Contrôle Intelligent, Planification et Optimisation de la Trajectoire d’un Drone Miniaturé

Abstract

Quadcopters are widely applied in surveillance, inspection, monitoring, and delivery, where precise trajectory tracking and energy efficiency are critical under uncertain environments. This work presents advanced control strategies that combine nonlinear design, heuristic optimization, and neural adaptation to address these challenges. First, a nonlinear backstepping controller is developed for the six-degree-of-freedom dynamics of the quadcopter. To overcome manual gain selection, nature-inspired algorithms such as Grey Wolf Optimizer (GWO), Garra Rufa Optimization (GRO), and Pelican Optimization Algorithm (POA) are employed for automatic tuning. Simulations confirm improved accuracy and robustness against wind disturbances compared to traditional methods. Second, energy-efficient flight control is achieved through a Linear Quadratic Regulator (LQR) with adaptive weighting matrix tuning. A hybrid scheme using GWO and feedforward neural networks (FNNs) adjusts performance and control effort weightings, reducing actuator load while maintaining stability. The first framework, particularly with POA, ensures robustness, adaptability, and accurate path tracking, while the second achieves reliable trajectory following with improved energy efficiency based on the enhancement of FNNs.