📝 SummarXiv

Abstract

Natural disasters, such as hurricanes and typhoons, pose significant challenges to public safety and infrastructure. While government agencies rely on multi million dollar UAV systems for storm data collection and disaster response, smaller drones lack the ability to autonomously adapt to rapidly changing environmental conditions, such as turbulent winds. This project investigates the implementation of advanced state estimation filters, Extended Kalman Filter (EKF), Unscented Kalman Filter (UKF), and Particle Filter (PF), to enhance the control and adaptability of quadrotor drones in uncertain wind profiles modeled by Von Karman turbulence. While the EKF relies on linearization techniques using the Taylor series expansion, which can struggle under high nonlinearity, the UKF leverages sigma points for better performance in nonlinear systems without requiring Jacobian computations. The PF, on the other hand, addresses non-Gaussian noise and severe nonlinearities by employing a large number of particles, albeit at a high computational cost. To enhance accuracy and minimize estimation errors, a genetic algorithm (GA) was employed to optimally tune the process and measurement noise covariances (Q and R matrices) as well as UKF specific parameters alpha, beta, and kappa. This study highlights the tradeoff between accuracy, computational efficiency, and smoothing capabilities across these filters. Despite its robustness, the PF suffered from computational inefficiencies due to the high state dimensionality, while the EKF demonstrated faster computation but lower adaptability in nonlinear conditions. The UKF emerged as a balanced approach, achieving superior performance in capturing dynamic wind disturbances.