Search

Formation flight has long been performed by many species of birds for its social and aerodynamic benefits. The traditional "V" shape formation flown by birds not only helped communication between individuals, but also decreased the induced drag for each trailing bird, and thus reduced the energy required for flying (Weimerskirch et al. 2001). The benefits of formation flight have also been evaluated for manned aircraft. However, due to the high level of risk, human-piloted close formations are rarely sustained for a long enough time to fully appreciate the aerodynamic benefits. Therefore, reliable autonomous formation control can be an attractive capability for both human-piloted aircraft and Unmanned Aerial Vehicles (UAVs). The formation control problem has been extensively discussed in recent years with numerous applications with ground mobile robots, aircraft systems, and space vehicles. In their survey paper, (Scharf et al. 2004) classified the spacecraft formation flight control algorithms into five architectures: • “Multiple-Input Multiple-Output, in which the formation is treated as a single multipleinput, multiple-output plan; • Leader/Follower, in which individual spacecraft controllers are connected hierarchically; • Virtual Structure, in which spacecraft are treated as rigid bodies embedded in an overall virtual structure; • Cyclic, in which individual spacecraft controllers are connected nonhierarchically; • Behavioral, in which multiple controllers for achieving different (and possibly competing) objectives are combined.” Similar classifications can be extended to the formation control of other types of vehicles. However, due to the complexity and non-linearity associated with the aircraft dynamics, the ‘leader-follower’ approach was by far the most popular method for aircraft formation flight control. The advantage of the ‘leader-follower’ approach lies in its conceptual simplicity, where the formation flight problem is reduced to a set of tracking problems that can be analyzed and solved using standard control techniques. In the early 1990s, a series of publications from D’Azzo and his colleagues (Dargan et al. 1992) (Buzogany et al. 1993) (Reyna et al. 1994) (Veth et al. 1995) outlined the foundation for the control of the ‘leader-follower’ formation flight using compensation-type controllers.

Synthetic sensors enable flight data estimation without devoted physical sensors. Within modern digital avionics, synthetic sensors can be implemented and used for several purposes such as analytical redundancy or monitoring functions. The angle of attack, measured at air data system level, can be estimated using synthetic sensors exploiting several solutions, e.g., model-based, data-driven, and model-free state observers. In the class of data-driven observers, multilayer perceptron neural networks are widely used to approximate the input-output mapping angle-of-attack function. Dealing with experimental flight test data, the multilayer perceptron can provide reliable estimation even though some issues can arise from noisy, sparse, and unbalanced training domain. An alternative is offered by regularization networks, such as radial basis function, to cope with training domain based on real flight data. The present work’s objective is to evaluate performances of a single-layer feed-forward generalized radial basis function network for AoA estimation trained with a sequential algorithm. The proposed analysis is performed comparing results obtained using a multilayer perceptron network adopting the same training and validation data.

An important issue in robust adaptive control is the estimation of realistic ranges for the parametric uncertainty to be used in the projection mechanism with the goal of guaranteeing the boundedne...

A crucial aspect that could facilitate the applications of adaptive control systems in aerospace applications is the development of effective validation and verification procedures. Most of the exi...

Wide-field optical flow information can benefit the navigation of small or micro unmanned aerial vehicles in GPS-degraded/denied environments, inspired by the study of insect/bird flights. This paper focuses on a flight-test evaluation of navigation information in wide-field optical flow, using flight data collected by a wide-angle camera installed on a fixed-wing unmanned aerial vehicle at altitudes ranging from 30 to 183 m above the ground. A comprehensive evaluation methodology is proposed for the comparison between optical flows computed from videos and the reference motion fields determined from conventional navigation sensors including GPS, a laser range finder, and inertial sensors. Seven sets of unmanned aerial vehicle flight data including a total of approximately 72,000 image frames are used for both full-flight temporal evaluation and representative interframe spatial evaluation of wide-field optical flow information for unmanned aerial vehicle navigation purposes. The evaluation results showed...

