Learn-to-Fly (L2F) is an advanced technology development effort aimed at assessing the feasibility of real-time, self-learning flight vehicles. Specifically, research has been conducted on merging real-time aerodynamic modeling, learning adaptive control, and other disciplines with the goal of using this “learn to fly” methodology to replace the current iterative vehicle development paradigm, substantially reducing the typical ground and flight testing requirements for air vehicle design. Recent activities included an aggressive flight test program with unique fully autonomous fight test vehicles to rapidly advance L2F technology. This paper presents an overview of the project and key components.

Rapid modeling and efficient testing methods are important in a number of aerospace applications. In this study efficient testing strategies were evaluated in a wind tunnel test environment and combined to suggest a promising approach for both ground-based and flight-based experiments. Benefits of using Design of Experiment techniques, well established in scientific, military, and manufacturing applications are evaluated in combination with newly developing methods for global nonlinear modeling. The nonlinear modeling methods, referred to as Learn-to-Fly methods, utilize fuzzy logic and multivariate orthogonal function techniques that have been successfully demonstrated in flight test. The blended approach presented has a focus on experiment design and identifies a sequential testing process with clearly defined completion metrics that produce increased testing efficiency.

Flight test and modeling techniques were developed to accurately identify global nonlinear aerodynamic models onboard an aircraft. The techniques were developed and demonstrated during piloted flight testing of an Aermacchi MB-326M Impala jet aircraft. Advanced piloting techniques and nonlinear modeling techniques based on fuzzy logic and multivariate orthogonal function methods were implemented with efficient onboard calculations and flight operations to achieve real-time maneuver monitoring, near-real-time global nonlinear aerodynamic modeling, and prediction validation testing in flight. Results demonstrated that global nonlinear aerodynamic models for a large portion of the flight envelope were identified rapidly and accurately using piloted flight test maneuvers during a single flight, with the final identified and validated models available before the aircraft landed.

The Ares I-X trajectory reconstruction produced best-estimated trajectories of the flight-test vehicle ascent through stage separation and of the first- and upper-stage entries after separation. The trajectory-reconstruction process combines onboard, ground-based, and atmospheric measurements to produce the trajectory estimates, using an iterated extended Kalman filter algorithm. The Ares I-X vehicle had a number of onboard and ground-based sensors that were available, including inertial measurement units, radar, air data, and weather balloons. However, due to problems with calibrations and/or data, not all of the sensor data were used. This paper describes the methodology and results of the trajectory-reconstruction process, including flight-data preprocessing and input uncertainties, trajectory-estimation algorithms and dynamic models, output transformations, and comparisons with preflight predictions. The results of the reconstruction indicate nominal vehicle performance that is well within the range of expected dispersions based on preflight Monte Carlo analysis.

Results of the Aeronautics Research Mission Directorate Seedling Project Phase I research project entitled Nonlinear Aerodynamics Modeling using Fuzzy Logic are presented. Efficient and rapid flight test capabilities were developed for estimating highly nonlinear models of airplane aerodynamics over a large flight envelope. Results showed that the flight maneuvers developed, used in conjunction with the fuzzy-logic system identification algorithms, produced very good model fits of the data, with no model structure inputs required, for flight conditions ranging from cruise to departure and spin conditions.

The Ares I-X trajectory reconstruction produced best estimated trajectories of the flight test vehicle ascent through stage separation, and of the first and upper stage entries after separation. The trajectory reconstruction process combines on-board, ground-based, and atmospheric measurements to produce the trajectory estimates. The Ares I-X vehicle had a number of on-board and ground based sensors that were available, including inertial measurement units, radar, air- data, and weather balloons. However, due to problems with calibrations and/or data, not all of the sensor data were used. The trajectory estimate was generated using an Iterative Extended Kalman Filter algorithm, which is an industry standard processing algorithm for filtering and estimation applications. This paper describes the methodology and results of the trajectory reconstruction process, including flight data preprocessing and input uncertainties, trajectory estimation algorithms, output transformations, and comparisons with preflight predictions.

The Ares I-X trajectory reconstruction produced best estimated trajectories of the flight test vehicle ascent through stage separation, and of the first and upper stage entries after separation. The trajectory reconstruction process combines on-board, ground-based, and atmospheric measurements to produce the trajectory estimates. The Ares I-X vehicle had a number of on-board and ground based sensors that were available, including inertial measurement units, radar, air-data, and weather balloons. However, due to problems with calibrations and/or data, not all of the sensor data were used. The trajectory estimate was generated using an Iterative Extended Kalman Filter algorithm, which is an industry standard processing algorithm for filtering and estimation applications. This paper describes the methodology and results of the trajectory reconstruction process, including flight data preprocessing and input uncertainties, trajectory estimation algorithms, output transformations, and comparisons with preflight predictions.

