Hakki Erhan Sevil

Estimation of Receiver Aircraft States and Wind Vectors in Aerial Refueling

A square root unscented Kalman filter is used to estimate the states of a receiver aircraft and the components of the wind the aircraft is exposed to as it flies behind the tanker aircraft. The wind is the superposition of time-varying prevailing wind, turbulence, and the tanker’s wake-induced wind. The system update of the estimator uses the nonlinear aircraft model, augmented by a nonlinear wind model, obtained from the translational dynamics of the aircraft. The aircraft state estimates are successfully used in controlling the position of the receiver aircraft relative to the tanker flying straight level and making turns. Although the performance of the estimation and control is very dependent on the covariance and correlation time constants of the sensor measurement errors, the estimated state feedback performs better than measured state feedback with higher levels of measurement noise and turbulence intensity.

Fault diagnosis in air data sensors for receiver aircraft in aerial refueling

Multiple air-data sensors placed on different positions on an aircraft are usually expected to provide identical measurements. This is because aircraft are assumed to fly in no-wind or uniform wind conditions. There are cases, however, where an aircraft flies in nonuniform wind vector field with significant variations in magnitude and direction over the dimensions of the aircraft. For example, in an aerial refueling operation, the receiver aircraft needs to fly within the nonuniform wind field induced by the wake of the tanker aircraft. This paper shows that an air-data sensor fault detection and isolation algorithm, if developed with an underlying assumption of identical sensor measurements, yields very high false detection rate when the aircraft enters the wake of another aircraft. This paper also presents a novel modification to the redundant-sensor-based fault detection and isolation algorithm that eliminates this problem. The new method is based on a model of the wind field as a function of relative ...

Construction of an obstacle map and its realtime implementation on an Unmanned Ground Vehicle

This paper presents the development of an obstacle mapping system based on the concept of a Probabilistic Threat Exposure Map (PTEM). The paper also discusses the realtime embedded implementation of this obstacle mapping system on a small Unmanned Ground Vehicle (UGV) to support realtime obstacle avoidance. These activities are a part of a larger effort to establish a theoretical foundation for autonomous and cooperative multi-UxV guidance solutions in adversarial environments.

False Fault Detection in Airdata Sensor due to Nonuniform Wind in Aerial Refueling

This paper presents a redundant-sensor-based Fault Detection and Isolation (FDI) algorithm, applied to a receiver aircraft in aerial refueling operation. The FDI algorithm is based on a statistical approach and uses measurements from three identical airdata sensors, placed at dierent locations. The airdata sensors measure airspeed, angle-of-attack and side slip angle. Simulation results show that the FDI algorithm successfully performs detection and isolation of sensor faults when the receiver aircraft ies solo or outside the wake of the tanker aircraft. However, simulation experiments also show that the FDI algorithm yields false fault detection when the receiver aircraft ies within the nonuniform wind eld, induced by the wake vortices of the tanker.

Trait-based heterogeneous populations plus (TbHP+ ) genetic algorithm

This study developed a variant of genetic algorithm (GA) model called the trait-based heterogeneous populations plus (TbHP+). The developed TbHP+ model employs a memory concept in the form of immunity and instinct to provide the populations with a more efficient guidance. Also, it has an ability to vary the number of individuals during the search process, thus allowing an automatic determination of the size of the population based on the individual qualities such as character fitness and credit for immunity. The algorithm was tested against the classical GA model in convergence and minimum error performance. For this purpose, 5 different mathematical functions from the literature were employed. The selected functions have different topological characteristics, ranging from simple convex curves with 2 variables to complex trigonometric ones having several hilly shapes with more than 2 variables. The developed model and the classical GA model were applied to finding the global minima of the functions. The comparison of the results revealed that the developed TbHP+ model outperformed the classical GA in faster convergence and minimum errors, which may be explained by the adaptive nature of the new paradigm.

