betaFeedback

WorksMathematics

Thrust Regulation in a Solid Fuel Ramjet using Dynamic Mode Adaptive Control

Abstract

This paper presents the application of a novel data-driven adaptive control technique, called dynamic mode adaptive control (DMAC), for regulating thrust in a solid fuel ramjet (SFRJ). A high-fidelity computational model incorporating compressible flow theory and equilibrium chemistry is used to simulate the combustion dynamics. An adaptive tracking controller is designed using the DMAC framework, which leverages dynamic mode decomposition to approximate the local system behavior, followed by a tracking controller synthesized around the identified model. Simulation results demonstrate that DMAC provides an effective and reliable approach for thrust regulation in SFRJs. In addition, a systematic hyperparameter sensitivity study is conducted by varying the tuning parameters over several orders of magnitude. The resulting responses show that the closed-loop performance and tracking error remain stable across wide parameter variations, indicating that DMAC exhibits strong robustness to hyper parameter tuning.

Introduction

Ramjet engines are well-suited for long-range, high-speed missions, owing to their ability to deliver sustained thrust over extended durations. Their operational simplicity, due to the absence of rotating turbomachinery, makes them easier to operate and maintain compared to air-breathing propulsion systems. Based on the type of fuel used, ramjets can be classified as either liquid-fuel (LFRJ) or solid-fuel (SFRJ) variants. Among these, SFRJs offer greater mechanical simplicity at comparable scales, as they do not require turbopumps, fuel bladders, injectors, or the associated plumbing. Furthermore, the higher volumetric energy density of solid fuels can enable SFRJs to achieve longer ranges than their liquid-fueled counterparts. An additional advantage of SFRJs is their combustion behavior: the flame front typically extends along the entire length of the fuel grain, which helps suppress combustion instabilities that are more common in LFRJs.

Figure 1 illustrates the typical geometry of a Solid Fuel Ramjet (SFRJ). The solid fuel grain lines the combustor wall, where the high-speed, high-temperature airflow vaporizes and ignites the fuel. The resulting combustion adds energy to the flow, producing thrust. Stable operation requires carefully maintained flow conditions within the combustor. If the airflow rate is too low, the heat release may be insufficient to generate the required thrust. On the other hand, an excessively high flow rate can lead to inlet unstart due to excessive heat addition or exceed the blowoff limits of the combustor [1, 2, 3]. Both situations increase the risk of flame extinction and thrust loss. As such, the thermodynamic state within the combustor, comprised of pressure, temperature, and mass flow rate, must remain within a narrow, controlled operating range. Thus, ensuring reliable operation of an SFRJ requires a control system that maintains the SFRJ’s state within acceptable limits and remains robust against parametric variations and external disturbances across the broadest possible operating envelope. However, due to the complexity of the multi-physics processes involved, including solid combustion, mixing, and high-speed flow, developing an analytical model that 1) predicts a stable operational range, 2) is computationally efficient, and 3) can be used for control system design, is highly challenging.

A typical SFRJ cross section

Figure 1. A typical SFRJ cross section.

In this paper, we consider the problem of developing an adaptive control system that does not require a system model, but instead uses limited measurements to update the control law and generate the required control signal to regulate the thrust generated by the SFRJ. This work complements our previous work described in [4, 5, 6], which investigated the application of the retrospective cost adaptive control to regulate the thrust generated by a quasi-static model of SFRJ and a computational model that simulates the multiphysics dynamics in an SFRJ.

The control design approach used in this work is based on the dynamic mode adaptive control (DMAC) algorithm, which is described in detail in [7]. Although our previous work based on the retrospective cost adaptive control (RCAC) framework has been successfully applied to the thrust control problem in ramjets [4, 5, 6, 8, 9, 10], RCAC requires the choice of a target filter that captures the essential modeling information required to update the control law. However, the fixed filter choice limits the operational envelope of the system and significantly complicates the design process in a multi-input, multi-output system. In contrast, DMAC does not require any modeling information and instead uses the measured data to identify a low-order dynamic approximation of the system along with an adaptive linear controller and thus potentially has a larger operational envelope.