This paper presents an analytical redundancy-based approach to the estimation of the true aircraft airspeed without the use of a Pitot probe. Measurements from an Inertial Measurement Unit (IMU), Global Positioning System (GPS), and angle of attack and sideslip angle vanes are used within a sensor fusion algorithm utilizing the kinematic model of a fixed-wing aircraft to co-estimate the airspeed, wind speed, and attitude of an aircraft. The presented model does not use information relative to the aircraft dynamic model, and therefore, is totally independent of the specific aircraft. Due to the necessity of the algorithm to determine the aircraft wind speed, a predictive model of the wind behaviour is necessary. This work compares two different stochastic wind models – the random walk (RW) and the Gauss–Markov (GM)–within the context of the airspeed estimation problem using flight data of a light aviation aircraft. The use of light aviation data is an important new consideration since previous work has focused on unmanned aircraft. The results of this study indicated that the RW model provides better performance over the GM model for airspeed estimation. Additionally, the light aviation data further reinforces the cross-platform capability of the considered algorithm and, furthermore, demonstrates the effectiveness of this method within manned flight.

Although very uncommon, the sequential failures of all aircraft Pitot tubes, with the consequent loss of signals for all the dynamic parameters from the Air Data System, have been found to be the cause of a number of catastrophic accidents in aviation history. This paper proposes a robust data-driven method to detect faulty measurements of aircraft airspeed, angle of attack, and angle of sideslip. This approach first consists in the appropriate selection of suitable sets of model regressors to be used as inputs of neural network-based estimators to be used online for failure detection. The setup of the proposed fault detection method is based on the statistical analysis of the residual signals in fault-free conditions, which, in turn, allows the tuning of a pair of floating limiter detectors that act as time-varying fault detection thresholds with the objective of reducing both the false alarm rate and the detection delay. The proposed approach has been validated using real flight data by injecting artificial ramp and hard failures on the above sensors. The results confirm the capabilities of the proposed scheme showing accurate detection with a desirable low level of false alarm when compared with an equivalent scheme with conventional “a priori set” fixed detection thresholds. The achieved performance improvement consists mainly in a substantial reduction of the detection time while keeping desirable low false alarm rates.

This paper presents a novel analytical redundancy approach for pitot tube failure accommodation using nonlinear Kalman filtering. This approach utilizes information from other sensors that are commonly implemented on aircraft in order to obtain an estimate of the airspeed which is independent from the pitot tube(s). This method was demonstrated to be independent of the aircraft model by providing experimental estimation results from two different unmanned aerial vehicle (UAV) research platforms.

The extended Kalman filter (EKF) and unscented Kalman filter (UKF) for nonlinear state estimation with both additive and nonadditive noise structures are presented and compared. Three different Global Positioning System (GPS)/inertial navigation system (INS) sensor fusion formulations for attitude estimation are used as case studies for the nonlinear state estimation problem. A diverse set of actual flight data collected from research unmanned aerial vehicles was used as empirical data for this study. Roll and pitch estimation results were comparedwith independent measurements from amechanical vertical gyroscope to evaluate the performance. The performance of the EKF and UKF is compared in terms of noise assumptions, covariance matrix tuning, sampling rate, initialization error, GPS outages, robustness to inertial measurement unit bias and scale factors, and linearization. Similar sensitivity for this GPS/INS attitude estimation problem was found between the EKF and UKF for most cases. Small differences were seen between EKF and UKF for initialization error and GPS outages: the UKF was found to be more robust to inertial measurement unit calibration errors, and the EKF was determined to be more computationally efficient.

Autonomous formation flight is a key approach for reducing energy cost and managing traffic in future high density airspace. The use of Unmanned Aerial Vehicles (UAVs) has allowed low-budget and low-risk validation of autonomous formation flight concepts. This paper discusses the implementation and flight testing of nonlinear dynamic inversion (NLDI) controllers for close formation flight (CFF) using two distinct UAV platforms: a set of fixed wing aircraft named “Phastball” and a set of quadrotors named “NEO.” Experimental results show that autonomous CFF with approximately 5-wingspan separation is achievable with a pair of low-cost unmanned Phastball research aircraft. Simulations of the quadrotor flight also validate the design of the NLDI controller for the NEO quadrotors.

Adaptive control-based schemes have been implemented to date in a variety of different applications. However, the ability to obtain a predictable transient closed-loop performance in adaptive systems is still a challenging problem from a verification and validation point of view. To face this problem we have recently introduced an analysis and design framework for adaptive control systems in the presence of bounded uncertainty and bounded adaptive control (the boundedness can be enforced, for instance, by a parameter projection mechanism) showing that the transitory performance of a MRAC system can be expressed, analyzed, and optimized via a convex optimization formulation based on Linear Matrix Inequality (LMI) requirements. A key feature of this framework is that it is possible to tune the adaptive control parameters rigorously so that the tracking error of the closed-loop system evolves within an a priori specified region of the error space whose size can be minimized by selecting a suitable cost function.

