Breaking the Cycle: Avoiding Aircraft-Pilot Coupling

• Introduction
• Basics
• Examples
• Sources

Introduction

PIO/APC are important to understand to maintaining the safety of flight control systems. Controller-aircraft coupling has led to many incidents and accidents and program time and cost overruns due to these issues are the norm rather than the exception. Pilot-induced oscillations are a failure of the design process. When designers simulate a new control system if the model used for control design doesn’t match the phase behavior of the actual aircraft, a PIO condition could be encountered during flight testing. These coupling oscillation events have been present in many aircraft development programs in US and Europe as we will see in the further examples. By the 90’s PIO/APC events accounted for more aircraft incidents and accidents than structural failure.

This article is for people who have a basic grasp of control theory and aerodynamics and want to learn more about the PIO/APC phenomenon and how it might jump out and ruin your aircraft development program.

We will go heavy into examples of PIO on different aircraft. What caused them, and how they were fixed. This is by no means an exhaustive list, but I tried to find examples that people might recognize as well as examples that included video footage. I wanted to give a quick summary for each of examples. I will not reproduce many of the figures, equations, or analysis, but I will include a source list below each of the summaries.

PIO/APC Basics

Pilot-Induced Oscillations are negative feedback between the pilot and the aircraft control system. First I want to make clear that this is really a misnomer as the root cause of these undesired dynamic behaviors is poor control system design and not poor piloting skills. Most of the literature still refers to the phenomenon as a pilot-induced oscillation, but the term aircraft-pilot coupling may be more representative of the phenomenon. Throughout this article I will use the two terms interchangeably based on the nomenclature of the source material.

Causes

A PIO results when the closed-loop pilot/vehicle system oscillates due to a loss in asymptotic stability. This happens when any pole or complex pair of poles of a system cross from the left-hand plane to the right-hand plane of the S-domain. The most important factor in handling qualities during landing is control system lags, which could create a PIO. They can also be caused by an excessive deadband in the control system as the lack of response has the same phase-lag increasing effect.

The typical pilot response time is 1-2 seconds, so divergent oscillations around this frequency may create a PIO incident if the aircraft response becomes 180 degrees out of phase with the pilot control inputs. Susceptible aircraft have either high stick sensitivity, excessive time delay, or phase lag. Unstable response modes and unusual coupling responses also contribute to the onset of PIO. Aircraft that have a crossover frequency less than 4 rad/s will have a problem if the response grows.

Neutral stability PIOs often develop at crossover frequencies between 0.5 and 1 Hz.When the rate limiter is saturated, the resulting phase shift reduces the closed-loop stability margins and increases the risk of a PIO. Aircraft with sidestick controllers can be more prone to adverse coupled oscilations. As we will see in the examples, high-gain situations like refueling or aggressive maneuvers can cause the pilot to tighten up their control to the point where a PIO-prone aircraft enters into unstable oscillations.

Effects

As a pilot manipulates the controls to achieve tighter control, it causes the same effect as increasing the gain of the controller(the pilot). The oscillations typically appear suddenly and without much or any warning for the pilot. Everything seems fine until they hit the “cliff” . Some less serious PIO problems are bobble and tracking oscillations.

Fixes

As the most important factor in handling qualities during landing is control system lags, the goal of all fixes is to reduce these lags. You can use a frequency-dependent gain reduction filter to eliminate PIO tendencies that are not able to be prevented in any other way. Typically, most PIO recovery or mitigation methods end up reducing the aircraft’s agility. By reducing the command path gain near the PIO frequencies there is less of a risk of control surface saturation. Control angles commanded at these frequencies should be considerably less than many other flight conditions.

Software limits can also placed on the control servo commands to prevent the hydraulic servos from rate limiting. Analysis methods deal with the shape of the transfer functions near the -180 phase angle where the crossover frequency may occur (1-6 rad/s). When designing a PIO-resistant system, the PIO frequency cannot be too high, the PIO gain cannot be too low, the phase delay cannot be too small, and the large amplitude responses cannot be limited too much.

Examples

• F-8
• YF-16
• YF-22
• F-4
• C-17
• Space Shuttle
• M2-F2
• SAAB Gripen
• YF-12
• B-2

F-8 DFBW

The F-8 aircraft that was used as a testbed for fly-by-wire control systems developed a PIO during one of the landings. This occurred at Dryden in 1978 as a part of tests of flight control system delays for testing the Space Shuttle control systems. For this particular flight, the approach speed was 265 kts with an added control delay of 100ms. During the recovery from the touch-and-go, the onboard software switched from the Command Augmentation System (CAS) mode to the Direct Mode. This removed any feedback stabilization, causing the aircraft to enter into a dangerous oscillation very close to the ground.

