Suko Wiyanto, Rini Nur Hasanah, Muhammad Aziz Muslim, Nanang Sulistiyanto, Raden Arief Setyawan
Quadcopter unmanned aerial vehicles (UAVs) are increasingly deployed in surveillance, logistics, agriculture, and defense, yet their control remains challenging due to nonlinear dynamics and underactuation. This systematic literature review traces the evolution of quadcopter control from classical proportional–integral–derivative (PID) schemes to reinforcement-learning-based and hybrid intelligent strategies. Publications from 2015 to 2025 were collected through Scopus, IEEE Xplore, ScienceDirect, and SpringerLink using PRISMA and ROBIS frameworks to ensure methodological rigor. The synthesis identifies four dominant themes: (1) learning-optimal control, including reinforcement learning, adaptive dynamic programming, and neural-enhanced model predictive approaches; (2) robust and adaptive nonlinear control augmented with neural or fuzzy systems; (3) learning-based estimation and fault detection integrating physics-informed neural networks with Kalman filtering; and (4) intelligent PID tuning and component-level modeling that enhance adaptability and energy planning. PID remains attractive for simplicity but lacks adaptability, while robust controllers offer stability at higher computational cost. Reinforcement learning enables autonomy yet suffers from limited sim-to-real transfer, and learning-enhanced estimation increases resilience but needs further validation. Hybridization of model-based guarantees with data-driven adaptability emerges as the most promising direction. Future work should establish standardized benchmarks, expand hardware validations, and develop hybrid frameworks bridging classical and intelligent paradigms toward resilient, energy-efficient UAV autonomy. © 2026 by authors and Galileo Institute of Technology and Education of the Amazon (ITEGAM).
Department of Electrical Engineering, Universitas Brawijaya, Malang, 65145, Indonesia