To simulate the flow inside an SFRJ, we use a high-fidelity computational model described in detail in [11]. Specifically, a truncated section of an SFRJ geometry is used to simulate the internal flow in order to keep computational cost low. To model the continuum hypersonic flows, we use the SU2-NEMO (NonEquilibrium MOdels) code that solves the Navier-Stokes equations for multi-species gases in thermochemical nonequilibrium [12, 13, 14].

The paper is organized as follows. Section II describes the computational model in detail and the simplified SFRJ geometry used in this work. Section III briefly reviews the dynamic mode adaptive control framework to regulate the thrust of the SFRJ model. Section IV presents simulation results to demonstrate the application of the DMAC technique to regulate the SFRJ thrust. Finally, the paper concludes in § 5.

Computational Model of SFRJ

This section briefly describes the computational model of the SFRJ considered in this work. The governing equations of the flow are described in detail in [6, 11]. The CFD software used for this work is SU2, a computational analysis and design package that has been developed to solve multiphysics analysis and optimization tasks using unstructured mesh topologies [15]. SU2 employs a median-dual finite-volume approach to solve the discretized governing equations. Further details of the governing equations and numerical schemes of SU2 can be found in Ref. [15].

The computational domain consists of a truncated SFRJ geometry that includes the inlet channel and the combustor, as shown in Figure 2. The truncated geometry is considered to reduce the computational cost of the simulation. The inlet channel is 40 mm in diameter and has a length of 0.2 m. The combustor is 70 mm in diameter and has a length of 0.978 m. The total length for the simplified geometry is 1.178 m. Since the geometry is axisymmetric about the centerline, only half of the two-dimensional cross-section at the center of the geometry is simulated. Figure 3 shows the baseline Mach number contours in the computational domain with a supersonic inlet velocity of 695 m/s, a static inlet pressure of 100,000 Pa, and a static inlet temperature of 300 K and with no heat addition.

Truncated and full SFRJ geometry

Figure 2. Truncated and the full SFRJ geometry. Only the truncated section is considered in this work.

Mach number contour for truncated SFRJ

Figure 3. Mach number contour for truncated SFRJ.

Dynamic Mode Adaptive Control

This section presents the dynamic mode adaptive control (DMAC) algorithm. As shown in Figure 4, consider a dynamic system in a basic servo loop architecture whose input is uk∈Rlu_k \in\mathbb{R}^{l} and the output is yk∈Rpy_k \in\mathbb{R}^{p}. The objective of the DMAC controller is to generate a discrete-time input signal uku_k such that the sampled output yky_k tracks the reference signal rkr_k. The DMAC controller consists of a dynamic mode approximation to approximate the local system behavior and a tracking controller designed based on the identified model.

Dynamic Mode Adaptive Control architecture

Figure 4. Dynamic Mode Adaptive Control (DMAC) architecture for model-free, data-driven, and learning-based control of dynamic systems.

Dynamic Mode Approximation

Let ξk∈Rlξ\xi_k \in\mathbb{R}^{l_\xi} denote the measured portion of the state of the system. Note that ξk\xi_k may or may not be the entire state of the system. To compute the control signal uku_k, we first approximate linear maps A∈Rlξ×lξA \in\mathbb{R}^{l_\xi\times l_\xi} and B∈Rlξ×luB \in\mathbb{R}^{l_\xi\times l_u} such that