• [1] NATO, Ed., Flight Vehicle Integration Panel Workshop on Pilot Induced Oscillations: = (Atelier sur le Pompage Piloté). in AGARD advisory report, no. 335. Neuilly-sur-Seine: AGARD, 1995
• [2] “Overcoming pilot-induced oscillations in the space shuttle,” IEEE Annals Hist. Comput., vol. 21, no. 4, pp. 69–70, Oct. 1999, doi: 10.1109/MAHC.1999.801535.

YF-16 First Flight

During a in 1974 high-speed taxi test in Fort Worth ,a PIO caused the YF-16 test pilot to make an unplanned first flight. The purpose of the taxi test was to perform a function check of various aircraft systems. The aircraft then entered a series of roll oscillations. There were 10 oscillations in 14.3 seconds involving full lateral roll commands (50 deg/s) which hit the rate and position limits. This oscillation caused the right stabilator and the left wingtip missile to strike the runway.

Due to the deviation from the runway heading, the pilot, Phil Oestericher, made the decision to fly out of the condition saving the aircraft. After this incident, it was discovered that the flaperon roll gain was higher than necessary. As a result, the trailing-edge flaperon gains were reduced. On further review, this condition was first observed in the NT-33A in-flight simulations but was overlooked.

• [1] B. J. Goszkowicz, “Sidestick Controllers During High Gain Tasks”.
• [2] RTO-TR-029
• [3] NATO, Ed., Flight Vehicle Integration Panel Workshop on Pilot Induced Oscillations: = (Atelier sur le Pompage Piloté). in AGARD advisory report, no. 335. Neuilly-sur-Seine: AGARD, 1995

YF-22 Flight Test

The YF-22 was involved in a PIO during a touch-and-go landing that led to a crash on April 25, 1992. This PIO was caused by time delays from rate and position saturation of the control surfaces and mode switching in the flight control system that changed the control stick gains. The landing gear was retracted at the same time as the afterburner was ignited to initiate a go-around. At this time the pilot was holding nose up trim that required operating near the forward deflection limits for the stick.

When the landing gear was retracted, it activated a mode switch in the flight control laws. With the landing gear up the command gradient and thrust vectoring allowed for greater commanded pitch rates. This caused the pilot to reverse the stick and enter into the oscillation where rate limiting of the horizontal tail and thrust-vectoring nozzles occurred. The oscillations started 40 feet above the runway and the aircraft to impact the runway after 4-5 oscillations. The graph below shows the gradient change between the two modes. Note that in the gear-down mode the thrust vectoring was turned off, but in the gear-up mode thrust vectoring is enabled. If we look at the graph of the pitch command gradient(dotted line), we can see that there is a sharp nose-down slope past the 10lb forward stick point. This gradient was developed for nose down recoveries from the post-stall AOA region but it was not adjusted for landing and go-around behaviors.

Upon analysis against the Smith-Geddes handling criterion, the F-22 was found to be PIO prone in the flaps-up up-and-away control law at low speeds. Any type of rate limiting would increase the time delay and make an oscillation much more likely. To remedy the system, the design goals for the control anticipation parameter and damping ratio are nearly identical along with the stick force gradient. Also the thrust vectoring is kept on for the gear down flight control law as not to change the response. This was similar to the earlier F-8 PIO that happened during the shuttle testing as they were both caused by the mode-switching of the control law.

• [1] RTO-TR-029
• [2] D. M. R. Anderson and A. B. Page, “UNIFIED PILOT-INDUCED OSCILLATION THEORY VOLUME III: PIO ANALYSIS USING MULTIVARIABLE METHODS”.
• [3] NATO, Ed., Flight Vehicle Integration Panel Workshop on Pilot Induced Oscillations: = (Atelier sur le Pompage Piloté). in AGARD advisory report, no. 335. Neuilly-sur-Seine: AGARD, 1995
• [4] J. Harris and G. Black, “F-22 control law development and flying qualities,” in 21st Atmospheric Flight Mechanics Conference, San Diego,CA,U.S.A.: American Institute of Aeronautics and Astronautics, Jul. 1996. doi: 10.2514/6.1996-3379.

