Particle Filter-Based State and Parameter Estimation for Delta Wing UAV

Online parameter estimation methods are essential for unmanned aerial vehicles (UAVs) as they enable real-time adaptation to changing flight conditions, enhance robustness against model uncertainties, and support fault detection by identifying deviations from expected behavior. The need for compact, cost-effective designs and operation in harsh conditions, such as wind, turbulence, and temperature variations, make sensors prone to noise and reduce measurement precision. Additionally, complex maneuvers cause rapid changes in orientation and acceleration, further challenging sensor accuracy and making it difficult to estimate the vehicle's state and parameters using these noisy measurements. Traditional filter error methods like Unscented and Extended Kalman Filters rely on linear approximations and Gaussian assumptions, which diminish their effectiveness in highly nonlinear and in the presence of non-Gaussian noise. To address these limitations, this paper proposes a Particle filter augmented with Recursive least square method for state and parameter estimation. This approach utilizes a set of samples (particles) to represent the probability distribution of the state and provides highly accurate and rapid state and parameter estimation for UAVs. Most parameters converging within 2 seconds and demonstrating significant improvements in root mean square error (RMSE) for state estimates. This swift and precise convergence enhances the effectiveness and robustness of reconfigurable control systems, ensuring optimal performance and stability in real-time operations.