Michael Creagh

Adaptive Control for a Hypersonic Glider using Parameter Feedback from System Identification

The design and simulation of an adaptive longitudinal control system for a Mach 8 hypersonic glider is detailed. The vehicle is equipped with two elevons for control surfaces. It is required to perform a flight-path angle pull-up manoeuvre during the reentry phase of a parabolic trajectory. This is achieved with a normal acceleration longitudinal controller. The controller uses the pole-placement technique to calculate gains in real time. The gain calculations are based on a set of error-minimised model parameters, which are a fusion of real-time measured parameters, found using recursive least squares and stored look-up table parameters. In addition, a first order filter is used to update the look-up table parameters so that future control system gain calculations are not dependent on the presence of system excitation. It is shown that such a system can provide a small performance improvement in overshoot. One fundamental problem with implementation of the system in real time is the lag between the fused parameter estimates and when such parameters are required by the autopilot. Typically the autopilot operates at a higher frequency than the fused parameter estimates and thus corrects a disturbance due to real parameter change before suitable update information is received for gain calculation. The system is suggested as a useful diagnostic tool for experimental flights of this nature, or for real-time use on a more suitable dynamic system.

L1 adaptive control augmentation configuration for a hypersonic glider in the presence of uncertainties

The descent longitudinal trajectory control methodology of a hypersonic glider is presented. The overall desired trajectory is presented, followed by a basic analysis of the control methodology used to carry out a pull up maneuver; bringing the glider from a descent trajectory to horizontal flight. A dynamic pole placement controller is implemented to carry out this maneuver. This controller is augmented with an L1 adaptive controller to cancel out the matched uncertainties. Only the longitudinal dynamics are considered in this analysis. The fundamental differences between only the pole placement and the L1 augmented pole placement with the L1 adaptive controller are presented in terms of robust stability and robust performance. When the INS/GPS module is turned on, of the altitude error, flight path angle error and angle of attack error are bounded compared to the unfiltered values. The dynamic pole placement controller is shown to have robustness in the presence of time invariant, Gaussian and time varying errors in the aerodynamic coefficients. The augmentation set up improves the performance of the baseline controller in the presence of these uncertainties. With the help of Monte Carlo Simulations the performance, robustness and stability bounds of the baseline and the L1 adaptive augmented controller are presented and are compared to the performance of the pole placement controller. All the simulations and implementations are carried out in CADAC++ which is written in C++. The system presented is a nonminimum phase system. The uncertainties presented in this paper are matched uncertainties. The main contribution of this paper lies in the application and determining robustness and performance of a piecewise constant L1 augmentation setup to a Linear Time Varying (LTV) System.

The SCRAMSPACE I scramjet flight design and construction

The design activities of the development of the SCRAMSPACE I scramjet-powered free-flight experiment are described in this paper. The objectives of this flight are first described together with the definition of the primary, secondary and tertiary experiments. The Scramjet configuration studied is first discussed together with the rocket motor system selected for this flight. The different flight sequences are then explained, highlighting the SCRAMSPACE I free-flyer separation and re-orientation procedures. A design trade-off study is then described considering vehicle stability, packaging, thermo-structural analysis and trajectory, discussing the alignment of the predicted performance with the mission scientific requirements. The global system architecture and instrumentation of the vehicle are then explained. The conclusions of this design phase are that a vehicle design has been produced which is able to meet the mission scientific goals and the procurement & construction of the vehicle are ongoing.

The SCRAMSPACE I Hypersonic Flight Experiment Feasibility Study

The conceptual design activities of the development of the SCRAMSPACE I scramjet-powered free-flight experiment are described in this paper. The objectives of this flight are first described together with the definition of the primary, secondary and tertiary experiments. The Scramjet configuration studied is first discussed together with the rocket motor system selected for this flight. The different flight sequences are then explained, highlighting the SCRAMSPACE I free-flyer separation and re-orientation procedures. A trajectory analysis and the derived thermo-structural analysis are presented, discussing the alignment of the predicted performance with the mission scientific requirements. The global system architecture and instrumentation of the vehicle are then explained. The conclusions of these Phase A technical activities are that the presented mission concept is in line with the SCRAMSPACE I scientific requirements and no major technical difficulties have been identified.

An attitude control strategy using four cold gas thrusters for the SCRAMSPACE 1 experiment

The design and simulation of an attitude control strategy using four cold gas thrusters is detailed. The system is designed to re-orient the hypersonic experimental vehicle SCRAMSPACE 1 outside the atmosphere so that on re-entry at 80 km altitude, it is within a 10.0 degree total angle of attack. This constraint is equates to 2.5 degrees total angle of attack at 32 km altitude where a scramjet is to be tested at Mach 8. Monte-Carlo simulations show that the final attitude accuracy at 80 km is constrained to a 1 degree error in pitch and yaw about the velocity vector. The reaction control strategy is to step through a number of control modes: nutational damping, de-spin and coarse roll angle acquisition, fine attitude tracking after de-spin, coarse pitch-over, midcourse fine tracking, terminal fine tracking and re-entry rate damping. Higher level logic allows for measurement of thruster force and re-starts of the algorithm. Planar Containment Guidance is developed and used during the pitch-over to constrain the vehicle to a plane whilst maintaining a constant pitch-rate. Re-entry rate damping is shown to widen the allowable angle of attack range at 80 km altitude to 19 degrees. The reaction control system presented is robust to trigonometric singularities and mathematical ambiguities.

