A Comprehensive Review of Recent Research Trends on UAVs

The paper situates UAV research within the broader rise of artificial intelligence (AI) across sectors such as robotics, healthcare, cybersecurity, education, energy, smart cities, transportation, and agriculture. AI is a critical enabler of UAV capabilities in navigation, object detection, and mission planning. Given UAVs’ rapid growth and diverse applications, numerous reviews have emerged across subdomains (e.g., open-source hardware/software, frame design, control, communication including 5G, AI integration, recognition/detection, and path planning). The authors highlight key research questions stemming from the interdisciplinary nature of UAVs (spanning aerospace, computer science, robotics, remote sensing), challenges and opportunities (regulation, privacy, safety), and future directions. They emphasize urgent inquiries on autonomy via ML/AI, robust navigation and sense-and-avoid, swarm intelligence, energy and power management, payload capacity, security and privacy, airspace integration, regulation, and human-UAV interaction. This review aims to guide new researchers by surveying classifications, architectures, open-source ecosystems, and tools, and by quantitatively mapping recent research directions (using Scopus) and their interconnections. The contributions include: a comprehensive reference collection; data-driven insights and predictions about fast-growing directions in the last three years; identification of open development axes; a systematic discussion of control algorithm selection; an overview of high-level software; and numerical analysis of interrelationships among UAV research directions.

The paper consolidates and synthesizes a large body of recent surveys and reviews covering UAV classifications (by flying principle, mission, weight, propulsion, control, altitude, configuration, purpose, launch method, payload, autonomy, size, endurance, and range), communication and antennas (including FANETs, 5G, satellite, mmWave), AI integration (object detection, tracking, navigation, swarms), perception and sensing (camera, LiDAR, radar), control methodologies (classical, modern, intelligent, adaptive, advanced techniques such as MPC, SMC, DRL), energy-efficient flight (aerodynamics, propulsion, solar, hybrid systems), human-UAV interaction (gestures, VR/AR, natural language), swarming (coordination, architectures, path planning), and application domains (agriculture, construction, marine, mining, military, wildlife and environmental monitoring). It also surveys open-source platforms and tools (PX4, ArduPilot, Paparazzi, ROS, MAVLink, QGroundControl, Mission Planner, AirSim, Gazebo, TensorFlow, OpenCV), providing a state-of-the-art context for the trends and open problems identified.

The review quantitatively analyzes recent UAV research trends using Scopus. Search strategy: TITLE-ABS-KEY queries for “UAV”, “drone”, “unmanned aerial vehicle”, and “unmanned aerial systems” restricted to PUBYEAR > 2019 and < 2024. The systematic search (fields: Title, Abstract, Keywords) yielded 47,635 references (2020–2023), collected on March 14, 2023. The authors categorized publications into major research directions (e.g., antennas, aircraft detection, remote sensing, AI including DL/RL/ML, aircraft control, IoT, trajectories, energy efficiency/utilization, etc.) and quantified document counts per direction and their interconnections. To assess growth dynamics (2020–2022), they fitted linear models of the form y = a x + b, where y is publication count, x is year, with parameter a representing rate of growth per year. They further computed an “acceleration of growth” ratio to indicate how rapidly interest increases across directions. The study complements this bibliometric mapping with narrative synthesis on open development axes, control methodologies, hardware/software architectures, applications, and open issues.