The ARES I-X vehicle was launched from NASA's Kennedy Space Center (KSC) on October 28, 2009 at approximately 11:30 EDT. ARES I-X was the first test flight for NASA s ARES I launch vehicle, and it was the first non-Shuttle launch vehicle designed and flown by NASA since Saturn. The ARES I-X had a 4-segment solid rocket booster (SRB) first stage and a dummy upper stage (US) to emulate the properties of the ARES I US. During ARES I-X pre-flight modeling and analysis, six (6) independent time domain simulation tools were developed and cross validated. Each tool represents an independent implementation of a common set of models and parameters in a different simulation framework and architecture. Post flight data and reconstructed models provide the means to validate a subset of the simulations against actual flight data and to assess the accuracy of pre-flight dispersion analysis. Post flight data consists of telemetered Operational Flight Instrumentation (OFI) data primarily focused on flight computer outputs and sensor measurements as well as Best Estimated Trajectory (BET) data that estimates vehicle state information from all available measurement sources. While pre-flight models were found to provide a reasonable prediction of the vehicle flight, reconstructed models were generated to better represent and simulate the ARES I-X flight. Post flight reconstructed models include: SRB propulsion model, thrust vector bias models, mass properties, base aerodynamics, and Meteorological Estimated Trajectory (wind and atmospheric data). The result of the effort is a set of independently developed, high fidelity, time-domain simulation tools that have been cross validated and validated against flight data. This paper presents the process and results of high fidelity aerospace modeling, simulation, analysis and tool validation in the time domain.

Flight and calibration results are presented for the Ares I-X 5-hole probe. The probe is calibrated by using a combination of wind tunnel, CFD, and other numerical modeling techniques. This is then applied to the probe flight data and comparisons are made between the vanes and 5-hole probe. Using this and other data it is shown the probe was corrupted by water rendering that measurement unreliable.

During the launch countdown of the Ares I-X test vehicle, engineers from Langley Research Center will use profiles of atmospheric density and winds in evaluating vehicle ascent loads and controllability. A schedule for the release of balloons to measure atmospheric density and winds has been developed by the Natural Environments Branch at Marshall Space Flight Center to help ensure timely evaluation of the vehicle ascent loads and controllability parameters and support a successful launch of the Ares I-X vehicle.

This paper describes the control challenges posed by the Ares I vehicle, the flight control system design and performance analyses used to test and verify the design. The major challenges in developing the control system are structural dynamics, dynamic effects from the powerful first stage booster, aerodynamics, first stage separation and large uncertainties in the dynamic models for all these. Classical control techniques were employed using innovative methods for structural mode filter design and an anti-drift feature to compensate for translational and rotational disturbances. This design was coded into an integrated vehicle flight simulation and tested by Monte Carlo methods. The product of this effort is a linear, robust controller design that is easy to implement, verify and test.

The flight test of the Ares I-X vehicle provides a unique opportunity to reduce risk of the design of the Ares I vehicle and test out design, math modeling, and analysis methods. One of the key features of the Ares I design is the significant static aerodynamic instability coupled with the relatively flexible vehicle - potentially resulting in a challenging controls problem to provide adequate flight path performance while also providing adequate structural mode damping and preventing adverse control coupling to the flexible structural modes. Another challenge is to obtain enough data from the single flight to be able to conduct analysis showing the effectiveness of the controls solutions and have data to inform design decisions for Ares I. This paper will outline the modeling approaches and control system design to conduct this flight test, and also the system identification techniques developed to extract key information such as control system performance (gain/phase margins, for example), structural dynamics responses, and aerodynamic model estimations.

Tare test data of a free-to-roll test rig with wind off are analyzed for the bearing friction by a method of differential corrections. The resulting data set is extended to transonic wind-on test conditions in oscillation frequency, amplitude, and normal force by assuming the friction coefficient to be inversely proportional to the frequency. The extended set of data is employed to establish a numerical model through a fuzzy logic algorithm. Some results showing the effects of frequency, amplitude, and normal force on the friction torque are illustrated in the paper. The predicted friction torque is then removed from the balance readings of rolling moments in the free-to-roll testing. Results of data analysis for the F-18C and F-16C models indicate that bearing-friction torques have very significant effects on the measured or experimentally determined roll damping. Without friction correction, the measured roll damping would be too small and, in some cases, even incorrect in sign. The effect of friction torque on the dihedral effect tends to be minor. Exceptions to these results appear when the configuration exhibits unstable roll damping in transonic flow.

