Manoranjan Sinha

Neural Partial Differential Method for Extracting Aerodynamic Derivatives from Flight Data

the partial differential of neural output is suggested and demonstrated. The performance of the two approaches (namely, the finite difference and the partial differential approach) is evaluated in terms of the estimated aerodynamic derivatives. It is shown that the two methods give comparable results; however, the partial differential approach provides first-order aerodynamic derivatives more accurately. Moreover, the results of the partial differentialapproacharecomparablewiththeresultsobtainedfromtheequationerrormethod.Moreover,gradient descent and scaled conjugate gradient methods are evaluated in terms of the mean square error for training the neural network. The proposed partial differential approach does not suffer from numerical instability and provides higher-orderderivatives;therefore,itcanbeextendedtothenonlinearaerodynamicregimeandtounstableaircraft.

Magnetocoulombic Attitude Control of Earth-Pointing Satellites

A magnetocoulombic satellite system is proposed and equations are developed to study the stability and control characteristics for a set of given initial conditions. An equation for charge required on coulomb shells is derived, which is used to find the available torque for actuating the satellite. It is shown that this system is similar to a satellite actuated by magnetic coils, except that the available torque vanishes for two different conditions. Controllability of the magnetocoulombic satellite rotating at a high angular velocity is proved. Moreover, stability is proved for a very high initial angular velocity, in addition to its exponential stability in the neighborhood of the origin. It is proved that the eigenvalues of the average control matrix, corresponding to different orientations and angular velocities of the satellite, in the body reference frame will converge to the same values as time tends to infinity. Moreover, it is proved that the eigenvalues of the average control matrix in the body ...

Dynamic Neural Units for Adaptive Magnetic Attitude Control of Spacecrafts

Aneural network based on dynamic neural units has the capability to handle any type of nonlinearity. In addition, it can adapt itself to parameter changes in real time. In this paper, such a dynamic neural network is used to design a controller through inverse modeling to address the attitude control of an Earth-pointing magnetically actuated spacecraft. Furthermore, normalization of weights of the dynamic neural units is proposed to ensure their convergence for proper learning. The dynamic neural controller developed in this paper, being adaptive, not only takes care of anyunknowndisturbance torques, but it is also robust and can adapt itself if there are any large changes in the parameters in theplant, such as themoments of inertia. It is shown, that the stabilization accuracy of the plant is better under the proposed neural controller as compared with a proportional-derivate controller. Proof of stability, for the dynamic neural units and the system as a whole, is also presented.

High Angle of Attack Parameter Estimation of Cascaded Fins Using Neural Network

Parameter estimation of cascaded fins at high angle of attack is carried out using a new neural-network approach. Two new lift models are proposed to overcome the problems faced by high angle-of-attack lift models based on Kirchhoff’s theory of flow separation. Lift modeling is carried out using a neural network for single and cascaded planar fins of airfoil and rectangular cross sections. Neural partial differentiation is applied to find the lift curve slope at zero angle of attack, which is more consistent compared to that based on average slope obtained using output-error method. Neural approach is applied to model the steady flow separation point on the fin. Separation and stall characteristic parameters are estimated using a new neural formulation. It is shown that the values of these parameters, obtained using the proposed neural method, better represent the observed lift data compared to those from the output-error method. Moreover, a modified equation is suggested to better model the steady separa...

Micro air vehicle path planning in fuzzy quadtree framework

Fuzzy quadtree framework has been utilized to develop a path planner for a fixed wing micro air vehicle (MAV). The fuzzy quadtree being computationally efficient can efficiently meet the computational requirements of a micro air vehicle, and therefore, does not require high capacity processor onboard. The proposed algorithm can provide optimal and safe path because it can avoid a pop up obstacle in real time while significantly reducing both the space and the time complexity. Some issues which are very pertinent to the MAV path planning like vehicle dimensions and safety measures for congested environment have been taken into account in the code developed. Besides, during the quadtree generation the constraint of turn rate kinematics of the MAV has been included.

Sensitivity analysis using neural network for estimating aircraft stability and control derivatives

This paper presents a method for aircraft parameter estimation using neural sensitivity analysis. The results are found to be superior to other ANN based methods.

FLIER: A Novel Sensor Fusion Algorithm

This paper proposes a novel sensor fusion algorithm to obtain instantaneous position and attitude estimates, which can either be used for aerial navigation or can be utilized to construct state feedbacks for camera stabilization. A divergence control strategy has also been formulated and the algorithm was embedded in real-time hardware. A comparative study between the proposed and conventional algorithm illustrates its efficacy.

A differential wideband bandpass filter

A wideband differential bandpass filter with enhanced common mode suppression and better out of band performance is presented in this paper. This filter consists of two back to back connected stepped impedance resonators (SIRs) coupled to each other through two dumbbell shaped slotline resonators. The size of the filter is 22 mm × 43 mm. The proposed filter has the differential passband bandwidth of 65 % centered around 4.85 GHz, with the common mode (CM) suppression better than 22 dB over the entire differential passband of the proposed filter. The insertion loss varies from 1.27 dB to 2.5 dB in the passband of the filter.

Finite-time continuous sliding mode magneto-coulombic satellite attitude control

A finite-time continuous sliding mode control is proposed and designed for the attitude control of an earth-pointing magneto-Coulombic satellite actuated using Coulomb shells and orbiting in a circular low earth orbit. A new nonlinear sliding manifold is proposed in terms of errors in quaternions and angular velocity and is used for average finite-time continuous sliding control formulation and closed loop stability analysis of averaged system dynamics. A real and finite-time continuous sliding control is proposed for real-time control of the system. and it is proved that this is reduced to the average finite-time continuous sliding control. Using the Lyapunov theorem, it is proved that the averaged closed loop system is globally stable and converges in finite time to the proposed sliding manifold in the presence of disturbance and destabilizing gravity gradient torque, which is followed by fast and finite-time convergence to the desired orientation and angular velocity (origin) along the sliding manifold. Simulation results of the sliding control are compared against those obtained using proportional-differential control, showing superiority of the sliding control in terms of less charge requirement and better convergence of states of the system. It is also shown that changing the distance between the shells and keeping moments of inertia constant does not alter the results qualitatively or quantitatively except for scaling of charge.

Dynamic Neural Units for an Adaptive Magnetic Attitude Control of a Satellite

Various controllers are available for the attitude control of magnetically actuated satellite including feedforward neural network. However, dynamic neural network has not been implemented for attitude control of satellite. Dynamic neural network based on dynamic neural units has the capability to handle any type of nonlinearity besides it can adapt itself in real time. The problem of attitude control for an earth pointing satellite using magnetic actuators and the adaptive neural controller, based on dynamic neural units through inverse modeling, has been addressed in this paper. Besides, weights normalization of dynamic neural units has been suggested to ensure their convergence for proper learning. Being adaptive, the proposed neural controller not only takes care of any unknown disturbance torque but also can adapt itself following the large parameter changes in the plant, and therefore, is robust to any unplanned change in the parameters of the plant such as moment of inertia. It has been shown that stabilization accuracy of the plant is better under neural controller as compared to the PD controller.