Evaluation of extant computer vision techniques for detecting intruder sUAS

In this study, we investigate the feasibility of detecting small intruder aircraft through camera images obtained onboard a small unmanned aircraft. The research group (Small Unmanned Aerial Vehicle Laboratory) from NASA Langley Research Center flew a set of missions with their small UAS (sUAS) where one of those vehicles is outfitted with three 4K resolution cameras located at the tips of the wings and one at the nose. We utilize the MathWorks Computer Vision System Toolbox components to process the video data that are provided by NASA. We demonstrate the capabilities of COTS (Commercial Off-The-Shelf) state-of-art algorithms to detect the intruder aircraft in the video files. In the evaluation of these algorithms, various parameters of each algorithm are tuned to improve the detection performance in the case of the NASA flights, and the results are presented. The aim is to analyze performance of existing COTS state-of-art algorithms in detecting intruder aircraft from the camera images.

Modeling of an unmanned ground vehicle for autonomous navigation and obstacle avoidance simulations

The aim of this effort is to develop a model of an actual unmanned ground vehicle system for computer simulations in order to evaluate guidance algorithms developed for autonomous waypoint navigation and obstacle avoidance. Simulation is a vital tool for the development of autonomous systems. Simulating individual parts and units of the system can help identify flaws in its design or implementation. In the Matlab-Simulink environment, a kinematic based model of an skid-steer ground vehicle is designed. Furthermore, a model of quadrature encoders for position estimation, and a laser range finder (LRF) sensor model for obstacle detection are also created. Two different groups of experiments are performed to test the performance of the proposed models. Experimental results indicate that the models can adequately simulate the actual vehicle behaviors. This effort is part of an ongoing research to create fully autonomous UxVs capable of waypoint navigation and obstacle avoidance.Copyright

Real-Time Obstacle Avoidance and Waypoint Navigation of an Unmanned Ground Vehicle

Real-time obstacle avoidance and navigation are key fields of research in the area of autonomous vehicles. The primary requirements of autonomy are to detect or sense changes and react to them without human intervention in a safe and efficient manner. The objective of this research is to develop autonomous way-point navigation and obstacle avoidance capabilities for an unmanned ground vehicle (UGV). This research consists of developing and implementing an environment mapping system capable of detecting and localizing potential obstacles using real-time sensor data. The real-time obstacle mapping system developed in this work automatically generates the Probabilistic Threat Exposure Map (PTEM). The PTEM construction algorithm successfully constructs a probabilistic obstacle map both in simulation and real-time. Autonomous waypoint navigation is also achieved for both simulation and real-time platforms. These activities are a part of a larger effort to establish a theoretical foundation and real-time implementation of autonomous and cooperative multi-UxV guidance solutions in adversarial environments.Copyright

Cost effective localization in distributed sensory networks

The most important mechanism to occur in biological distributed sensory networks (DSNs) is called lateral inhibition, (LI). LI relies on one simple principle. Each sensor strives to suppress its neighbors in proportion to its own excitation. In this study, LI mechanism is exploited to localize the unknown position of a light source that illuminated the photosensitive sensory network containing high and low quality sensors. Each photosensitive sensor was then calibrated to accurately read the distance to the light source. A series of experiments were conducted employing both quality sensors. Low quality array was allowed to take advantage of LI, whereas the high quality one was not. Results showed that the lateral inhibition mechanism increased the sensitivity of inferior quality sensors, giving the ability to make the localization as sensitive as high quality sensors do. This suggests that the networks with multitude of sensors could be made cost-effective, were these sensory networks equipped with LI.

Determining intruder aircraft position using series of stereoscopic 2-D images

The aim of this study is to investigate methods for computing the position of an intruder aircraft relative to an observer aircraft with onboard stereo cameras. To focus on relative position estimation rather than the intruder aircraft detection through image processing, the first phase is to generate camera images given the relative position information. This process uses a simple pinhole camera method where cameras are characterized by focal length, angle of view, and resolution. The second phase is to develop two methods to estimate the relative position based on the generated camera images. Both methods employ epipolar geometry of stereo vision based on two cameras placed on the aircraft with lateral separation. Various cases are run in a Matlab/Simulink simulation environment. Simulation cases are designed to evaluate the relative position estimation methods with different aircraft trajectories, different camera separations, and different camera resolutions. Simulation results show that relative position can be estimated while both aircraft are flying along any trajectories as long as the intruder aircraft is visible by both cameras. The estimation accuracy degrades as the relative distance between the aircraft increases. The larger lateral separation seems to improve the estimation accuracy. Image resolution seems to have little to no impact on estimation accuracy.