Award

Modern aircraft require a complex set of systems, sensors, and hardware, all working in concert to ensure the successful completion of its mission and the safe return of its crew / passengers. To do this, pilots and flight crews rely on accurate knowledge of the vehicle’s physical configuration and dynamic states via sensor measurements, enabling aircrews to make critical flight safety and mission decisions. However, sensor faults can lead to incorrect or conflicting information, reducing situational awareness (SA) and potentially leading to accidents. Onboard instrumentation failures are a relatively common cause of aborted approaches and in the presence of multiple inconsistent instrument data, whatever the cause, the pilot may not be able to determine the correct state of the aircraft. Beyond sensor failures, information available from the vehicle management system (VMS) such as pilot entered aircraft weight, are of limited use in the aircraft due to potential errors in pilot entry without a systematic backup or check to this value. Therefore, a real-time in-flight robust algorithmic estimation of aircraft states and configuration from existing sensor data is required to reduce pilot burden and reduce vehicle sensor redundancy requirements.Estimating the true state of a complex dynamical system from noisy observations in real-time is one of the most fundamental tasks in signal processing and control. The low-complexity implementation of the Kalman filter (KF), combined with its sound theoretical basis, makes it the standard state estimation algorithm. Conventional KF methods are all model-based however, and thus whenever the dynamical system cannot be accurately characterized as a computationally accurate state-space model, degradation in performance of the state estimation occurs.  Data-driven alternatives have been proposed to capture the subtleties of complex processes and replace the need to explicitly characterize the system dynamics. However, these alternative methods do not incorporate domain knowledge such as structured state-space models in a principled manner, and thus require many trainable parameters and large data sets even for simple sequences and lack the interpretability of model-based methods. To enable a robust estimate of aircraft states using in-flight sensors and address these issues, a team led by STI proposes to develop Advanced Real-Time Efficient Aircraft State Estimation (ARTEASE). The software toolset will leverage a novel Data-Augmented State Estimation (DASE) algorithm in development at Pennsylvania State University (PSU) that uses a combined approach of a physics-based model along with efficient data-driven components to enhance the quantitative accuracy of the estimation. DASE will provide an estimate of aircraft configuration parameters such as gross weight, center of gravity, etc., as well as traditional dynamic aircraft states including attitude, rate, etc.