been a marked increase in experimental rigor, evidenced by the sharp rise in the use of high-precision motion capture systems for ground-truth evaluation since 2009.
Looking toward the next decade, the authors identify three emerging frontiers: the development of holonomic UAVs (capable of moving in any direction without tilting), advanced localization in complex environments, and more sophisticated human-drone interaction. While commercial platforms continue to advance, the need for open, technically detailed academic research remains vital for the continued evolution of the field.
Summary updated