An alternative attitude control strategy for SCRAMSPACE 1 experiment

The current reaction control strategy for SCRAMSPACE 1 involves an exo-atmospheric reorientation using attitude and rate feedback from the Inertial Navigation System. An alternative design methodology of an attitude control system is presented that only uses rate feedback. This eliminates the need for an attitude determination system altogether. This methodology is designed to re-orient the hypersonic experimental vehicle SCRAMSPACE 1 by using rate damping and natural aerodynamic moments. It is desired that the re-orientation produces a 2.5 degrees angle of attack at 32km altitude on the decent trajectory. This angle of attack is necessary in order to test the scramjet at Mach 8. The re-entry rate damping is initialized at an altitude of 55km on the descent trajectory. The worst case scenario angle of attack of 150 degrees is taken as the initial conditions at 200km altitude. For the worst case scenario, thrusters of size 10N, 15N, 20N, 25N and 30N are able to acquire the desired angle of attack. The final angles of attack for all the aforementioned thruster sizes are less than 1 deg. The mean of the angle of attack during the experimental window is under 0.5 degrees for all tested cases of thruster force, initial spin rate and initial angle of attack. However, when a sensible fuel restriction of 5kg is placed, the recommended thruster size is 20N.

A Kalman-Filter Based Inertial Navigation System Processor for the SCRAMSPACE 1 Hypersonic Flight Experiment

The design of a navigation processor for the SCRAMSPACE 1 hypersonic flight experiment is detailed. Two cascaded extended Kalman filters provide a fusion of gyroscope, accelerometer, magnetometer, GPS and aerodynamic database sources for vehicle state estimation. Two quantities of great interest for SCRAMSPACE 1 are angle of attack and side-slip angle. The navigation processor performs estimations in two stages. The first stage tracks the vehicles navigation states with a commercial grade Inertial Measurement Unit (IMU)/GPS combination, starting at the conclusion of a despin manoeuvre. It is initialised by a navigation- grade Inertial Navigation System (INS). A 25-run Monte Carlo sweep shows that without a navigation filter, the first stage cannot hold attitude via raw gyroscope integration over the 468 seconds necessary. The filtered estimate has a 72% chance of holding attitude over this time, with 5.5, 6.3 and 14.9 degrees of standard deviation of roll pitch and yaw error. A prototype proof-ofconcept attitude determination system is shown to confirm the performance in reduction of attitude drift, compared with that of integrating the gyroscope measurements directly. In a stationary test over 240 seconds, the integrated gyroscope solution drifts 160 degrees in roll, 12 degrees in pitch and 47 degrees in yaw. The magnetometer-integrated solution drifts 5 degrees in roll, 1 degree in pitch and 10 degrees in yaw. In a dynamic test over 64 seconds, in which pure rotations are induced, the integrated gyroscope solution drifts 40 degrees in roll, 10 degrees in pitch and 20 degrees in yaw. The second stage of the experiment utilises known vehicle dynamics to estimate angle of attack and sideslip on re-entry to the atmosphere. The estimations are shown to be within 0.1 degrees of the truth, even when the INS-computed relative wind angles are an order of magnitude out and the wrong sign.

H-infinity Control for Attitude Manoeuvres of a Spinning Asymmetric Vehicle

The design and simulation of an H-infinity controller for the purposes of attitude guidance for a spinning asymmetric vehicle is detailed. The controller takes pitch and yaw rate commands from an attitude guidance law and issues flap deflection commands. The modelled vehicle has a rectangular body cross-section and spins at 1.5 Hz with independent pitch and yaw flap sets. The simulation is performed in a six degree-offreedom program. The vehicle is shown to perform an attitude manoeuvre pitching up 6o and yawing to the right 4o with a settling time of less than 0.25 seconds. A previous implementation of the guidance law used two SISO autopilots, which resulted in a settling time of approximately 0.5 seconds. The H-infinity controller demonstrates a significant improvement in settling time performance and is a potential candidate for implementation in a flight experiment, despite an H-infinity norm of 100.1 at a frequency of 0.

Comparison of reference frames in the linearisation of flight dynamics for spinning vehicles

The derivation and linearization of equations of motion for spinning vehicles is presented in this paper. Two reference frames are used for the basis of the derivations. The non-rotating body frame and the rotating body frame are used to establish state-space representations. An eigenvalue and eigenvector analysis presents the modes of flight for each reference frame using an asymmetrical body. The non-rotating body frame equations of motion results in five roots; three stable real poles and two identical complex conjugate pairs. This frame is found to be useful for control and linearised simulation of axisymmetrical vehicles, where the control mechanism can be decoupled from the spinning body. This frame is not useful for modelling an asymmetrical spinning vehicle. The analysis of the rotating body frame equations results in six non-zero roots; three real, stable poles, two different complex conjugate pairs and a marginally unstable real pole. Two of the real poles are dependant on the reference roll angle chosen for the analysis. It was found that the rotating body frame is appropriate for controller design, where the controller input is a requested angle of attack, sideslip angle, pitch rate or yaw rate. It is not applicable to a controller with a requested pitch or yaw angle.

Attitude Guidance for Spinning Vehicles with Independent Pitch and Yaw Control

The design and simulation of an attitude guidance and control scheme for a spinning aerospace vehicle is detailed. The guidance law issues pitch and yaw rate commands to two single-input, single-output controllers, which in turn issue flap deflection commands. A non-axisymmetrical vehicle spinning at 1.5 Hz with independent pitch and yaw flap sets is simulated in a six degree-of-freedom simulation. It is shown to perform attitude manoeuvres using either one or both sets of flaps. Using one set of flaps results in a settling time of 1.5 s, whilst the settling time for both flap sets is 0.7 s. The guidance law has a number of possible applications. These include munitions guidance, stability augmentation and redundancy, alternatives in fin configurations for vehicles and reduction in number of necessary control surfaces.