In this paper, several Global Positioning System/inertial navigation system (GPS/INS) algorithms are presented using both extended Kalman filter (EKF) and unscented Kalman filter (UKF), and evaluated with respect to performance and complexity. The contributions of this study are that attitude estimates are compared with independent measurements provided by a mechanical vertical gyroscope using 23 diverse sets of flight data, and that a fundamental difference between EKF and UKF with respect to linearization is evaluated.

This paper discusses the evaluation of a neurally augmented fault-tolerant flight control scheme for a high-performance military aircraft featuring an adaptive actuator and sensor failure detection, isolation, and identification algorithm in a motion-based flight simulator. The design of the fault-tolerant control scheme is based on a nonlinear dynamic inversion schemewithaneuralnetwork-basedaugmentationforreducingthedynamicinversionerrors associated with the occurrence of an actuator failure while a set of online learning neural observers is used for dealing with specific sensor failures. The failure detection, isolation, and identification scheme is based on an adaptive threshold technique for estimating failure bounds associated with both actuator and sensor failures. Also, an ‘ad-hoc’ parameter is proposed here for the novel task of evaluating the pilot workload in compensating for both actuatorandsensorfailuresonboardtheaircraft.Ageneraloutcomeoftheeffortisademonstration of the importance of realistic motion-based simulation environments for evaluation of this specific class of flight control laws. The study also demonstrated the importance of the neural augmentation for failure accommodation purposes and the effectiveness of the proposed adaptive threshold technique for failure identification purposes.

A systematic methodology is described for calculating structured residuals with high fault diagnostic capabilities for detecting sensor and actuators failures. The effort addresses implementation issues for real-time applications such as residual computation complexity and sensitivity to measurement noise. These specific requirements have been rigorously introduced through a cost function measuring the quality of the residual signal. A structural analysis approach of the nonlinear model of the system in conjunction with the unknown variables elimination method is used to derive subsets of residual equations. An algorithm is proposed for selecting the residual equations with maximum "failure isolability" and minimum cost, according to the selected performance criteria. The methodology has been applied to the design of a real-time residual generator for a nonlinear model of a remotely controlled semi-scale YF-22 research aircraft.

The design of a simulation environment is described for a machine vision (MV)-based approach for the problem of aerial refueling (AR) for unmanned aerial vehicles (UAVs) using the USAF refueling method. MV-based algorithms are implemented within the proposed scheme to detect the relative position and orientation between the UAV and the tanker. Within this effort, techniques and algorithms for the visualization the tanker aircraft in a virtual reality (VR) setting, for the acquisition of the tanker image, for the feature extraction (FE) from the acquired image, for the feature matching (FM) of the features, for the tanker-UAV pose estimation (PE) have been developed and extensively tested in closed-loop simulations. Detailed mathematical models of the tanker and UAV dynamics, refueling boom, turbulence, wind gusts, and tanker's wake effects, along with the UAV docking control laws have been implemented within the simulation environment. This paper also presents the results of a study relative to the use of passive markers versus feature extraction for the problem of estimating in real time the UAV-tanker relative position and orientation vectors.

Even small differences in atmospheric and/or flight conditions can potentially impact significantly the evaluation of the performance of the control laws and prevent a correct comparison, especially in the case of reduced size aircraft (autonomous or remotely piloted). Consistent deterministic control inputs can only be guaranteed through some form of computer-based on-board excitation system. In this paper, a methodology is proposed for flight data reduction with the purpose of accounting for non-homogeneous atmospheric conditions and inconsistent pilot inputs. The method is developed for the specific purpose of comparing model-following adaptive control laws. Performance evaluation parameters based on angular rate tracking errors are defined and used for the comparison. As a result of this approach, an additive correction is applied to the angular rate measurements to compensate for non-homogeneous turbulence effects. A multiplicative correction factor is applied to the angular rate tracking error to take into account non-identical pilot inputs. The procedure is validated with simulation and flight data obtained in the process of designing a set of fault tolerant control laws based on non-linear dynamic inversion with neural network augmentation for the reduced size WVU YF-22 aircraft model.

