<p dir="ltr">Methodologies for optimizing UAVs' control for varied environmental conditions have become crucial in the recent development for UAV control sector, yet they are lacking. This research focuses on the dynamism of the Gazebo simulator and PX4 Autopilot flight controller, frequently referenced in academic sectors for their versatility in generating close-to-reality digital environments. This thesis proposed an integrated simulation system that ensures realistic wind and gust interactions in the digital world and efficient data extraction by employing an industrial standard control communication protocol called MAVLink with the also the industry standard ground control software QGroundControl, using real and historical weather information from NOAA database. This study also looks into the potential of reinforcement learning, namely the DDPG algorithm, in determining optimal UAV safety distance, trajectory prediction, and mission planning under wind disruption. The overall goal is to enhance UAV stability and safety in various wind-disturbed conditions. Mainly focusing on minimizing potential collision risks in areas such as streets, valleys, tunnels, or really anywhere has winds and obstacles. The ROS network further enhanced these components, streamlining UAV response analysis in simulated conditions. This research presents a machine-learning approach to UAV flight safety and efficiency in dynamic environments by synthesizing an integrated simulation system with reinforcement learning. And the results model has a high accuracy, reaching 91%, 92%, and 97% accuracy on average in prediction of maximum shifting displacement, and left/right shifting displacement, when testing with real wind parameters from KLAF airport. </p>

<p>  </p> <p>Aircraft noise is one of the most significant environmental concerns for the aviation industry, and it adversely affects the physical and mental health of community members who are in close proximity to airports. The operations and expansion of airports and land use planning are affected because of the community’s adverse reaction to such annoyances. Aircraft operations and fleet mix information are required when airport managers and stakeholders execute the Aviation Environmental Design Tool (AEDT) to compute the noise metrics; however, these data are unavailable from over 2,000 United States non-primary General Aviation (GA) airports that lack full-time air traffic control facilities or personnel. </p> <p>This study developed a low-cost noise assessment technology for non-towered GA airports. The Automatic Dependent Surveillance-Broadcast (ADS-B) messages were obtained using an inexpensive ADS-B receiver. A barometric pressure calibration was applied to improve the aircraft operations estimation. A fleet mix database was created by linking the collected ADS-B data to an FAA-registered aircraft database containing U.S.-registered aircraft information (such as types of aircraft and engines). Specific aircraft information was obtained by filtering the International Civil Aviation Organization (ICAO) identification code from the obtained ADS-B records. A set of 20 advanced aircraft performance parameters was constructed to determine the operation mode and corresponding power setting. The corresponding noise levels were determined using the EUROCONTROL Aircraft Noise and Performance (ANP) database.</p> <p>The testing and validation results from the case study at the Purdue University Airport (ICAO Code: KLAF) demonstrated the developed low-cost approach could identify aircraft noise events, and the accuracy of modeled noise data was assessed with an average error of 4.50 dBA. Therefore, the developed approach appears to be an affordable means of monitoring aircraft noise at non-towered GA airports.  </p>

