A Comparison of Two Methods Used to Deal with Saturation of Multiple, Redundant Aircraft Control Effectors
SummaryĀ (6 min read)
List of Figures
- Star represents the starting vertex of the AMS.
- The method of control vector scaling is implemented.
- The method of pitch prioritization is implemented.
- 42 5.7 Desired moment time histories plotted with attained moments for the offset left approach using the method of scaling the moment direction.
- . . . 46 5.10 Longitudinal and lateral stick time histories for the offset high left approach using both methods.
1.1 Background
- Classically, aircraft flight controls have been designed to use a single control effector to produce a desired moment about a specific rotational degree of freedom.
- These additional arise from freeing opposing moment generators and controlling them independently.
- Left and right horizontal tails can be used individually to produce roll, pitch or yaw moments.
- With this arrangement the number of control effectors on an aircraft could total close to 20.
- Most of these solutions are not realistic for an actual aircraft due to control limits created by the physical geometry of the effector and aerodynamic constraints on the surfaces.
1.2 Allocation Problem
- The distribution of controls to obtain a certain objective is the general allocation problem.
- These equations come from Newtonās Second Law of Motion and are frequently linearized about some reference condition and written xĢ = Ax+Bu (1.1).
- The problem is solved by finding a control vector to produce desired aircraft dynamics.
- This research examines such a method in dealing with these unattainable moments.
- Note that other difficulties arise in control allocation such as the fact that the linearized equation md =.
1.3 Allocation Methods
- A schematic of the general arrangement of a control system, with control law and control allocator, is presented in Figure 1.1.
- Dynamic inversion allows one to tailor the shape of the responses by calculating the forces and moments required of the controls.
- Some of these methods of control allocation are presented in Bordignon [4].
- Some methods such as daisy chaining and generalized inverses are often used [4].
- All of these methods provide a simple linear approach to solving the allocation.
1.4 Attainable Moments
- The term attainable refers to the set of moments that can be generated given the physical limits of the effectors.
- Control vector solutions that are within the physical limits are admissible solutions.
- The calculation of the attainable moment subset (AMS) is simple and defines individual facets of the polytope [5].
- One facet is defined by two controls free to vary, with all other controls being saturated.
- An additional constraint to the control allocation problem could involve the rate limits for each effector.
1.5 Research Objectives
- It was the purpose of this research to provide an insight into the problem of unattainable moments.
- A method of pitch prioritization was investigated to allocate controls in the case of an unattainable moment and compared with the method of scaling the moment direction.
- Pitch prioritization allocates controls while preserving the stability augmentation in the longitudinal axis of an unstable airframe.
- An F-18-like airframe was used as a testbed to perform the research and evaluate the performance of prioritization.
- Prioritization of a single objective during situations that ask for unattainable moments was.
2.1 Introduction
- Before the days of redundant control effectors and complex control laws the control allocation problem was simply dealt with by the use of mechanical linkages between the pilotās stick and pedals, and the control surfaces.
- Control systems since then have increased in complexity such as using an aileron/rudder interconnect to offset the adverse yaw effect during a turn.
- Even more complexity comes about when additional controls are developed and added to the airframe in an attempt to provide more maneuverability.
- Redundancy in aircraft controls has significantly altered the complexity of the control allocation problem.
- Several algorithms have been developed over the years to solve this problem.
2.2 Background
- Less complex algorithms such as generalized inverses have been used successfully in the past.
- The most popular of these control allocation algorithms is the Minimum Norm solution.
- This solution arranges the controls to produce the desired objectives while minimizing control energy.
- This method simply uses the pseudo-inverse solution to minimize the 2-norm of the control vector solution.
- There are several well-known ways to derive the pseudo-inverse solution that may be found in any text dealing with the underdetermined systems of linear 9.
2.3 Cascading Generalized Inverses
- One method called Cascading Generalized Inverse (CGI) was developed by Bordignon [4].
- This method uses the idea that if a generalized inverse commands a solution in which one or more controls exceed a limit, so long as not all are saturated, the controls are set at the limits, and the rest of the controls are redistributed to achieve the desired objectives.
- In other words, if a control is commanded past its limit, it is set at the limit, and the effects of the control at saturation are subtracted from the desired objectives.
- The method is repeated every time a control position is exceeded.
- If all controls become saturated, the moment is unattainable using this method.
2.4 Facet Search
- The allocation method used in this research is based on the original research by Durham [1,2] and later refined by Bordignon and Durham [3, 4].
- This method of control allocation takes the control effectiveness matrix B, the set of admissible controls, and some desired objective.
2.5 Bisecting Edge Searching Algorithm
- The method described above is more or less a brute force method of determining the intersection of md with the surface of the AMS.
- The method involved looking at a single facet defined by a pair of controls and determining whether the desired moment direction pointed toward this facet.
- There is no guarantee that the intersection would be found before the last facet was tested.
- The potential of searching the entire AMS meant that the number of floating point operations and computation time could be very high with the addition of control effectors.
- The reader is referred to Durham and Scalera [6, 7] for the theory behind the algorithm and a comparison of the performance with other allocators.
2.5.1 Two-Dimensional Problem
- The two dimensional problem is introduced as part of the solution to the three dimensional problem.
- That is, the three dimensional problem can be solved by systematically repeated solutions of the two dimensional projections of the three dimensional problem.
- B will then consist of two rows, with each element describing the effect each control has on the objectives.
- Figure 2.1 depicts an attainable moment subset, AMS, created from a four-control two-objective problem.
- The vertex with the maximum x-component is identified on the AMS.
2.5.2 Three-Dimensional Problem
- For the three-dimensional problem, consider md to consist of three objectives, causing the control effectiveness matrix to have three rows.
- From this view point the problem resembles the two-dimensional problem and is solved in the same way.
- There is an angle where the AMS can be rotated about the +x-axis such that the intersection of md, which is aligned with the +x-axis, with the AMS will lie exactly on a limb of the AMS as viewed from the +z-axis.
- Small perturbations in this angle will identify two edges that define the facet that contains the intersection.
- The rotation angle can not be calculated analytically, however, the angle is not needed, just.
3.2 F/A-18 Testbed
- The testbed airframe used in this research was based off of an F/A-18A airframe simulation.
- The greatest area in which the test-bed simulation differs from the original F/A-18A airframe is in the treatment of control effectors.
- These effectors were retained to make the simulation compatible with existing F/A-18 aircraft simulations.
- These controls are only used for initial trim and subsequent scheduling.
- The F/A18 in PA mode schedules leading-edge flaps based on angle-of-attack.
3.3 Airframe Simulation
- There are six files that were used in the simulation of the airframe: AERO.F, AEROPA.F, CONTROL.F, CONSTANTS.F, ENGINE.F, AND ALLOC.F.
- This section explains the purpose of each routine.
3.3.1 Aero.f
- AERO.F is an executive subroutine that calls subroutines based on the flight phase of the aircraft.
- In this specific case AERO.F calls AEROPA.
- The code then combines the aerodynamics from the non-linear scheduled/trimmed flight condition with the aerodynamics from the UGLOBAL array to produce the total aerodynamic forces and moments.
3.3.2 Aeropa.f
- AEROPA.F is taken from the F/A-18 simulation and modified slightly to mesh with AERO.F and to include proper control scheduling.
- This is the only code that gives the airframe F/A-18 like characteristics.
3.3.3 Control.f
- Stick and rudder pedal commands are taken as inputs and converted into an angle-of-attack command Ī±cmd, sideslip command Ī²cmd, and a roll rate command pcmd.
- These commands are input to a simple dynamic inversion control law that generates desired moments for the control allocation subroutine.
3.3.6 Alloc.f
- This code is the BESA control allocator that produces required control deflections for desired moments.
- The BESA method was explained in Chapter 2.
3.4 Simulation Environment
- Naval Air Systems Command provided the original F/A-18 simulation code in FORTRAN that was implemented on a Silicon Graphics Origin 2000TM Deskside System with two CPUās running at 180 MHz with 256 MB of RAM and a 4 GB disk.
- CASTLE is a 6-degree-of-freedom non-linear aircraft simulation architecture developed by Naval Air Warfare Centerās Manned Flight Simulator branch (MFS) [11].
- The visual scene is calligraphic depicting a dusk or night environment.
- The visual database includes scenes for a carrier landing approach, and a naval air station.
- The stick feel is produced using an electronic control loader and can be modified through software to produce a realistic feel of any actual aircraft stick.
3.5 Airframe Validity
- The airframe in this research was modeled after the F-18A airframe for ease of implementation on other users computers that already had the original F-18 airframe.
- The modifications that were made were simply added subroutines to transform the airframe as needed.
4.1 Introduction
- The method of prioritization has rarely been used for problems such as unattainable moments that lead to inadmissible control solutions.
- The use of pitch prioritization provides one method of dealing with unattainable moment commands.
- As previously stated, pitch axis prioritization was chosen because of the popularity of relaxing the static margin of an aircraft to reduce trim drag.
- To stabilize the airframe the control law requires a certain amount of pitching moment.
- Prioritization the pitching-moment requirement will ensure that the maximum amount of pitch can be attained during any maneuver.
4.2 Sizing the Attainable Moment Subset
- Control effectiveness was determined by the linearization of the F/A-18 aerodynamic database described in Chapter 3.
- The controls that were added to the testbed airframe were given the 25.
4.3 Method of Pitch Prioritization and Moment Direc-
- Tion Preservation Scaling the control solution vector preserves the moment direction while decreasing the magnitude.
- Preserving the moment direction will simply decrease the magnitude of the moment vector, using the solution on the boundary of the AMS.
- The light green arrow indicates the desired moment for a specific time during the maneuver.
- Prioritizing the pitch axis, the control effectors are able to provide the full amount of desired pitching moment.
5.1 Background
- Maneuvers that cause the control law to require unattainable moments and require the majority of the controls to be saturated are used to investigate the comparison between the two methods of dealing with unattainable moments.
- An important aspect of the results in this research was how the airplane felt to the pilot in addition to the hard data that was plotted.
- One evaluation pilot was used to fly all the data represented in this paper.
- Several flights were flown by other pilots, however those results are not included here.
- The test pilot has a background in Navy fighters, carrier suitability, and flight testing.
5.2 Offset Carrier Approach maneuver
- Aerodynamic control effectiveness reduces with a decrease in the dynamic pressure.
- As a result, an aircraft requires larger control deflections to produce the desired amount of force and moment.
- An offset powered approach to an aircraft carrier was chosen for evaluation in this research since it provided the low dynamic pressure flight condition as well as large stick inputs and a high pilot workload.
5.2.1 Offset Carrier Approach maneuver: Description
- For this maneuver the static margin of the airframe was relaxed so that the airframe was unstable in the longitudinal axis.
- The control law kept the airframe stable during normal flight with small stick inputs.
- Offset Left - Airframe was offset to the left of centerline eight degrees and required more lateral maneuvering than pitch to correct the offset.
- The approach was offset eight degrees left of centerline and three degrees above the desired glide-slope.
- The pilot began the maneuver at the same point.
5.2.2 Offset Left Carrier Approach maneuver: Results
- The following data represents two different offset left approaches flown from the same initial conditions.
- One approach implemented the method of scaling the moment direction while the other used the method of pitch prioritization.
- Figure 5.2 plots the saturation of each method for an offset left maneuver.
- The yellow hatching indicates when the controls are saturated.
- The area of saturation for the lateral-directional requirement occured when the pitch requirement was near saturation indicating that most of the controls had been used for pitching moment.
5.2.3 Offset High Left Carrier Approach maneuver: Results
- An additional offset approach was used to further evaluate the two methods of dealing with unattainable moments.
- The prioritization mode was able to achieve all of the pitching moment without saturating any of the controls on the airframe.
- The saturation plot for the scaling method ends early because the airframe diverged in pitch as soon as the corrective maneuver was applied.
- These percentages are calculated as previously explained in section 5.2.2.
- Figure 5.10 plots the stick movement throughout the entire flight using both methods.
5.2.4 Offset Carrier Approach maneuvers: Pilot Interpretation
- The main trend seen in all the test flights was the pilot adaptation to how the aircraft handled.
- This learning curve made it difficult to investigate all of the differences between the two different methods.
- The one disconcerting tendency of the airframe while in pitch prioritization mode was the continuously changing amount of lateral control power during flight.
- The method of scaling the desired moment direction gave the aircraft an unstable feel to the pilot during any kind of aggressive maneuver.
- Overall, the airframe with pitch prioritization was āmore smoothā compared to the flight with the method of scaling the moment direction implemented.
Summary and Conclusions
- A comparison of two methods to deal with unattainable moments in an aircraft was performed using a testbed airframe that resembled an F/A-18 with a highly redundant control effector suite.
- Control allocation utilizing the Bisecting, Edge Searching Algorithm was implemented on this airframe and a dynamic inversion control law was used to produce the desired aircraft handling qualities.
- In extreme unstable cases prioritizing the pitch axis produced favorable characteristics with respect to the alternative of preserving the moment direction by scaling the control solution vector that the allocator produces.
- Representative maneuvers that offered real-time evaluation of the two methods were flown with a pilot-in-the-loop.
- The validity of these tasks were subject to pilot compensation, to the point where the pilot was required to remove any learning of how the aircraft feels from the desired task.
Did you find this useful? Give us your feedback
Citations
2Ā citations
Cites background from "A Comparison of Two Methods Used to..."
...(Durham 1993; 1994a; 1994b; Bordingnon and Durham 1995; Durham 2001; Nelson and Durham 2001) have focused on optimal methods in control allocation in order to maximize attainable torque and therefore improving performance....
[...]
...Durham and her colleagues (Scalera and Durham 1999; Nelson and Durham 2001; Beck 2002; Durham et al. 2017) have compared different allocation algorithms in terms of maximum available torque space....
[...]
References
2,837Ā citations
"A Comparison of Two Methods Used to..." refers methods in this paper
...The original control set was linearized in a PA mode using a linearization method borrowed from Stevens and Lewis [12]....
[...]
527Ā citations
"A Comparison of Two Methods Used to..." refers methods in this paper
...Even with additional methods to improve the solution, it is shown that these methods can not obtain the maximum moment generating capabilities [1]....
[...]
...I would like to thank my advisor Dr. Wayne Durham for his persistence and patience with me and my academic career....
[...]
...The research done by Durham and Scalera focused on the restoring aspects of allocation as well as control reconfiguration in the event of a surface failure....
[...]
...Durham and Scalera demonstrate that this method of control allocation provides a guarantee that the full moment capability is utilized in the control solution....
[...]
...Durham [6] introduced the Bisecting, Edge-Searching Algorithm (BESA) to efficiently allocate controls....
[...]
208Ā citations
"A Comparison of Two Methods Used to..." refers background or methods in this paper
...To Mike Henry, Kevin Waclawicz, Dan Hart, Todd Norell, Trevor Wallace, Roger Beck, Josh Durham, Bill Oetjens, and others that mean so much to me....
[...]
...A second method developed by Durham is presented in several papers [1ā4]....
[...]
...The allocation method used in this research is based on the original research by Durham [1,2] and later refined by Bordignon and Durham [3, 4]....
[...]
...Further work by Durham and Scalera [7] explores this algorithm and compares the performance to an existing F-15 ACTIVE (Advanced Control Technology for Integrated VEhicles) allocator using real-time pilot-in-the-loop simulations....
[...]
...I would like to thank my advisor Dr. Wayne Durham for his persistence and patience with me and my academic career....
[...]
141Ā citations
"A Comparison of Two Methods Used to..." refers methods in this paper
...I would like to thank my advisor Dr. Wayne Durham for his persistence and patience with me and my academic career....
[...]
...The research done by Durham and Scalera focused on the restoring aspects of allocation as well as control reconfiguration in the event of a surface failure....
[...]
...Durham and Scalera demonstrate that this method of control allocation provides a guarantee that the full moment capability is utilized in the control solution....
[...]
...Durham [6] introduced the Bisecting, Edge-Searching Algorithm (BESA) to efficiently allocate controls....
[...]
...A second method developed by Durham is presented in several papers [1ā4]....
[...]
137Ā citations
"A Comparison of Two Methods Used to..." refers background or methods in this paper
...This method is described in Bordignon [4]....
[...]
...It has been shown that these generalized inverse routines, including minimum norm solutions and daisy chaining, cannot guarantee admissible control solutions for all attainable moments [4]....
[...]
...Some of these methods of control allocation are presented in Bordignon [4]....
[...]
...I would like to thank my advisor Dr. Wayne Durham for his persistence and patience with me and my academic career....
[...]
...The research done by Durham and Scalera focused on the restoring aspects of allocation as well as control reconfiguration in the event of a surface failure....
[...]