F-4 Phantom accident

During an attempt to set a low-level speed record, the F-4 phantom experienced a PIO followed by a subsequent breakup of the aircraft as structural loads were increased beyond design limits. This accident occurred in white sands on May 18 1961. At 200ft altitude and Mach 1.1 the rapid onset of unstable pitch dynamics reached -4 and +14 Gs after only 3 oscillations. It is not clear what caused the oscillations, but one source lists a pitch damper failure as the cause of the accident. Others say that the pilot had previously use the procedure of holding 20lb of push forces during low-altitude high-speed runs to create a nose-up safety margin if he relaxed pressure on the controls due to distraction. Regardless of the cause, PIO tendencies were pronounced if the stick forces were not completely trimmed out.

• [1] RTO-TR-029
• [2] “18 May 1961 | This Day in Aviation.” Accessed: Sep. 30, 2024. [Online]. Available: https://www.thisdayinaviation.com/18-may-1961/

C-17

The C-17 PIO was caused by an attempt to use the fly-by-wire control system to increase performance to give fighter-like responsiveness. This led to a control system prone to rate-sustained lateral and pitch PIOs. These oscillations were exacerbated by low elevator(11 deg/s) rates and moderate aileron rates(40 deg/s). Excitation of the wing structural frequencies coupled with the stick inputs gave a “roll-ratcheting” effect. Most of these oscillations occurred during the landing phase and due to large control inputs causing instantaneous rate saturation.Some oscillations occurred due to normal control inputs during high gain tracking tasks such as refueling. The Aileron rate saturations could occur with only 25% of the available lateral stick throw because the roll command gain schedule resulted in a doubling of the gain at around 200 kts.. The left images shows the original roll control law and the right image shows the response.

An outer loop electronic rate limiter in the command path was used to prevent saturation of the inner loop. The minimizing of actuator rate limits was the most significant change that reduced the PIO tendency. If you look at the roll control schematic, were a total of five 2nd-order structural filters in the roll-rate feedback path that dded too much time delay. These were reduced to just two 2nd-order structural filters in the upgraded system. Lateral-Directional axes needed different changes to reduce the sensitivity, slow down the roll mode, and increase the Dutch-Roll damping. Reducing the control sensitivity was required as simply reducing the phase lag didn’t completely resolve the issue.

• [1] RTO-TR-029
• [2] NATO, Ed., Flight Vehicle Integration Panel Workshop on Pilot Induced Oscillations: = (Atelier sur le Pompage Piloté). in AGARD advisory report, no. 335. Neuilly-sur-Seine: AGARD, 1995.
• [3] “Minimizing pilot-induced-oscillation susceptibility during C-17 development.” Accessed: Jul. 10, 2024. [Online]. Available: https://arc.aiaa.org/doi/epdf/10.2514/6.1997-3497

Space Shuttle

ALT-5

The first PIO occurred when landing at Edwards in 1977 during the approach test and landing flight. During approach and landing, the shuttle had a touchdown point 500 feet beyond the runway threshold. A 3.5 rad/s PIO started 7 seconds before the first skip with continuous elevon limiting. After the first skip a Lateral PIO occurred that was followed by other skipping and hopping motions caused by overcontrol. This event was caused by a 270ms effective time delay in the flight control system loop.

STS-3

During STS-3, a cross-wind caused an overshoot of the final approach with a late adjustment. This put the aircraft into a high-gain situation on short final which caused a PIO to start just as the orbiter touched down. This 1/5 cycle oscillation that caused the nose to slam down on the runway. The estimated time delay for the control system was between 200 and 300 ms.

These PIO tendencies were corrected by using a frequency-dependent PIO suppression filter that reduced the gain as the frequency of the pilot input increased.

• [1] RTO-TR-029
• [2] T. T. IVlyers, D. E. Johnston, and D. T. McRuer, “Space Shuttle Flying Qualities and :Flight ~Control System Assessment Stu«:ly — Phase II”.
• [3] NATO, Ed., Flight Vehicle Integration Panel Workshop on Pilot Induced Oscillations: = (Atelier sur le Pompage Piloté). in AGARD advisory report, no. 335. Neuilly-sur-Seine: AGARD, 1995.
• [4] “Overcoming pilot-induced oscillations in the space shuttle,” IEEE Annals Hist. Comput., vol. 21, no. 4, pp. 69–70, Oct. 1999, doi: 10.1109/MAHC.1999.801535.
• [5] S. Pilot-Lndn, “NASA Technical Memorandum 81349”.

M2F2

These severe PIOs occurred on 3 occasions in the low AOA, final approach configuration. The aircraft dynamics formed a coupled roll-spiral mode that when added to the closed-loop pilot dynamics, generated an instability. the open-loop root locus plot is shown below.