Detailed analysis of wind-tunnel data in free-to-roll testing of one aircraft model, the preproduction F/A-18E, in a transonic tunnel is presented. The main purpose is to identify possible uncommanded rolling motions of the full-scale aircraft by examining the roll dynamic characteristics of the model in the tunnel. To improve the tunnel balance data, the bearing-friction effect on the balance rolling-moment coefficient was removed. The corrected rolling-moment coefficients are then modeled through a fuzzy-logic algorithm. The resulting aerodynamic models are employed in calculating all roll derivatives by a central-difference scheme. The proposed wing-drop theory relies on the values of the relative aerodynamic stiffness in the rolling equation of motion, which is assumed to be composed of two terms: derivatives with respect to the roll angle alone and second-order derivatives of the rolling-moment coefficient with respect to the roll angle and angle of attack. Wing drop is predicted if the relative aerodynamic stiffness changes sign from that of the overall motion and if the contributions to the rolling-moment coefficient from both the first-order and second-order derivatives are of the same sign and are small in value. This is equivalent to the vanishing of a frequency with damping in the vibration theory. It is found that at a low angle of attack and a transonic Mach number, the dynamic motion is wing rock. As the angle of attack is increased, the wing-drop condition is initially exhibited with a single event, then with multiple occurrences of wing drop at a higher angle of attack, with the magnitude of roll-off angles changing with time. It is also found that wing-rock motion in the tunnel on a free-to-roll test rig is mostly caused by the unstable effect of time rate of sideslip angle, not by the traditional roll damping due to roll rate.

The free-to-roll technique is used as a tool for-predicting areas of uncommanded lateral motions. Recently, the NASA/Navy/Air Force Abrupt Wing Stall Program extended the use of this technique to the transonic speed regime. Using this technique, this paper evaluates various wing configurations on the Joint Strike Fighter (F-35) aircraft. The configurations investigated include leading- and trailing-edge flap deflections, leading-edge flap gap seals, and vortex generators. These tests were conducted in the NASA Langley 16-Foot Transonic Tunnel. The analysis used a modification of a figure of merit developed during the Abrupt Wing Stall Program to discern configuration effects. The results showed how the figure of merit can be used to schedule, wing flap deflections to avoid areas of uncommanded lateral motion. The analysis also used both static and dynamic wind-tunnel data to provide insight into the uncommanded lateral behavior. The dynamic data were extracted from the time history data using parameter identification techniques. In general, sealing the gap between the inboard and outboard leading-edge flaps on the Navy version of the F-35 eliminated uncommanded lateral activity or delayed the activity to a higher angle of attack.

An overview of dynamic test techniques used at NASA Langley Research Center on scale models to obtain a comprehensive flight dynamics characterization of aerospace vehicles is presented. Dynamic test techniques have been used at Langley Research Center since the 1920s. This paper will provide a partial overview of the current techniques available at Langley Research Center. The paper will discuss the dynamic scaling necessary to address the often hard-to-achieve similitude requirements for these techniques. Dynamic test techniques are categorized as captive, wind tunnel single degree-of-freedom and free-flying, and outside free-flying. The test facilities, technique specifications, data reduction, issues and future work are presented for each technique. The battery of tests conducted using the Blended Wing Body aircraft serves to illustrate how the techniques, when used together, are capable of characterizing the flight dynamics of a vehicle over a large range of critical flight conditions.

Several aircraft airframe modeling approaches are currently being used in the DoD community for acquisition, threat evaluation, training, and other purposes. To date there has been no clear empirical study of the impact of airframe simulation fidelity on piloted real-time aircraft simulation study results, or when use of a particular level of fidelity is indicated. This paper documents a series of piloted simulation studies using three different levels of airframe model fidelity. This study was conducted using the NASA Langley Differential Maneuvering Simulator. Evaluations were conducted with three pilots for scenarios requiring extensive maneuvering of the airplanes during air combat. In many cases, a low-fidelity modified point-mass model may be sufficient to evaluate the combat effectiveness of the aircraft. However, in cases where high angle-of-attack flying qualities and aerodynamic performance are a factor or when precision tracking ability of the aircraft must be represented, use of high-fidelity models is indicated.

The free-to-roll technique is used as a tool for predicting areas of uncommanded lateral motions. Recently, the NASA/Navy/Air Force Abrupt Wing Stall Program extended the use of this technique to the transonic speed regime. Using this technique, this paper evaluates various wing configurations on the pre-production F/A-18E aircraft and the Joint Strike Fighter (F-35) aircraft. The configurations investigated include leading and trailing edge flap deflections, fences, leading edge flap gap seals, and vortex generators. These tests were conducted in the NASA Langley 16-Foot Transonic Tunnel. The analysis used a modification of a figure-of-merit developed during the Abrupt Wing Stall Program to discern configuration effects. The results showed how the figure-of-merit can be used to schedule wing flap deflections to avoid areas of uncommanded lateral motion. The analysis also used both static and dynamic wind tunnel data to provide insight into the uncommanded lateral behavior. The dynamic data was extracted from the time history data using parameter identification techniques. In general, modifications to the pre-production F/A-18E resulted in shifts in angle-of-attack where uncommanded lateral activity occurred. Sealing the gap between the inboard and outboard leading-edge flaps on the Navy version of the F-35 eliminated uncommanded lateral activity or delayed the activity to a higher angle-of-attack.