Quadrotor attitude control under perturbations with neural network and state observer

• Propose an improved sparrow search algorithm with chaotic mapping for BPNN weight optimization. • Enable adaptive PID tuning through ISSA-optimized BPNN. • Develop an integrated ISSA-BP-PID-LESO framework for quadrotor attitude control. • Reduce tracking error by 65.5%, overshoot by 88.9%, and steady-state error by 98.3% compared with PID. • Improve RMSE by 33.4%, validating robustness under uncertain flight conditions. This paper proposes an integrated control framework for quadrotor attitude regulation that addresses the challenges of nonlinear dynamics, external disturbances, internal uncertainties, and real-time adaptability. The framework incorporates an improved sparrow search algorithm to optimize the parameters of a neural network, enabling accurate approximation of complex system behaviors. A linear extended state observer is employed to estimate and compensate for time-varying external disturbances in real time, while a dynamic regulation module ensures rapid and stable system response. Simulation results demonstrate that the proposed method significantly enhances control accuracy and robustness under complex and uncertain conditions, outperforming other state-of-the-art methods.