- Scale and prominence of directions (2020–2023): - Antennas: 22,150 documents (largest direction), with strong links to aircraft detection, AI, remote sensing, control, and IoT. - Aircraft detection: 5,604 documents; strong interconnection with antennas. - AI: 5,789 documents; includes DL (3,631) and RL (1,618). - Remote sensing: 3,983 documents. - Aircraft control: 2,504 documents. - IoT: 1,702 documents. - Additional topics noted: Machine Learning (1,761), Energy efficiency (1,293), Energy utilization (134), Agricultural robots (1,606). - Interconnections (examples from antenna-centric links): - Antennas ↔ Aircraft detection: 3,749 documents; Antennas ↔ AI: 4,231; Antennas ↔ Remote sensing: 2,176; Antennas ↔ Aircraft control: 1,707; Antennas ↔ IoT: 1,092. - Growth dynamics (average rate per year; acceleration ratio): - Antennas: +980 docs/year; acceleration ≈ 1.20. - Aircraft detection: +447/year; acceleration ≈ 1.19. - AI: +865/year; acceleration ≈ 1.33. - IoT: +155/year; acceleration ≈ 1.34. - Remote sensing: +93/year; acceleration ≈ 9.08 (notable surge in interest). - Energy-efficient: +289/year; acceleration ≈ 2.86. - Control: +197/year; acceleration ≈ 2.43. - Swarm: +54/year; acceleration ≈ 0.81. - Thematically, antennas/communication lead the ecosystem and tightly couple with sensing, AI, and control; AI methods (DL, RL) are pivotal across detection, trajectory optimization, and autonomy. - Open development axes highlighted: AI integration (including Generative AI/ChatGPT as natural language interfaces and for high-level planning/code generation), environmental monitoring and conservation (including Antarctic studies), Urban Air Mobility, miniaturization, swarming and cooperative control, BVLOS, long-range/high-altitude with renewable energy, flight safety, suspension payload capabilities, transformability/convertibility (morphing wings, foldable/reconfigurable frames, variable pitch propellers, transformable rotors/VTOL). - Control landscape: classical (PID), modern (state-space, LQR), intelligent (NN, fuzzy, genetic), adaptive (MRAC); advanced techniques include MPC, SMC (with chattering mitigation and reaching laws), DRL, neural control, cooperative and fault-tolerant control (FDD/FTC), and Prescribed Performance Control. - Hardware/software architectures: layered avionics (flight computer/controller, sensors, actuators/ESCs, power/battery, comms, payload, structure) and software stacks (firmware, OS, middleware, applications). Open-source ecosystems (PX4, ArduPilot, Paparazzi, ROS, MAVLink, QGroundControl, Mission Planner, AirSim, Gazebo, TensorFlow, OpenCV, PyTorch, DroneKit) enable rapid R&D. - Applications: agriculture, construction/civil, marine, mining, environmental and wildlife monitoring, disaster response, surveillance, logistics, Antarctic science, among others. - Open issues: operability (autonomy, planning, swarming, energy/payload limits), technology/regulatory (comms, networking, sensing, standardization), and safety/privacy/security (cybersecurity, GPS spoofing, anti-drone tech). Future research directions include swarm systems, security and privacy (e.g., blockchain, physical-layer security), improved path planning, enhanced charging/energy systems (solar, WPT), and Optical Wireless Communications.

The mapping shows UAV research as highly interconnected, with antennas/communications acting as a central hub connecting sensing, AI, and control. The strong cross-links suggest that integrated solutions—co-designing communications, perception, and control—are essential to advance reliable autonomy, BVLOS operations, and multi-UAV collaboration. Rapid growth in remote sensing and energy-efficient flight indicates rising demand for data-rich, long-endurance missions, while accelerating AI publications underscore AI’s role in perception, planning, and decision-making. The results address the research questions by clarifying where activity is concentrated, how subfields interact, and which directions (e.g., AI, remote sensing, control, energy efficiency) are accelerating. This informs priorities such as: developing robust sense-and-avoid and safety mechanisms; advancing swarm coordination with scalable comms/control architectures; improving power systems (hybrid/solar) and energy-aware planning; and fostering human-UAV interfaces (including language-based control) to safely unlock new applications. The open-source toolchain further lowers barriers for reproducible research and accelerates translation from simulation to field testing.

This review offers a comprehensive synthesis of UAV research over the last three years, combining a bibliometric analysis of Scopus-indexed literature with an in-depth narrative across classifications, architectures, control methodologies, software ecosystems, and applications. Key contributions include quantifying prominent directions (notably antennas/communications, AI, detection, remote sensing, and control), revealing strong interconnections among subfields, and identifying open development axes (AI integration, conservation and environmental monitoring, UAM, miniaturization, swarming/cooperative control, BVLOS, long-range/high-altitude, safety, payload suspension, and transformability). The paper also systematizes control approaches from classical to advanced (MPC, SMC, DRL, PPC, FTC) and surveys open-source platforms enabling rapid development. Future work should emphasize integrated comms–perception–control co-design, robust safety/security and privacy, energy technologies (including wireless/solar charging), standardized regulations and interoperable protocols, and human-centric interaction paradigms (e.g., language-based control). These directions will help realize scalable, safe, and sustainable UAV deployments across diverse domains.

- Data source restricted to Scopus and specific keywords (UAV, drone, unmanned aerial vehicle/systems); relevant works outside these terms or databases may be undercounted. - Time window focused on recent years (2020–2023; growth modeling emphasized 2020–2022), potentially biasing against earlier foundational research or very recent publications after March 14, 2023. - Publication counts include heterogeneous document types (journals, conferences, books), which may weigh fields differently. - Linear growth modeling (y = a x + b) and the reported acceleration ratios simplify complex temporal dynamics and may not capture non-linear surges or plateau effects. - Interconnection counts inferred from co-occurrence within queries may reflect terminology overlap rather than deep methodological integration.