Performance models provided by aircraft manufacturers are used by aircraft operators to perform flight path simulations aiming to reduce aircraft fuel consumption. However, performance models are generic and does not account for the performance deviations of each aircraft individual. The performance deviations, particularly in terms of fuel consumption, will affect the dynamic programming of flight path simulations. This may result in a less optimal flight path and ultimately lead to higher fuel consumption than expected. In hope of reducing this risk, a collection of local performance factors were derived. These factors describe the percentual deviation between the real fuel flow and the levels predicted by the performance model, and are allocated with respect to a range of flight parameters in a data library known as the performance library. A test environment is then constructed to simulate a continuous flow of flight data, where a new performance library is derived from the flight data of every month. The local performance factors of the previous month are then updated with the current; a learning process based on the weighted average ensemble approach. Further, the local performance factors are used in conjunction with the performance model to estimate the aircraft fuel consumption during cruise. The observed average prediction error is noticeably smaller than that of an equivalent global, scalar performance factor used by airlines today. The result also reveals that the prediction accuracy and versatility of the performance library is mainly determined by its resolution - higher resolution generally offers better accuracy at a cost of requiring more flight data, whereas lower resolutions are more versatile but of lower accuracy. Finally, the performance libraries of two identical aircraft are used to trace the performance deviation between them. The weighted average of all local performance factors in the performance library of respective aircraft reveal that the average fuel consumption is roughly -1.9 % and -2.5 % lower than the estimates by the performance model, ultimately proving that it is feasible to detect overall fuel efficiency deviation between two identical aircraft. / Prestandamodeller tillhandhållna av flygplanstillverkarna används oftast av flygbolagen för att utföra flygruttsimuleringar i syfte att bespara bränsle. Dock är prestandamodellerna generiska och tillgodoräknar inte prestandaavvikelserna som förekommer hos varje flygplansindivid. Dessa prestandaavvikelser, speciellt i form av bränsleförbrukning, kommer att påverka den dynamiska programmeringen i flygruttssimulationen. Följde när flygrutter som kan leda till högre förbrukningar än de ursprungligen uppskattades. I hopp om att minimera denna risk beräknades mängder av lokala prestandafaktorer, vilka grundar på prestandamodellens avvikelse från verkliga flygdata. Dessa koefficienter allokerades sedan till ett databibliotek (prestandabibliotek) med avseende på en samling av flygparametrar. En testmiljö konstruerades i följd för att simulera ett kontinuerligt dataflöde. Vidare skapades ett prestandabibliotek för varje månadsflygdata, där de nyskapade lokala prestandafaktorerna viktas med de motsvarandeparterna i föregående månadens prestandabibliotek, vilket är en inlärningsprocessbaserad på viktad medelvärdesensemble. Prestandabiblioteket applicerades sedan över prestandamodellen och det snittliga uppskattning felet observerades vara märkbart mindre än det från en motsvarande global, skalärbaserad prestandafaktor. Resultatet antyder också på att prestandabibliotekets uppskattningsnoggrannhet och allsidighet beror huvudsakligen på dess upplösning - en hög upplösning leder generellt till ökad uppskattningsnoggrannhet med på bekostnad av mer flygdata, medan lägre upplösningar tenderar att vara mer allsidiga men med mindre uppskattningsnoggrannhet. Slutligen användes prestandabiblioteken av två identiska flygplan för att spåra prestandaavvikelser som förekommer mellan dem. Viktat medelvärde av alla prestandafaktorer i respektiveflygplanets prestandabibliotek tyder på att snittförbrukningen är ungefär 1,9 % respektive2,5 % lägre än det som uppskattades av prestandamodellen. Härmed bevisades att det är möjligt att spåra varianser i snittförbrukningen mellan två identiska flygplan.

<p dir="ltr">Cyber-physical-human systems (CPHSs) integrate human cognitive capabilities into the decision and control processes of complex dynamical systems. While artificial intelligence (AI) has shown promise in controlling such systems, it often encounters challenges such as conflict with human behavior and brittleness. Moreover, even successful AI implementations may lead to negative impacts on humans, such as the degradation of manual skills and diminished situation awareness, thereby weakening humans' ability to effectively monitor and intervene in off-nominal conditions as the final decision-makers of the systems. To address these unique challenges within CPHSs, this dissertation proposes three key approaches. First, human behavior modeling approaches are proposed to enhance understanding and prediction of human behavior from the perspective of AI. Accurate modeling enables better calibration of AI's expectations regarding human teammates' intentions and skill-levels. Second, a novel shared control approach is developed to expedite human training for complex dynamic control tasks. An assistant agent supports human novices in emulating human experts by leveraging human behavior models to gauge the human's skill-levels and provide tailored assistance to help improve one's skill. Lastly, human-autonomy teaming (HAT) design is addressed from a resource allocation perspective. A systematic computational simulation approach is proposed to optimize function and attention allocation to manage trade-offs in performance, situation awareness, workload, and other considerations. The proposed frameworks are demonstrated via examples in drone applications. Numerical and experimental results, utilizing simulation platforms and human subjects, validate the efficacy of the proposed approaches. This dissertation presents significant progress in the design and implementation of CPHSs in that it offers insights and methodologies to enhance collaborative interactions between humans and autonomous systems in complex environments.</p>