Fundamentals, Challenges, and Improved Algorithms of PID Control for Quadrotor UAVs

Abstract

The quadrotor unmanned aerial vehicles (UAVs) are widely employed in power line inspections, agricultural monitoring, cargo delivery, and surveying operations because of their straightforward design and remarkable agility. However, traditional proportional-integralderivative (PID) controllers perform poorly under wind, load changes, and sensor noise, which limits the UAV’s precision and flight reliability. Thus, this paper reviews the current development of PID control for quadcopter UAVs, examines the limitations of traditional PID controllers, and compares the performance and applications of various improved algorithms, offering guidance for controller selection and parameter tuning. By analyzing relevant literature on improvement strategies, such as fuzzy adaptive PID, evolutionary learning-optimized PID, data-driven adaptive PID, deep reinforcement learning PID (DRL-PID), and hybrid model PID, and integrating existing simulation and experimental data, key performance indicators such as stability, speed, disturbance rejection, and computational load are summarized. The results show that data-driven PID like DRL-PID achieves the highest accuracy, while fuzzy adaptive PID remains widely used in engineering for its low computational load and real-time performance. Each algorithm has its strengths in different applications: fuzzy adaptive PID is suitable for real-time tasks, RL-GA and hybrid models are suitable for high-precision mapping, and WRLS adaptive PID is suitable for logistics with frequently changing loads. Meanwhile, lightweight DRLPID and multi-UAV cooperative control are promising future directions.