Birds fly in formation to benefit from the upwash wake vortices created by their peer leader. Similar strategies can be employed on aircraft for fuel saving or for range extension. However, further understanding of wake vortices in the near field and mid field is needed for close formation flight (< 10 wingspans). This paper introduces a new method for wake vortex detection using small UAV formation. A NonLinear Dynamic Inversion (NLDI) based formation control law is adapted to the WVU Phastball UAV platform. The designed controller showed good tracking performance during close formation flight (5-10 wingspans) with typical errors of less than 3 meters during straight legs. Then, multiple sensor measurements from leader-follower UAV formation flights were used for the wake vortex detection, including angle of attack, angle of sideslip, accelerometers, and gyros. The collected flight data showed the feasibility of using UAV formation for the wake encounter detection.

This paper focuses on the analysis of a scheme for sensor failure, detection, identification and accommodation (SFDIA) using experimental flight data of a research aircraft model. Recent technical literature has shown the advantages of time-varying estimators and/or approximators. Conventional approaches are based on different versions of observers and Kalman filters while more recent methods are based on different approximators based on neural networks (NNs). The approach proposed in the paper is based on the use of on-line learning nonlinear neural approximators. The characteristics of three different neural architectures were compared through different sensor failures. The first architecture is based on a multi layer perceptron (MLP) NN trained with the extended back propagation algorithm (EBPA). The second and third architectures are based on a radial basis function (RBF) NN trained with the minimal resource allocating network (MRAN) and extended-MRAN (EMRAN). The MRAN and EMRAN algorithms have recently been developed for RBF networks and have shown remarkable learning capabilities at a fraction of the memory requirements and computational effort typically associated with conventional RBF NNs. The experimental data for this study are flight data acquired from the flight-testing of a th semi-scale B777 research model designed, built, and flown at West Virginia University (WVU). Copyright © 2002 John Wiley & Sons, Ltd.

Wind and turbulence, including wakes induced by leading aircraft, have a large impact on flight performance and flight safety of both manned and unmanned aircraft. An accurate real-time wind estimation technique is crucial for tasks such as increasing air traffic capacity, commercial formation flight, or aerial refueling, etc. A leader-follower formation flight of Phastball Unmanned Aerial Vehicles (UAVs) were used as the experimental platform for the above problem. The air data system of Phastball UAV was developed with pitot-tube and flow-angle sensors. Using the designed system, two Unscented Kalman Filters (one standalone UKF and one cooperative UKF) were developed for the wind field estimation with and without using the wake information from the leader aircraft. For close formation flights, the wake of the leader is assumed to be predictable by certain wake models for the follower aircraft. Flight data showed the effectiveness of the standalone EKF for the wind estimation compared with the ground weather station measurements. Simulation results showed the advantage of the cooperative UKF over the standalone UKF.

Close Formation Flight is a key potential approach for reducing greenhouse gas emissions and managing traffic in future high density airspace. This paper discusses the implementation and flight testing of a formation flight controller. Experimental results show that an autonomous close formation flight with approximately 5-wingspan separation is achievable with a pair of low-cost unmanned research aircraft.

This paper describes the results of a study where an on-line parameter identification (PID) technique is used for determining on-line the mathematical model of an aircraft that has sustained damage to a primary control surface. The mathematical model at post-failure conditions can then be used by a failure accommodation scheme to compute on-line the compensating control signal to command the remaining healthy control surfaces for a safe continuation and/or termination of the flight. Specific criteria for the use of an on-line PID for these critical flight conditions are first discussed. The methodology is illustrated through simulations of a fighter jet at subsonic flight conditions featuring a novel modeling procedure to characterize the post-failure/damage aerodynamic conditions. The simulations have shown the potential of this on-line PID within a fault tolerant flight control system. The results have also highlighted the importance of conducting an ‘ad hoc’ small amplitude and short-duration PID maneuver immediately following a positive failure detection to enhance the reliability of the on-line estimated parameters used in the accommodation scheme.

In recent years, due to the globally increasing trend in air traffic volume, the aviation community has been touched by the occurrence of a number of crashes, although the overall aviation safety is actually improving in most countries. In the US the National Transportation and Safety Board (NTSB) begins its investigation by analyzing the wreckage along with the information from flight data recorder (FDR) and cockpit voice recorder (CVR). In most instances this set of information is enough for the NTSB to discover the cause of the crash; unfortunately, this is not always the case. Until a few years ago FAA regulations mandated the recording of 11–17 flight parameters without specifying the recording of the deflection of primary control surfaces. Following a few accidents where control surface failures were believed to be a likely cause of the crash, the FAA recently required the US-based airlines to retrofit the fleet with newer digital FDRs capable of recording a much larger number of parameters, including, of course, the deflection of primary control surfaces. This rule has a multi-year compliance period. However, some airlines are or have been seeking exemptions from this rule for some specific aircraft soon to be retired from service. Furthermore, only the US commercial fleet is affected by this ruling. Therefore, there is a need for a scheme that can reconstruct additional aircraft time histories to aid investigators for crashes with limited CVR information and where control surface failure is believed to be a factor. This paper describes a scheme formulated to reconstruct the aircraft primary surface deflection using data available from the current FDRs recording only 11–17 parameters. The scheme consists of two neural networks. The first is used to simulate the aircraft dynamics, while the second is used to reconstruct the primary surface deflections. The methodology is applied to simulated maneuvers from the non-linear model of an F-16 from a commercially available flight simulation software.

