Lags between pilot inputs and aircraft responses may lead to a pilot-vehicle system entering into a pilot-induced oscillation. A model reference control element can compensate for these lags, avoiding pilot-induced oscillations and improving tracking performance. Neural network compensation drives the combined controller-plant system to exhibit a closed-loop response similar to that of an idealized system, immune to the factors that trigger pilot-induced oscillations. Actuator rate limiting has been a historically ubiquitous cause of pilot-induced oscillations; meanwhile, aeroelastic effects pose a threat to future lightweight, flexible aircraft designs. Model reference control can reduce pilot-induced oscillation tendencies caused by either of these factors. This paper presents the methods used to train the model reference controller, including conditions for system stability under model reference control. It then showcases results from a pair of simulation experiments applying the control design to compensate for rate limiting and aeroelasticity, respectively. Results demonstrate that the model reference control scheme reduces pilot-induced oscillation tendencies and improves closed-loop tracking performance.

The "bandwidth criteria" serve as simple yet effective model-based guidelines for predicting an aircraft's pilot-induced oscillation tendencies. However, these criteria were designed for linear, rigid-body models of aircraft. Their effectiveness on a new generation of lightweight, flexible aircraft is yet to be determined. This paper evaluates the pilot-induced oscillation bandwidth criteria on a variety of flexible aircraft dynamics models. The bandwidth criteria were first applied to predict the pilot-induced oscillation tendencies of each model. To validate these predictions, simulated pilot models performed close pitch tracking tasks using each set of vehicle dynamics. This paper recommends design methods for simulation tasks and appropriate pilot models to uncover pilot-induced oscillation tendencies in flexible aircraft. These simulations identified which configurations experienced pilot-induced oscillations during these tracking tasks. By comparing the bandwidth criteria predictions to the time history analyses, results showed that the bandwidth criteria accurately predicted pilot-induced oscillation tendencies for the flexible models.

The design of complex engineering systems leads to solving very large optimization problems involving different disciplines. Strategies allowing disciplines to optimize in parallel by providing sub-objectives and splitting the problem into smaller parts, such as Collaborative Optimization, are promising solutions.However, most of them have slow convergence which reduces their practical use. Earlier efforts to fasten convergence by learning surrogate models have not yet succeeded at sufficiently improving the competitiveness of these strategies.This paper shows that, in the case of Collaborative Optimization, faster and more reliable convergence can be obtained by solving an interesting instance of binary classification: on top of the target label, the training data of one of the two classes contains the distance to the decision boundary and its derivative. Leveraging this information, we propose to train a neural network with an asymmetric loss function, a structure that guarantees Lipshitz continuity, and a regularization towards respecting basic distance function properties. The approach is demonstrated on a toy learning example, and then applied to a multidisciplinary aircraft design problem.

View Video Presentation: https://doi.org/10.2514/6.2021-3063.vid Collaborative Optimization is a two-level design optimization architecture that allows the solution of multidisciplinary problems as a series of single-disciplinary problems with very little infrastructure overhead. It can be a very practical option in cases where practitioners do not have a fully coupled multidisciplinary optimizers at hand, but have access to multiple single-disciplinary optimizers. However, in its current form the method can require a large number of single-disciplinary evaluations and be slow to converge. In this work, we propose an improvement to the original method. We show how, at each iteration of the Collaborative Optimization framework, orthogonal projections on the feasible set of every discipline are performed. Mimicking the Bayesian Optimization framework, we propose to learn probabilistic representations of the feasible sets with the projection information, and use them to choose the next sample point in a principled manner. We also show that ensembles of Lipshitz neural networks are well performing probabilistic representations in this case. The approach is applied to a tailless aircraft range maximization problem.

System identification is a key step for model-based control, estimator design, and output prediction. This work considers the offline identification of partially observed nonlinear systems. We empirically show that the certainty-equivalent approximation to expectation-maximization can be a reliable and scalable approach for high-dimensional deterministic systems, which are common in robotics. We formulate certainty-equivalent expectation-maximization as block coordinate-ascent, and provide an efficient implementation. The algorithm is tested on a simulated system of coupled Lorenz attractors, demonstrating its ability to identify high-dimensional systems that can be intractable for particle-based approaches. Our approach is also used to identify the dynamics of an aerobatic helicopter. By augmenting the state with unobserved fluid states, a model is learned that predicts the acceleration of the helicopter better than state-of-the-art approaches. The codebase for this work is available at https://github.com/sisl/CEEM.

System identification is the process of building a mathematical model of an unknown system from measurements of its inputs and outputs. It is a key step for model-based control, estimator design, and output prediction. This work presents an algorithm for non-linear offline system identification from partial observations, i.e. situations in which the system's full-state is not directly observable. The algorithm presented, called SISL, iteratively infers the system's full state through non-linear optimization and then updates the model parameters. We test our algorithm on a simulated system of coupled Lorenz attractors, showing our algorithm's ability to identify high-dimensional systems that prove intractable for particle-based approaches. We also use SISL to identify the dynamics of an aerobatic helicopter. By augmenting the state with unobserved fluid states, we learn a model that predicts the acceleration of the helicopter better than state-of-the-art approaches.