ξk+1=Aξk+Buk,(1)\xi_{k+1}=A\xi_k+Bu_k, \tag*{(1)}

which can be reformulated as

ξk+1=Θϕk.(2)\xi_{k+1}=\Theta\phi_k. \tag*{(2)}

where

Θ≜[AB]∈Rlξ×(lξ+lu),ϕk≜[ξkuk]∈Rlξ+lu.(3)\Theta\triangleq\begin{bmatrix} A & B \end{bmatrix} \in\mathbb{R}^{l_\xi\times(l_\xi+l_u)}, \qquad\phi_k \triangleq\begin{bmatrix} \xi_k \\ u_k \end{bmatrix} \in\mathbb{R}^{l_\xi+l_u}. \tag*{(3)}

A matrix Θ\Theta such that (2) is satisfied may not exist. However, an approximation of such a matrix can be obtained by minimizing

Jk(Θ)≜∑i=0kλk−i∥ξi−Θϕi−1∥22+λktr⁡(ΘTRΘΘ),(4)J_k(\Theta) \triangleq\sum_{i=0}^{k} \lambda^{k-i}\left\lVert\xi_i-\Theta\phi_{i-1}\right\rVert_2^2+\lambda^k\operatorname{tr}\left(\Theta^{T}R_\Theta\Theta\right), \tag*{(4)}

where RΘ∈R(lξ+lu)×(lξ+lu)R_{\Theta} \in\mathbb{R}^{(l_{\xi}+l_{u})\times(l_{\xi}+l_{u})} is a positive definite regularization matrix that ensures the existence of the minimizer of (4) and λ∈(0,1]\lambda\in(0,1] is a forgetting factor. In nonlinear or time-varying systems, Θ\Theta approximated by minimizing (4) in the case where the state varies significantly in the state space may not be able to capture the local linear behavior at the current state. Thus, incorporating a geometric forgetting factor to prioritize recent data over older data improves the linear approximation and prevents the algorithm from becoming sluggish.

Proposition III.1. Consider the cost function (4). For all k≥0k \ge0, define the minimizer of (4) as

Θk≜min⁡Θ∈Rlx×(lx+lu)Jk(Θ).(5)\Theta_k \triangleq\min_{\Theta\in\mathbb{R}^{l_x \times(l_x+l_u)}} J_k(\Theta). \tag*{(5)}

Then, the minimizer Θk\Theta_k satisfies

Θk=Θk−1+(ξk−Θk−1ϕk−1)ϕk−1TPk,(6)\Theta_k = \Theta_{k-1} + (\xi_k-\Theta_{k-1}\phi_{k-1})\phi_{k-1}^{\mathsf{T}}P_k, \tag*{(6)}
Pk=λ−1Pk−1−λ−1Pk−1ϕk−1γk−1−1ϕk−1TPk−1,(7)P_k = \lambda^{-1}P_{k-1}-\lambda^{-1}P_{k-1}\phi_{k-1}\gamma_{k-1}^{-1}\phi_{k-1}^{\mathsf{T}}P_{k-1}, \tag*{(7)}

where, for all k≥0k \ge0, γk≜λ+ϕk−1TPk−1ϕk−1\gamma_k \triangleq\lambda+\phi_{k-1}^{\mathsf{T}}P_{k-1}\phi_{k-1}, and Θ0=0\Theta_0=0, P0≜RΘ−1P_0 \triangleq R_{\Theta}^{-1}.

Proof. See Proposition V.2 in [7]. □

Note that the cost function (4) is a matrix extension of the cost function typically considered in engineering applications [16]. As shown in [16, 17, 18], persistency of excitation is required to ensure that 1) the estimate converges and 2) the corresponding covariance matrix PkP_k remains bounded. To ensure the persistency of excitation, in this paper, we introduce a zero-mean white noise in the control signal to promote persistency in the regressor ϕk\phi_k, as discussed in Section III.B.