In recent years neural networks have been proposed for identification and control of linear and non-linear dynamic systems. This paper describes the performance of a neural network-based fault-tolerant system within a flight control system. This fault-tolerant flight control system integrates sensor and actuator failure detection, identification, and accommodation (SFDIA and AFDIA). The first task is achieved by incorporating a main neural network (MNN) and a set of n decentralized neural networks (DNNs) to create a system with n sensors which has the ability to detect a wide variety of sensor failures. The second scheme implements the same main neural network integrated with three neural network controllers. The contribution of this paper focuses on enhancements of the SFDIA scheme to allow the handling of soft failures as well as addressing the issue of integrating the SFDIA and the AFDIA schemes without degradation of performance in terms of false alarm rates and incorrect failure identification. The results of the simulation with different actuator and sensor failures with a non-linear aircraft model are presented and discussed.

The results are presented of a comparative study evaluating the performance of neural network (NN) and fuzzy logic reconstructors (FLRs) for the development of a virtual flight data recorder (VFDK). Typical flight data recorders (FDRS) on commercial airliners do not record the aircraft control surface deflections. These dynamic parameters are critical in the investigation of an accident or an uncommanded maneuver. The results are shown relative to a VFDR based on a neural network simulator (NNS) along with a neural network reconstructor (NNR) or a FLR The NNS is trained off-line, using available flight data for the particular aircraft, for the purpose of simulating any desired dynamic output recorded in current FDRs. The NNS is then interfaced with the NNR or with the FLR. The output of the two reconstructors are the control surface deflections which minimize a performance index based on the differences between the available data from the FDR and the output from the NNS. The study tested with night data from a B737-300 shows that both schemes, the one with the NNR and the one with the FLR, provide accurate reconstructions of the control surface deflections time histories.

This paper presents preliminary results of the development of a virtual flight data recorder (VFDR) for commercial airliners. Federal Aviation Administration (FAA) regulations, currently being revised, mandate the recording of 11 dynamic parameters, not including the control surface deflections. The absence of these data can be critical for crash investigation purposes. This paper proposes the introduction of a VFDR based on a neural network simulator (NSS) and a neural network reconstructor (NNR). The NNS is trained, using flight data for the particular aircraft, to simulate any desired control surface deflections (or any other parameters of interest not recorded by the FDR), minimizing a cost function based on the differences between the available data from the FDR and the output from the NNS. The VFDR scheme has been introduced, tested, and validated with flight data from a Boeing 737-300 with an FDR with extended recording capabilities showing accurate reconstruction of the control surface deflections’ time histories. The VFDR can be considered a tool for crash investigations where control surface failures are believed to be a factor.

This paper proposes the implementation of on-line learning neural controllers in the autopilot control laws of a modern high-performance aircraft. A first advantage of this design philosophy consists in avoiding the precomputation, storing, and interpolation between thousands of feedback gains of a typical flight control system. Another advantage is the ability to compensate for nonlinearities and model uncertainties. In addition, an on-line learning time-varying neural architecture will avoid the time-consuming gain recalculation following any modification to the aircraft or to its control system during its operative life. The implementation of these types of alternative controllers is made possible by the recent simultaneous advances in neural-network technology with regard to the availability of efficient and fast learning algorithms, along with progress in digital microprocessors. The approach is shown using a six-degree-of-freedom nonlinear simulation code. The traditional gain-scheduling-based control laws for typical autopilot functions are replaced by on-line learning neural architectures trained with the extended back-propagation algorithm. This algorithm has shown substantial improvements over the conventional back-propagation method in learning speed and accuracy. The results of the simulation with and without a man in the loop are presented and discussed.

