Autopilot design for fixed wing aircraft under colored noise

Abstract

Flight safety and performance of aerial vehicles are among the top priorities in the modern aviation industry. Accurately determining the vehicle's states during flight is vital for ensuring a safe and stable operation. To this end, state estimation algorithms are employed, aiming to provide the best possible prediction of the vehicle's true state by accounting for various disruptive effects in sensor data, such as noise and wind disturbances. Autopilot systems are advanced control mechanisms that enable an aircraft to autonomously fly along predefined routes and at specific speeds with high accuracy. For these systems to operate effectively, continuous and precise access to the vehicle's state information—such as position, velocity, and orientation—is essential. Therefore, state estimation algorithms become a fundamental component of the autopilot. Advanced algorithms like Kalman filters help reduce noise in sensor data during this process, allowing for more reliable and accurate state predictions. The Kalman filter provides highly successful results, especially when the system model is accurate and the statistical properties of the noise are well known. However, its performance can be negatively affected in the presence of noise types known as 'colored' noise. In such cases, the filter makes assumptions that do not align with the actual noise structure, which can lead to increased estimation errors. Therefore, to effectively handle the complex and diverse noise types encountered in real flight conditions, additional adaptation techniques or alternative filtering methods may be employed. For instance, different variants of the Kalman filter or multivariate statistical approaches can be used to model or mitigate the effects of colored noise. Through these advanced algorithms, parameters such as position, velocity, and orientation of the aircraft can be estimated more reliably and accurately, thereby enhancing flight safety and performance. In this thesis study, an autopilot control system for the longitudinal motion of a fixed-wing aircraft has been developed, along with the design of a filter capable of performing effectively under colored noise conditions. First, a detailed model of the selected aircraft was developed in Simulink. The equations derived from the created model were linearized at the specified trim point, and subsequently, the state-space matrices of the linearized system were obtained. After comparing the nonlinear and linear models and confirming the consistency of the results, the control phase was initiated. Due to the system's MIMO nature, the LQR control method was chosen, and an appropriate feedback structure was established. The gain matrix required for LQR control was obtained by selecting suitable weighting matrices, and the necessary gain values were incorporated into the feedback loop. Finally, filter designs were carried out for state estimation, and the filter block was integrated into the model. Eight different simulation scenarios were defined for the completed model. White and colored noises were applied to the system outputs, and various tests were conducted using both the Kalman filter and the designed filter under these noise conditions. Additionally, the controller performance was evaluated with different reference inputs. As a result of the simulations, the filters' performance under various noise influences was observed and compared. The performance of the autopilot designed for longitudinal motion was examined, and the results across different scenarios were analyzed. The obtained results and comprehensive studies demonstrate that advanced filtering techniques, which provide high accuracy and reliability even in colored noise environments, can significantly enhance aircraft performance and flight safety.