Tracking Controller

This subsection presents the algorithm to compute the control signal uku_k using the dynamics approximation computed in Section III.A. To track the reference signal rk∈Rr_k \in\mathbb{R}, the DMAC algorithm uses the fullstate feedback controller with integral action. Note that the full state refers to the state ξk\xi_k and not the system state xkx_k. In particular, the control law is

uk=Kξ,kξk+Kq,kqk+vk,(8)u_k = K_{\xi,k}\xi_k + K_{q,k}q_k + v_k, \tag*{(8)}

where

ξk≜[Va,kPa,k]∈R2,(9)\xi_k \triangleq\begin{bmatrix} V_{a,k} \\ P_{a,k} \end{bmatrix} \in\mathbb{R}^{2}, \tag*{(9)}

denotes the DMAC state vector. Here, Va,kV_{a,k} and Pa,kP_{a,k} represent the normalized average flow velocity and average pressure at the nozzle outlet, respectively. The normalization is performed by dividing each state value by its corresponding reference value obtained under nominal heat flux conditions of 2×106 W/m22 \times10^{6}\ \mathrm{W}/\mathrm{m}^{2}. That is,

Va,k=Vout,kVref,Pa,k=Pout,kPref,(10)V_{a,k} = \frac{V_{\mathrm{out},k}}{V_{\mathrm{ref}}}, \qquad P_{a,k} = \frac{P_{\mathrm{out},k}}{P_{\mathrm{ref}}}, \tag*{(10)}

where VrefV_{\mathrm{ref}} and PrefP_{\mathrm{ref}} denote the outlet velocity and pressure measured at the nominal heat flux condition. This normalization ensures that the regressor ϕk\phi_k used in the dynamics approximation remains well-scaled and numerically stable across operating points.

The matrices Kξ,k∈Rlu×lξK_{\xi,k} \in\mathbb{R}^{l_u \times l_{\xi}} and Kq,k∈Rlu×lyK_{q,k} \in\mathbb{R}^{l_u \times l_y} are the time-varying fullstate feedback gain and the integrator gain, computed using the technique shown in Appendix A of [7] and vk∼N(0,σvIlu)v_k \sim\mathcal{N}(0,\sigma_v I_{l_u}) is a zero-mean white noise signal added to the control to promote persistency in the regressor ϕk\phi_k used in the dynamic mode approximation step. Note that the integrator state qkq_k satisfies

qk+1=qk+zk,(11)q_{k+1}=q_k+z_k, \tag*{(11)}

where zk≜rk−ykz_k \triangleq r_k-y_k is the output error.

The gain matrices Kξ,kK_{\xi,k} and Kq,kK_{q,k} are computed using the well-known linear-quadratic-integral control [19], which requires the AA, BB, and CC matrices of the system. The dynamics matrix AA and the input matrix BB are given by the dynamic mode approximation described in III.A. The computation of the output matrix CC is described below.

In this work, the system output is modeled using a neural network. The neural network is trained using simulation data from the SFRJ model described in Section II. Specifically, the thrust y∈Ry \in\mathbb{R} is modeled as a nonlinear function of ξ\xi, that is,

y=NN(ξ),(12)y = NN(\xi), \tag*{(12)}

where NN:R2→RNN : \mathbb{R}^{2} \to\mathbb{R} is a feedforward neural network [20].

In this work, the neural network is trained using MATLAB’s feedforwardnet function. The training data consists of two sets of high-fidelity simulation results, totaling 500 samples. Each sample consists of the pair (Va,Pa)(V_a, P_a) as inputs and the corresponding generated thrust yy as the output. The neural network architecture used in this work, shown in Figure 5, consists of two hidden layers, each with 10 neurons and tansig activation functions, followed by a linear output layer. The network is trained to minimize the mean squared error (MSE) loss between the predicted and actual thrust using the Levenberg–Marquardt algorithm. The dataset is randomly split into training (70%), validation (15%), and testing (15%) sets. Figure 6 shows the mean squared error (MSE) on the training, validation, and test sets over epochs. All three error curves decrease consistently, indicating proper learning without overfitting. The final validation error closely matches the training and test errors, suggesting that the trained network generalizes well to unseen data.

Figure 5. Artificial neural network architecture used in this work, consisting of two hidden layers with 10 neurons each and tansig activation functions.