This paper describes an innovative approach to the problem of the on-line determination of a control law in order to achieve a dynamic reconfiguration of an aircraft that has sustained extensive damage to a vital control surface. The approach consists of the use of on-line learning neural network controllers that have the capability of bringing an aircraft, whose dynamics can become unstable after a substantial damage, back to an equilibrium condition. This goal has been achieved through the use of a specific training algorithm, the extended back-propagation algorithm (EBPA), and proper selection of the architectures for the neural network controllers. The EBPA has recently shown remarkable improvements over the back-propagation algorithm in terms of convergence time and local minimum problems. The methodology is illustrated through a nonlinear dynamic simulation of a typical combat maneuver for a high-performance aircraft.

This paper presents the modeling of the aerodynamic and thrust forces for the West Virginia University (WVU) Propulsion Assisted Control (PAC) unmanned research aircraft. Specifically, a state space model of the aircraft was estimated from flight data, and the stability and control derivatives were extracted and optimized to determine the aerodynamic forces and moments. Dynamic thrust was modeled using wind tunnel experiments to evaluate its dependencies upon wind speed, RPM, and angle of attack. This was used to incorporate differential thrust into the aircraft model. The effects of differential thrust were analyzed through flight tests, where prolonged failures required active compensation from the pilot to keep the aircraft along its original flight path. This suggests that successful implementation of differential thrust could lead to an extra degree of freedom in compensating for lateral-directional failures within a fault tolerant flight control system.

In this paper, a neural network is proposed as an approach to the task of failure detection following damage to an aerodynamic surface of an aircraft flight control system. Several drawbacks of other failure detection techniques can be avoided by taking advantage of the flexible learning and generalization capabilities of a neural network. This structure, used for state estimation purposes, can be designed and trained on line in flight and generates a residual signal indicating the damage as soon as it occurs. From an analysis of the cross-correlation functions between some key state variables, the identification of the damage type can also be achieved. The results of a nonlinear numerical simulation for a damaged control surface are reported and discussed.

A particular algorithm is applied to the aircraft flight control reconfiguration problem. The determination of the desired control law, which can adapt in a very short period of time to major damage of a control surface, is obtained by making use of the recent control and response time histories. The estimated model of the damaged aircraft used in this technique is obtained by using a multiple model Kalman filtering approach. The model estimation and the control algorithm have been codified in a computer simulation program for a six degree-of-freedom aircraft model. The simulation results of the reconfiguration are presented.

This paper describes the design of a simulation environment for a GPS / Machine Vision (MV)-based approach for the problem of Aerial Refueling (AR) for Unmanned Aerial Vehicles (UAVs) using the USAF refueling method. MV-based algorithms are implemented within this effort as smart sensor in order to detect the relative position and orientation between the UAV and the tanker. Within this effort, techniques and algorithms for the visualization the tanker aircraft in a Virtual Reality (VR) setting, for the acquisition of the tanker image, for the Feature Extraction (FE) from the acquired image, for the Point Matching (PM) of the features, for the tanker-UAV Pose Estimation (PE) have been developed and extensively tested in closed loop simulations. Detailed mathematical models of the tanker and UAV dynamics, refueling boom, turbulence, wind gusts, and tanker’s wake effects, along with the UAV docking control laws and reference path generation have been implemented within the simulation environment. Mathematical model of the noise produced by GPS, MV, INS and pressure sensors are also derived. This paper also presents an Extended Kalman Filter (EKF) used for the sensors fusion between GPS and MV systems. Results on the accuracy reached for the estimation of the relative position are also provided.

The use of a combined Machine Vision (MV) and GPS-based approach has been recently proposed in simulation efforts as an alternative approach to ‘pure GPS’ for the problem of Autonomous Aerial Refueling (AAR) for Unmanned Aerial Vehicles (UAVs). While MV has appealing capabilities, a few critical issues need to be addressed for the actual implementation of MV for the AAR problem. For this purpose a simulation environment was developed featuring an interaction with a 3D Virtual Reality (VR) interface that generates an image stream of the AAR maneuver. The image flow is processed by the MV algorithm, providing, as output, a vector of the estimates of the relative tanker-UAV distance and attitude. This signal is then used by the UAV feedback control laws for ‘tracking & docking’ to the refueling boom. The MV algorithm specifically provides image processing for the isolation of optical markers, which are located at specific points on the tanker, extraction of the marker center of gravity, marker matching algorithm, and pose estimation algorithm for the final evaluation of the relative distance vector. Within this effort emphasis was placed on the development of an ‘ad-hoc’ feature matching algorithm followed by a comparative analysis of the performance of different matching algorithms. The paper presents a detailed analysis of the results from open loop and closed loop simulation of the different MV algorithms.