We can see that at angles of attack around 2 degrees this coupled roll-spiral mode would become unstable. This roll-spiral coupling was attributed to the large effective dihedral, negative AOA region, large positive yawing moment due to roll rate, low natural roll damping, and large adverse yaw due to aileron deflection. At AOAs around 0, high aileron-rudder-connection gains would lead to the oscillation. With the SAS and the ARI included the bank angle to aileron transfer function numerator has second-order zeros in the left half plane and the denominator becomes 2 second-order factors.[^4] This means that the roll-mode has coupled with the spiral mode to form a long-period oscillation. With the SAS system on, the roll and spiral modes were initially separated for higher angles of attack, but again, they merged and passed into the unstable plane around -2 degrees.

• [1] R. W. Kempel, “Analysis of a coupled roll-spiral-mode, pilot-induced oscillation experienced with the M2-F2 lifting body”.

Saab JAS-39 Gripen

The test aircraft crashed in 1989 due to a PIO which was partially caused by the control surface servo rate limits. A response to lateral turbulence caused the rate limits to engage, which increased the control system delay. The oscillation initially developed in roll, then in pitch, causing a subsequent crash. The commanded roll rates saturated the elevons in the pitch axis. The “mini-stick” was a contributing factor as it had a skewed axis and could demand full control with very small movements. Another contributing factor was the augmentation of the dihedral effect by the control system, which made the aircraft sensitive in the roll axis.

During a 1993 public demonstration in Stockholm, another oscillation incident started as the pilot made an aggressive wings-level roll for a pass in front of the crowd. During this maneuver the roll-rate limits were engaged which caused the aircraft to roll more than expected. The rate limiting caused the time delay from the pitch control input to the pitch acceleration response to increase from 100ms to 800 ms. With the reverse roll control input applied, the rate limits were hit again and the aircraft entered a PIO. Due to the time delay caused by the rate-limiting, the inner pitch stabilization loop was no longer effective and the aircraft departed controlled flight.

The solution to this problem was to use phase compensating rate limiters. The feedback phase compensation for the controller was based on anti-windup methods. Low-pass filters were used to eliminate the biases and issues caused by high-frequency roll-off. This phase compensation technique was used in place of normal rate limiters in the control system of the production fighter.

• [1] “Flight Control Law Design: An Industry Perspective – ppt video online download.” Accessed: Feb. 11, 2023. [Online]. Available: https://slideplayer.com/slide/5738737/
• [2] G. Stein, “The practical, physical (and sometimes dangerous) consequences of control must be respected, and the underlying principles must be clearly and well taught.,” IEEE Control Systems Magazine, 2003.
• [3] RTO-TR-029
• [4] L. Rundqwist and R. Hillgren, “Phase compensation of rate limiters in JAS 39 Gripen,” in 21st Atmospheric Flight Mechanics Conference, San Diego,CA,U.S.A.: American Institute of Aeronautics and Astronautics, Jul. 1996. doi: 10.2514/6.1996-3368.
• [5] NATO, Ed., Flight Vehicle Integration Panel Workshop on Pilot Induced Oscillations: = (Atelier sur le Pompage Piloté). in AGARD advisory report, no. 335. Neuilly-sur-Seine: AGARD, 1995.
• [6] R. H. Saab ab, “CAREFREE MANOEUVRING AND AUTOMATIC RETURN TO NORMAL FLIGHT ENVELOPE JAS 39 GRIPEN.,” IFAC Proceedings Volumes, vol. 40, no. 7, pp. 115–120, Jan. 2007, doi: 10.3182/20070625-5-FR-2916.00021.

YF-12

The YF-12 had 2 types of PIO. The first PIO occurred when the pilot interacted with the SAS rate limits short-period poles, and structural bending zeros. The combination of these elements at 1Hz to created a small-amplitude PIO. These mostly occurred during refueling as the aircraft is PIO prone during refueling due to the higher pilot gain at a frequency around 1 Hz.

The large-amplitude more serious PIO experienced -1 to 3g excursions with a frequency of 0.5 Hz. In one case an overshoot in longitudinal trim due to a faulty trim switch caused an oscillation that almost exceeded the aircraft G-limits. This condition caused the pilot to tighten up their control, creating the conditions for the oscillation.

Here is a time trace of the control command during a PIO event showing rate and positing limiting for the YF-12 during a PIO

This is a good example as we can see the characteristic of position limiting where it looks like the controls are “cut off” at a negative threshold. The rate limiting appears in the time trace as constant slope lines creating a very angular looking control trace.

• [1] J. W. Smith and D. T. Berry, “ANALYSIS OF LONGITUDINAL PILOT-INDUCED OSCILLATION TENDENCIES OF YF-12 AIRCRAFT”.

B-2

The title of the video is misleading. This is not an example of aeroelastic flutter, instead it is a unique coupling of the structure and control system that can be induced by a commanded pitch doublet at velocities outside the operational envelope. During flight expansion, the aircraft was flow to M0.82, but in operation, the envelope is limited to Mach 0.8. This mode was called a Residual Pitch Oscillation and it occurred at true air speeds just outside this operational limit. This was a potential source of an inadvertent over-G of the airframe.

At high speed, residual pitch oscillations involved the airplane’s short period, the first wing bending mode and certain weight and CG configurations. This was caused by a servo-aero-elastic phenomenon. Transonic shocks formed on the upper and lower wing surfaces. These caused an aft shift in the aerodynamic center, which increased the static stability and increased the rigid-body short-period frequency. These dynamics coupled with the deadband of the control surface actuators and caused constant amplitude oscillations.

This response was not predicted by analytical methods, nor observed in low-speed or high-speed wind-tunnel tests. Linear aerodynamic models were unable to capture this behavior. The high-speed wind-tunnel test did not detect this as the model was rigidly-mounted. Initial flutter flight tests also didn’t capture this mode as it requires a heavy payload with a forward CG location. With critically heavy weight configurations zero damping was observed.

• [1] PeninsulaSrsVideos, B-2 Flight Test, (Jul. 07, 2017). Accessed: Jul. 08, 2024. [Online Video]. Available: https://www.youtube.com/watch?v=l-Fo44a5oO
• [2] R. T. Britt, J. A. Volk, D. R. Dreim, and K. A. Applewhite, “Aeroservoelastic Characteristics of the B-2 Bomber and Implications for Future Large Aircraft”.

Conclusion

Hopefully, this was informative and not too dense. If you would like to read more, there are many sources, some of which are linked under their respective aircraft examples. Any other sources that I used are listed below. In future articles, we will explore more about designing stable and robust flight control systems and provide further examples of good and bad designs.

Sources

• [1] P. Dr. Hamel, “Advances in Aerodynamic Modeling for Flight Simulation and Control Design,” Jan. 2001.
• [2] E. Field, “The application of a C* flight control law to large civil transport aircraft,” 1993, Accessed: May 15, 2023. [Online]. Available: https://dspace.lib.cranfield.ac.uk/handle/1826/186
• [3] AIAA SciTech 2023
• [4] B. J. Goszkowicz, “Sidestick Controllers During High Gain Tasks”.
• [5] RTO-TR-029
• [6] “Wing Rock Prediction Method for a High Performance Fighter Aircraft”.
• [7] INCAS − National Institute for Aerospace Research “Elie Carafoli” B-dul Iuliu Maniu 220, Bucharest 061126, Romania iursu@incas.ro, U. Ioan, and T. Adrian, “Dealing with actuator rate limits. Towards IQC-based analysis of aircraft-pilot system stability,” in INCAS BULLETIN, Jun. 2013, pp. 53–72. doi: 10.13111/2066-8201.2013.5.2.7.
• [8] R. Bailey, B. Powers, and M. Shafer, “Interaction of feel system and flight control system dynamics on lateral flying qualities,” in 15th Atmospheric Flight Mechanics Conference, Minneapolis,MN,U.S.A.: American Institute of Aeronautics and Astronautics, Aug. 1988. doi: 10.2514/6.1988-4327.
• [9] D. M. R. Anderson and A. B. Page, “UNIFIED PILOT-INDUCED OSCILLATION THEORY VOLUME III: PIO ANALYSIS USING MULTIVARIABLE METHODS”.
• [10] NATO, Ed., Flight Vehicle Integration Panel Workshop on Pilot Induced Oscillations: = (Atelier sur le Pompage Piloté). in AGARD advisory report, no. 335. Neuilly-sur-Seine: AGARD, 1995.
• [11] J. Harris and G. Black, “F-22 control law development and flying qualities,” in 21st Atmospheric Flight Mechanics Conference, San Diego,CA,U.S.A.: American Institute of Aeronautics and Astronautics, Jul. 1996. doi: 10.2514/6.1996-3379.
• [12] W. Smith, “Induced and Controller Coupled Oscillations Experienced on the F-16XL Aircraft During Rolling Maneuvers”.
• [13] Andreas, “Aircraft-Pilot Coupling – Understanding why and how it happens.,” Engineering Pilot. Accessed: Jan. 14, 2023. [Online]. Available: https://www.engineeringpilot.com/post/aircraft-pilot-coupling-understanding-why-and-how-it-happens
• [14] P. Steinmetz, “Pilot-Induced Oscillation Research: Status at the End of the Century”.