Figure 6. Neural network training performance plot showing the mean squared error (MSE) on the training, validation, and test sets over epochs.

It follows from (12) that the linearized output matrix is given by

Ck=∂NN(ξ)∂ξ∣ξ=ξk,(13)C_k = \left.\frac{\partial NN(\xi)}{\partial\xi}\right|_{\xi=\xi_k}, \tag*{(13)}

where the Jacobian is numerically computed using the central difference scheme, that is,

∂NN(ξ)∂ξ≈12ε[NN(ξ+εe1)−NN(ξ−εe1)NN(ξ+εe2)−NN(ξ−εe2)].(14)\frac{\partial NN(\xi)}{\partial\xi} \approx\frac{1}{2\varepsilon}\begin{bmatrix} NN(\xi+\varepsilon e_1)-NN(\xi-\varepsilon e_1) \\ NN(\xi+\varepsilon e_2)-NN(\xi-\varepsilon e_2) \end{bmatrix}. \tag*{(14)}

Note that e1,e2∈R2e_1,e_2 \in\mathbb{R}^2 are the standard basis vectors, and, in this work, we set ε=10−7\varepsilon=10^{-7}.

Simulation Results

This section presents numerical examples that demonstrate the application of the DMAC technique for regulating the thrust generated by the SFRJ. The DMAC hyperparameters were selected through a simple grid search to achieve a satisfactory transient response under nominal conditions.

The control signal uku_k is used to modulate the heat flux wkw_k as

wk=w0−Kw×uk,(15)w_k=w_0-K_w\times u_k, \tag*{(15)}

where w0w_0 is the nominal heat flux, uku_k is the adaptive control signal generated by the DMAC algorithm, and KwK_w is the scaling factor. The linear map (15) is chosen such that the magnitude of the adaptive control signal uku_k remains close to O(1)O(1) to ensure the numerical stability of the DMAC algorithm. In this work, the nominal heat flux w0=2×106w_0=2\times10^6 W/m2^2 and the scaling factor KwK_w is set to 10510^5.

Command Following

First, we consider the problem of regulating the SFRJ thrust to a constant value. In particular, the SFRJ is commanded to generate a constant thrust value of r=1000r=1000 N. Since ξk∈R2\xi_k\in\mathbb{R}^2 and uk∈Ru_k\in\mathbb{R}, it follows that Θk\Theta_k is a 2×32\times3 matrix. In DMAC, we set RΘ=102I3R_{\Theta}=10^2I_3 and the forgetting factor λ=0.995\lambda=0.995. The LQR weights in the tracking controller are R1=I3R_1=I_3 and R2=1R_2=1.

Figure 7 shows the closed-loop response of the SFRJ with the DMAC algorithm updating the controller, where a) shows the commanded thrust r=1000r=1000 N and the generated signal yky_k, b) shows the control signal uku_k, c) shows the absolute value of the tracking error zk≜yk−rz_k \triangleq y_k-r on a logarithmic scale, and d) shows the estimate matrix Θk\Theta_k computed by DMAC. Note that the output error approached zero.

Figure 7. Closed-loop response of SFRJ with DMAC. a) shows the output yky_k and the reference signal rr, b) shows the control signal uku_k, and c) shows the absolute value of the tracking error zkz_k on a logarithmic scale.

Next, the SFRJ is commanded to follow a double-step command. Specifically, the thrust command is r=1000r=1000 N for k∈(0,200)k\in(0,200) and r=1200r=1200 N for k≥200k\ge200. Note that DMAC hyperparameters are kept the same as in the previous case. Figure 8 shows the closed-loop response of the SFRJ with the DMAC algorithm updating the controller, where a) shows the commanded thrust rr and the generated signal yky_k, b) shows the control signal uku_k, c) shows the absolute value of the tracking error zk≔yk−rz_k \coloneqq y_k-r on a logarithmic scale, and d) shows the estimate matrix Θk\Theta_k computed by DMAC. Note that the output error approached zero.