This paper describes the results of an effort on the analysis of the performance of specific ‘pose estimation’ algorithms within a Machine Vision-based approach for the problem of aerial refuelling for unmanned aerial vehicles. The approach assumes the availability of a camera on the unmanned aircraft for acquiring images of the refuelling tanker; also, it assumes that a number of active or passive light sources – the ‘markers’ – are installed at specific known locations on the tanker. A sequence of machine vision algorithms on the on-board computer of the unmanned aircraft is tasked with the processing of the images of the tanker. Specifically, detection and labeling algorithms are used to detect and identify the markers and a ‘pose estimation’ algorithm is used to estimate the relative position and orientation between the two aircraft.Detailed closed-loop simulation studies have been performed to compare the performance of two ‘pose estimation’ algorithms within a simulation environment that was specifically developed for the study of aerial refuelling problems. Special emphasis is placed on the analysis of the required computational effort as well as on the accuracy and the error propagation characteristics of the two methods. The general trade offs involved in the selection of the pose estimation algorithm are discussed. Finally, simulation results are presented and analysed.

This paper will present the results obtained from a formation flight experiment using 3 YF-22 research UAVs built and developed at West Virginia University. In the planned 3-aircraft flight configuration, a Radio Control (R/C) pilot maintains manual control of the 'leader' aircraft while two autonomous 'follower' aircraft maintain a pre-defined position and orientation with respect to the lead vehicle. The flight-testing program associated with this effort was quite substantial and involved approximately 100 sorties involving all three aircraft models in the program. Initial flight-testing phases evaluated a variety of issues, including; aircraft handling qualities, dynamic characteristics, electronic payload performance; specifically the data acquiring capabilities and the real time execution of control laws. During the final flight season program, a total of five formation flight experiments were successfully performed; specifically four 2-aircraft formations and one 3-aircraft formation flight demonstration.

This paper describes the hardware development for an autonomous formation flight research project at West Virginia University. Each aircraft test bed was outfitted with an Inertial Navigation System (INS), GPS receiver and telemetry system tailored to perform the formation flight experiment, This paper provides a detailed summary about the aircraft test-bed, on-board electronic payload, and the various flight testing phases leading to the final 3-aircraft formation demonstration.

Two aspects of formation flight are addressed in this paper: dynamic modeling and formation control. In formation flight aircraft dynamics are coupled by aerodynamic effects due to the vortices leaving the lifting surfaces, such as changes in lift and drag forces and lateral/directional effects that do not appear in a steady-level ‘isolated’ flight. These aerodynamic effects are properly modeled with a three dimensional code based on a Distributed Horse-Shoe Technique. A formation controller allowing both trajectory tracking and formation geometry keeping is then designed. It is shown that the designed controller yields satisfactory performance in a two-aircraft formation.

This paper presents identification, control synthesis and simulation results for an YF-22 aircraft model designed, built, and instrumented at West Virginia University. The ultimate goal of the project is the experimental demonstration of formation flight for a set of 3 of the above models. In the planned flight configuration, a pilot on the ground maintains controls of the leader aircraft while a wingman aircraft is required to maintain a pre-defined position and orientation with respect to the leader. The identification of both a linear model and a nonlinear model of the aircraft from flight data is discussed first. Then, the design of the control scheme is presented and discussed with an emphasis on the amount of information, relative to the leader aircraft, needed by the wingman to maintain formation. Using the developed nonlinear model, the control laws for a maneuvered flight of the formation are then simulated with Simulink® and displayed with the Virtual Reality Toolbox®. Simulation studies have been performed to evaluate the effects of specific parameters and the system robustness to atmospheric turbulence. The conclusions from this analysis have allowed the formulation of specific guidelines for the design of the electronic payload for formation flight.

A critical limitation for the current use of Unmanned Aerial Vehicles (UAVs) is represented by the lack of aerial refueling capabilities. This paper describes the results of an effort on the modeling of the UAV aerial refueling problem and on the design of the docking control scheme. The control of the docking maneuver is based on a fuzzy sensor fusion strategy featuring GPS and Machine Vision (MV) data. The design for a smooth docking maneuver under the presence of wake effects is performed; desirable performances were achieved with LQR-based control laws for the docking of the UAV to the probe–drogue refueling system. Simulations of the proposed docking scheme are presented and discussed.