In a pullout maneuver an initially diving aircraft is returned to level flight. Depending on the initial condition, aircraft characteristics and control inputs, altitude loss may be significant and minimizing it can be important to avoid collision with the ground. A motivating example is that of stall/spin recoveries, where the pullout represents a majority of the total altitude lost. This paper presents a solution of the minimal altitude loss pullout maneuver by posing it as an infinite horizon optimal control problem and solving it using dynamic programming techniques on a reduced-order point mass model for a lowwing general aviation aircraft. The computed optimal policy results in a “bang-bang” type controller, typical of shortest path problems, with maximum lift coefficient and bank rate applied at each point in time. The effect of maximum lift coefficient on the minimum altitude loss is analyzed, showing that attaining the highest lift coefficient possible throughout the pullout is critical. Based on these results a pullout flight control system is designed, with the optimal policy acting as an outer loop issuing commands to two inner loops that track lift coefficient and roll rate, respectively. The proposed pullout controller is tested on 6 DOF simulations, and shown to be effective at recovering the aircraft with close to optimal altitude loss.

This paper presents methodologies and techniques for stall/spin flight testing using commercially off-the-shelf model aircraft and flight data systems. Flight test procedures allowing systematic and efficient testing are outlined, as well as piloting techniques to obtain quality spin flight data. Two spin flight test campaigns with different testbeds are described, including the aircraft, flight data system and ground station setups. Flight test data for the two testbeds are presented and the incipient, developed and recovery characteristics are analyzed. The spin data is consistent across spin maneuvers for a given aircraft, and also similar between aircrafts of different scale, showing the accuracy of the flight test methodology and instrumentation.

Aircraft flown in formation can realize significant reductions in drag by flying in regions of wake upwash. However, most transports fly at transonic speeds where the impact of compressibility on formation flight is not well understood. This study uses an Euler solver to analyze the inviscid aerodynamic forces and moments of transonic wing/body configurations flying in a two-aircraft formation. Formations with large streamwise separation distances (10–50 wingspans) are considered. This work indicates that compressibility-related drag penalties in formation flight may be eliminated by slowing 2–3% below the nominal out-of-formation cruise Mach number, either at fixed lift coefficient or fixed altitude. The latter option has the additional benefit that the aerodynamic performance of the formation improves slightly at higher lift coefficients. Although optimal in-formation lift coefficients are not as high as those estimated by incompressible analyses, modest increases in altitude can yield further improveme...

We quantify the fuel and cost benefits of applying extended formation flight to commercial airline operations. Central to this study is the development of a bi-level, mixed-integer real formation flight optimization framework. The framework has two main components: 1)a continuous-domain aircraft mission performance optimization and 2)an integer optimization component that selects the best combination of optimized missions to form a formation flight schedule. The mission performance reflects the effects of rolled-up wakes, formation heterogeneity, and formation-induced compressibility. The results show that an airline can use formation flight to reduce fuel burn by 5.8% or direct operating cost by 2.0% in a long-haul international schedule. The savings increase to 7.7% in fuel or 2.6% in cost for a large-scale, transatlantic airline alliance schedule. These results include the effects of a conservative fuel reserve for formation flight. Sensitivity studies show that a modest reduction in the cruise Mach number may be sufficient to manage the impact of formation-induced compressibility effects on system-level formation flight performance. We demonstrate that the potential savings from extended formation flight (an operational improvement using existing aircraft) can approach those claimed for advanced vehicle technologies and unconventional configurations.

Sonic boom tailoring techniques for aircraft roughly the size and weight of a business jet have been demonstrated in flight-test,1 suggesting that the design of low-boom or even boom-less aircraft may be possible; however, the computational cost and complexity associated with current state-of-the-art tools makes e↵ective conceptual-level design infeasible when using traditional optimization approaches. Ongoing e↵orts in low sonic boom design have focused on gradient-driven shaping of the signature in the near-field2,3 or on the ground4,5 to directly drive the high-fidelity shape optimization. Gradient-driven and gradient-free approaches incorporating an integrated objective in the form of a noise level are described by Alonso6 et al. Low-fidelity, linearized (area-rule) approaches are also described in Ref. 6; however, these methods by definition are unable to accurately capture the non-linear features — namely, shocks — present in supersonic flows. While invaluable during design exploration, details of the shape produced by such approaches do not readily carry over to equivalent high-fidelity representations of the geometry. This paper presents an approach that decouples the low-boom aircraft design problem into two discrete components. Sonic boom minimization is performed using an approach based on linear theory, where a low-dimensionality, parametrized form of the near-field pressure signal enables rapid discovery of candidate target signal shapes. These target signals are then used in an inverse-design CFD framework, where a gradient-driven optimizer coupled with a discrete adjoint approach e ciently solves for the vehicle shape whose near-field signal matches the supplied target. A parametric geometry modeler is used to generate the surface triangulations and surface sensitivities required by this approach. A multi-disciplinary conceptual design tool is used to provide bounds for the high-fidelity shape optimization, ensuring that the final design is able to meet performance goals such as cruise range and takeo↵ field length. The current work starts with a discussion of the components comprising the design approach. Particular attention is paid to the linear component, which integrates a variety of tools to provide a robust, automated system for sonic boom minimization. Studies with simple axisymmetric bodies are then performed to characterize the mesh required for accurate resolution of near-field pressure signals in the CFD domain. This is followed by application of the adjoint-driven inverse-design component, again using a simple axisymmetric

Traditionally, aircraft shaping to minimize sonic boom has been done using methods 1,2 that generate equivalent area distributions which produce two-shock ground signals. The aircraft is then designed to meet this target equivalent area distribution. Although the area distributions produced by these methods achieve reduced ground signature overpressures compared to the traditional N-wave, the drag of the corresponding aircraft may be large. Additionally, these methods are restricted to finding area distributions that result in ground signals with only two shocks. We propose a method that addresses both issues. First, increased flexibility in the shape of the area distribution permits trade off between ground signal noise and aircraft performance for a specific signal shape. Second, this new method is capable of finding equivalent area distributions that result in ground signals containing more than two shocks, and is also able to eliminate the aft shock. We conclude by demonstrating the multi-shock inverse design method on several ground signals: a simple two-shock ground signal, ground signals containing five and nine shocks, and a two-shock ground signal with no aft shock.

Robust and optimal trajectories are computed for a dynamic soaring unmanned aerial vehicle. Dynamic soaring is technique for extracting energy from atmospheric wind gradients that large seabirds such as the albatross use as their primary means of propulsive power. The stochastic nature of the soaring problem is explicitly included in the optimization formulation by using an uncertainty quantication technique known as stochastic collocation. This non-intrusive method can generate accurate estimates of solution statistics with signicantly fewer samples than a Monte Carlo based approach. An analysis of variance is rst performed to determine which environmental parameters, aircraft properties, and initial conditions contribute most to nal energy and altitude variance. Robust dynamic soaring trajectories are then computed by using stochastic collocation over carefully chosen nested sparse grids.

An analysis of propeller-win g combinations in inviscid incompressible flow reveals some of the fundamental interactions which affect the performance of an installed propulsion system. A generalized version of Munk's stagger theorem is used in a rapid, approximate calculation of optimal lift distributions and installed efficiency. Results indicate that the distribution of lift over the wing which maximizes overall efficiency differs markedly from elliptic loading. Swirl recovery by the wing leads to increments in net propeller efficiency of 6% in example cases. The maximum installed efficiency is computed for single-rotation (up-inboard and up-outboard designs) and counter-rotating systems. Results suggest that some of the performance advantages attributed to counterrotation may be less dramatic for well-integrated wing-propeller designs than for isolated systems. Nomenclature An = amplitude of nth harmonic of wing lift & = wing aspect ratio b = wingspan c =wing chord CL =wing lift coefficient CT = thrust coefficient, =ir2T/pw2R4 D =drag IU,IW = definite integrals, Eqs. (10) and (11), respectively / = advance ratio, = irU00/uR I = section lift L =lift N = number of blades Obj = objective function Q = propeller torque R = propeller radius r - radial coordinate T = thrust Uw = freestream velocity u = axial perturbation velocity V = induced velocity vt = tangential induced velocity w = induced downwash x = stream wise coordinate y = spanwise coordinate .Vprop = spanwise location of propeller F = circulation 6 =dimensionless spanwise coordinate K =Goldstein's radial velocity correction X =Lagrange multiplier p = density Subscripts int = interference component = wing w = propeller

network Network optimization objective function Jsolo Optimal fuel burn or cost for each solo route from mission optimization JthreeAC Optimal fuel burn or cost for each three aircraft formation from mission optimization JtwoAC Optimal fuel burn or cost for each two aircraft formation from mission optimization uPh.D. Department of Aeronautics & Astronautics. Student Member AIAA †Ph.D. Department of Aeronautics & Astronautics. Student Member AIAA ‡Ph.D. Department of Aeronautics & Astronautics. Student Member AIAA §Professor, Department of Aeronautics & Astronautics. Fellow AIAA

This paper presents a compact formulation of the Vortex Lattice Method (VLM) which allows fast computation of the nonlinear inviscid aerodynamics of fixed-wing aircraft at low speed and low angle of attack flight. The formulation captures the nonlinearities present in this flight regime, but avoids the costly re-computation that regular VLMs require. Instead ap reprocessing step reduces the calculation of inviscid aerodynamic force and moment to simple quadratic expressions in terms of the flight and control variables. The size of these quadratic expressions is independent of the number of vortices used in the vortex lattice, allowing for fine discretization of lifting surfaces with no penalty in re-computation time. Full six degree of freedom flight simulations that include a VLM aerodynamic model can now be executed very efficiently using this approach. Furthermore, this formulation provides an analytic aerodynamic model allowing direct mathematical manipulation, such as analytic differentiation with respect to flight and control variables, making it useful for optimization problems where analytic gradients could reduce computational costs as well as improve convergence characteristics.