Figure 8. Closed-loop response of SFRJ with DMAC. a) shows the output yky_k and the reference signal rr, b) shows the control signal uku_k, and c) shows the absolute value of the tracking error zkz_k on a logarithmic scale.

Effect of Hyperparameters

To investigate the robustness of the DMAC algorithm to its tuning hyperparameters RΘR_{\Theta}, λ\lambda, R1R_1 and R2R_2, we vary each of the hyperparameters systematically by keeping other hyperparameters at their nominal values. Figure 9 shows the effect of the DMAC hyperparameters on the closed-loop response yky_k and the absolute tracking error ∣zk∣|z_k|. From top to bottom, the rows correspond to variations in RΘR_{\Theta}, λ\lambda, R1R_1, and R2R_2, respectively. Note that, in each case, the hyperparameter is varied by a few orders of magnitude, suggesting that DMAC is robust to tuning parameters.

Figure 9. Effect of DMAC hyperparameters on the closed-loop performance.

Conclusion

This paper considered the problem of regulating the thrust generated by a solid fuel ramjet (SFRJ) using limited in-situ measurements, without requiring an analytical model of the system. The simulation results presented in this work demonstrate that the dynamic mode adaptive control (DMAC) method is a viable and effective technique for thrust regulation in SFRJs. In particular, its minimal measurement requirements enhance its practicality for real-world implementation.

In addition, a systematic hyperparameter sensitivity analysis was conducted to evaluate the robustness of the DMAC framework with respect to its tuning parameters. The results indicate that the closed-loop performance and tracking error remain stable across wide variations in these parameters, suggesting that DMAC exhibits strong robustness to hyperparameter tuning. This property further supports its suitability for deployment in uncertain and dynamically changing propulsion environments.

Acknowledgment

This research was supported by the Office of Naval Research grant N00014-23-1-2468.