For a research aircraft, "conventional" control laws (CLs) are implemented on a "baseline" flight computer (FC) while research CLs are typically housed on a dedicated research computer. Therefore, for an experimental aircraft used to test specific fault tolerant flight control systems, a safety logic scheme is needed to ensure a safe transition from conventional to research CLs (while at nominal conditions) as well as from research CLs at nominal conditions to conditions with "simulated" failures on specific control surfaces. This paper describes the design of such a safety scheme for the NASA Intelligent Flight Control System (IFCS) F-15 Program. The goals of the IFCS F-15 program are to investigate the performance of a set of fault tolerant CLs based on the use of dynamic inversion with neural augmentation. The different transitions are monitored using information relative to flight conditions and controller-related performance criteria. The testing of the scheme is performed with a Simulink-based flight simulation code and interface developed at West Virginia University for the NASA IFCS F-15 aircraft.

The objective of this document Is to show the capabilities of parallel hardware-based on-line learning neural networks (NNs). This specific application is related to an on-line estimation problem for sensor validation purposes. Neural-network-based microprocessors are starting to be commercially available. However, most of them feature a learning performed with the classic back-propagation algorithm (BPA). To overcome this lack of flexibility a customized motherboard with transputers was implemented for this investigation, The extended BPA (EBPA), a modified and more effective BPA, was used for the on-line learning, These parallel hardware-based neural architectures were used to implement a sensor failure detection, identification, and accommodation scheme in the model of a night control system assumed to be without physical redundancy in the sensory capabilities. The results of this study demonstrate the potential for these neural schemes for implementation in actual flight control systems of modern high performance aircraft, taking advantage of the characteristics of the extended back-propagation along with the parallel computation capabilities of NN customized hardware.

This paper presents the results of an investigation focused on parameter identification for the NASA F/A-18 HARV. This aircraft was used in the high alpha research program at NASA/Dryden. In this study the longitudinal stability derivatives are estimated from flight data using the maximum likelihood method coupled with a Newton-Raphson minimization technique. The objective is to estimate an aerodynamic model describing the aircraft dynamics over a range of angle of attack from 5 to 60 deg. The mathematical model is built using the traditional static and dynamic derivative buildup. Flight data used in this analysis were from a variety of maneuvers, including large amplitude multiple doublets, optimal inputs, frequency sweeps, and pilot pitch stick inputs. The parameter estimation code pEst, developed at NASA/Dryden, was used in this investigation. Results of the estimation process from alpha = 5 to 60 deg are presented and discussed. (Author)

This paper presents the results of an investigation related to the estimation of the lateral-directional aerodynamic characteristics of the NASA F/A-18 HARV from flight data. This aircraft was used in the high alpha research program at NASA/Dryden. In this study the lateral aerodynamic and control derivatives are estimated from flight data using the maximum likelihood method along with a Newton-Raphson minimization technique. The mathematical model is based on a classical static and dynamic derivative buildup. Flight data were collected over a range of 5-60-deg angle of attack from a variety of maneuvers, including large amplitude multiple doublets, optimal inputs, and pilot stick and rudder inputs. The parameter estimation code pEst, developed at NASA/Dryden, was used in this work. Results of the parameter estimation process from alpha = 5 to 60 deg are presented and discussed. (Author)

The subject of this paper is the determination of the aerodynamic stability and control derivatives from flight data using the Maximum Likelihood method for the NASA FIA-18 HARV at high angles of attack. The parameter identification (PID) code pEst, developed at NASA Dryden, was provided by NASA and modified for an alternative modeling approach for high angle of attack conditions. Estimates were obtained for longitudinal dynamics parameters.

This paper presents a neural-network-based approach for the problem of sensor failure detection, identification, and accommodation for a flight control system without physical redundancy in the sensors. The approach is based on the introduction of on-line learning neural network estimators. For a system with n sensors, a combination of a main neural network and a set of n decentralized neural networks achieves the design goal. The main neural network and the ith decentralized neural network detect and identify a failure of the ith sensor, whereas the output of the ith decentralized neural network accommodates for the failure by replacing the signal from the failed ith sensor with its estimate. The on-line learning for these neural network architectures is performed using the extended back-propagation algorithm. The document describes successful simulations of the sensor failure detection, identification, and accommodation process following both soft and hard sensor failures. The simulations have shown remarkable capabilities for this neural scheme.