References

  1. [1]Arjun, P., and Nagaraja, S., “Unstart phenomenon in a scramjet engine isolator,” Recent Advances in Thermofluids and Manufacturing Engineering: Select Proceedings of ICTMS 2022, Springer, 2022, pp. 195–204.DOI
  2. [2]Varshney, M., and Baig, M., “Unstart control in scramjet engines,” AIAA Scitech 2019 Forum, 2019, p. 0297.DOI
  3. [3]Yuceil, K., Valdivia, A., Wagner, J., Clemens, N., and Dolling, D., “Active control of supersonic inlet unstart using vortex generator jets,” 39th AIAA Fluid Dynamics Conference, 2009, p. 4022.DOI
  4. [4]Oveissi, P., Trivedi, A., Goel, A., Tumuklu, O., Hanquist, K. M., Farahmandi, A., and Philbrick, D., “Learning-based adaptive thrust regulation of solid fuel ramjet,” AIAA SICTECH 2023 Forum, 2023, p. 2533.DOI
  5. [5]Oveissi, P., Goel, A., Tumuklu, O., and Hanquist, K. M., “Adaptive Combustion Regulation in Solid Fuel Ramjet,” AIAA SICETCH 2024 Forum, 2024, p. 0743.DOI
  6. [6]Oveissi, P., Dorsey, A., Khokhar, G. T., Hanquist, K. M., and Goel, A., “Adaptive Combustion Regulation in High-Fidelity Computational Model of Solid Fuel Ramjet,” AIAA SciTech 2025 Forum, 2025, p. 0352.
  7. [7]Oveissi, P., and Goel, A., “Model-free Dynamic Mode Adaptive Control using Matrix RLS,” , 2025. URL https://arxiv.org/abs/2505.11844.
  8. [8]DeBoskey, R., Oveissi, P., Narayanaswamy, V., and Goel, A., “An In-situ Solid Fuel Ramjet Thrust Monitoring and Regulation Framework Using Neural Networks and Adaptive Control,” 2025 IEEE Conference on Control Technology and Applications (CCTA), IEEE, 2025, pp. 377–382.arxiv.org/abs/2506.08157
  9. [9]Goel, A., Duraisamy, K., and Bernstein, D. S., “Retrospective cost adaptive control of unstart in a model scramjet combustor,” AIAA Journal, Vol. 56, No. 3, 2018, pp. 1085–1096.DOI
  10. [10]Oveissi, P., Dorsey, A., McBeth, J., Hanquist, K. M., and Goel, A., “Learning-Based Thrust Regulation of Solid-Fuel Ramjet in Flight Conditions,” AIAA SciTech 2025 Forum, 2025, p. 2805.
  11. [11]Khokhar, G. T., McBeth, J., Hanquist, K. M., Oveissi, P., and Goel, A., “Investigation of Solid Fuel Ramjets Using Analytical Theory and Computational Fluid Dynamics,” AIAA SCITECH 2025 Forum, 2025, p. 0392.
  12. [12]Maier, W. T., Needels, J. T., Garbacz, C., Morgado, F., Alonso, J. J., and Fossati, M., “SU2-NEMO: An open-source framework for high-Mach nonequilibrium multi-species flows,” Aerospace, Vol. 8, No. 7, 2021, p. 193.DOI
  13. [13]Economon, T. D., Palacios, F., Copeland, S. R., Lukaczyk, T. W., and Alonso, J. J., “SU2: An open-source suite for multiphysics simulation and design,” AIAA Journal, Vol. 54, No. 3, 2016, pp. 828–846. https://doi.org/10.2514/1.J053813.
  14. [14]Maier, W. T., Needels, J. T., Garbacz, C., Morgado, F., Alonso, J. J., and Fossati, M., “SU2-NEMO: An Open-Source Framework for High-Mach Nonequilibrium Multi-Species Flows,” Aerospace 2021, Vol. 8, Page 193, Vol. 8, No. 7, 2021, p. 193. https://doi.org/10.3390/AEROSPACE8070193, URL https://www.mdpi.com/2226-4310/8/7/193/htmlhttps://www.mdpi.com/2226-4310/8/7/193.
  15. [15]Economon, T. D., Palacios, F., Copeland, S. R., Lukaczyk, T. W., and Alonso, J. J., “SU2: An open-source suite for multiphysics simulation and design,” AIAA Journal, Vol. 54, No. 3, 2016, pp. 828–846. https://doi.org/10.2514/1.J053813, publisher: American Institute of Aeronautics and Astronautics Inc.
  16. [16]Goel, A., Bruce, A. L., and Bernstein, D. S., “Recursive least squares with variable-direction forgetting: Compensating for the loss of persistency [lecture notes],” IEEE Control Systems Magazine, Vol. 40, No. 4, 2020, pp. 80–102.DOI
  17. [17]Y. Mareels, I., and Gevers, M., “Persistence of excitation criteria,” Proc. Conf. Dec. Contr, 1986, pp. 1933–1935. https://doi.org/10.1109/CDC.1986.267349.
  18. [18]Mareels, I. M., and Gevers, M., “Persistency of excitation criteria for linear, multivariable, time-varying systems,” Mathematics of Control, Signals, and Systems, Vol. 1, No. 3, 1988, pp. 203–226. https://doi.org/10.1007/BF02551284.
  19. [19]Young, P. C., and Willems, J., “An approach to the linear multivariable servomechanism problem,” International journal of control, Vol. 15, No. 5, 1972, pp. 961–979.DOI
  20. [20]Rozario, T., Oveissi, P., and Goel, A., “Matrix-Based Representations and Gradient-Free Algorithms for Neural Network Training,” 2024 International Conference on Machine Learning and Applications (ICMLA), IEEE, 2024, pp. 325–332.

Paper details

vlnc:W_m9z2jfca
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 9

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents