Airbuses side stick improves crew comfort and control, but is there a hidden cost?
The Airbus FBW side stick flight control has vastly improved the comfort of aircrew flying the Airbus fleet, much as the original Airbus designers predicted (Corps 1988). But the implementation also expresses the Airbus approach to flight control laws and that companies implicit assumption about the way in which humans interact with automation and each other. Here the record is more problematic.
The ‘FBW side stick’ trope
Before we start lets discuss and dispose of one particular trope, that the Airbus side stick controller interface was an inevitable (and integral) part of Airbuses FBW implementation. In practice only the use of a passive (no feedback) controller is truly a result of Airbuses initial decision to implement flight control ‘protection laws’ within the flight software. With protection laws performed by the computer no feedback of force/displacement to the pilot was required to prevent him over stressing the aircraft (for example) and a passive stick could be implemented (1). One could also just as easily implement a passive interface via a traditional yoke controller as a side stick. So we should remember when discussing side stick controllers that the use of a side stick is a design decision made, in the words of Airbus itself, to improve the human factors of the cockpit via increased space, comfort (2) and display visibility (3 ) not as an integral part of the FBW design.
Problems with crew coordination
Of course passenger aircraft are not flown by single pilots so the design also needed to address how pilots would transfer control between them. And as a consequence the necessity to coordinate such control. An initial ‘selected priority’ concept using a single central changeover switch was investigated by Airbus but rejected over concerns about the resultant single point of failure as well as potential crew conflict over it’s use and whether the switch would be easily located and used in an emergency.
Airbus then opted for pure electronic coupling with a control displacement of above a certain threshold (75%) initiating a transfer of control. However after one minor flight test incident (Corps 1988) the designers realised that in an emergency situation control could just as quickly be inadvertently taken back by the other pilot, and that a controller located priority override button with a visual indicator of who has control was a better implementation. This priority override was eventually to become the A320 launch implementation.
However, in operation, incidents subsequently occurred in which the Pilot Not Flying (PNF) initiated control inputs without transferring control. Investigation by Airbus categorised these inputs as spurious (unintentional), comfort (minor and intentional corrections) or instinctive (large inputs usually in response to a perceived threat). Of these, naturally, the instinctive inputs are the greatest risk.
“[In response to a TCAS advisory] …The Captain initiated an immediate descent. The Captain did not make a verbal announcement that he was taking command of the left side stick control.”
US NTSB Factual Report FTW96LA269 1996.
So while a priority override works under ‘normal’ circumstances, it appears that an error trap had been built into the system that emerges in situations of high stress, workload and time compression. We should note that a similar general coordination problem has also arisen in the past with aircrew forgetting to disconnect autopilots.
To address this potentially hazardous situation the side-stick priority display was augmented by Airbus to trigger an aural ‘Dual Input’ warning and illuminate the side stick priority warning light if control inputs of greater than 2 degrees were made by the PNF. But although this provides greater annunciation of ‘whose got the ball’, it turns out that in the circumstances identified as triggering instinctive responses the value of such alerts is degraded due to the inevitable attentional tunnelling that operators experience in high stress situations. So in the real world it appears that this implementation is less than effective (ATSB 2007).
The investigation was unable to determine whether the copilot was aware of the pilot in command’s dual sidestick inputs, even though they resulted in aural ‘DUAL INPUT’ synthetic voice messages…The copilot’s focus on correcting the aircraft’s attitude and trajectory, together with the numerous FWS synthetic voice messages, may have resulted in the copilot not comprehending the significance of the aural ‘DUAL INPUT’ warnings…”
ATSB Occurrence Report, 2007
As the original designers noted ‘having two sets of hands on the controls is a no no..’ (Corps 1988) but in an emergency with a traditional yoke control the instinctual grab by the PNF is something that is clearly visible and also inherently self resolves the control contention, usually by one pilot yelling at the other. A series of aural alerts and visual priority symbology on a display, while technically achieving the same effect, is a more complex coordination mechanism and, it appears, more likely to break down during high stress events. We continue to see this behaviour in flight crew to this day, see for example the unannounced transfer of control between pilots during the AF 447 accident, so this appears to be a recurrent procedural violation (4).
But what do pilots prefer?
In a 1987 evaluation of side stick controllers Summers et al (1987) found that under simulated ‘surprise’ hand overs pilots Cooper Harper rating of the schemes were (in descending order):
- Coupled sides sticks with algebraically summed inputs (1.4),
- Uncoupled side sticks with algebraically summed inputs and disconnect switch (final A320 implementation) (1.8),
- Uncoupled with algebraically summed inputs and priority logic (original A320 implementation) (3.3), and
- Uncoupled side sticks with with algebraically summed inputs (3.4).
Summers noted that pilots preferred a coupled system because it gave force feedback and ‘a sense of urgency’ (Summers 1987). From the results above we can see that, while acceptable, the disconnect switch is not the ‘best’ solution from a pilots perspective. As Summers study was conducted using a simulator one should also note that the full effects of performance shaping factors such as workload, fatigue and stress could not be fully evaluated. In a real world flight environment these ratings may well change. An important caveat to keep in mind when statements are made as to the ‘pilot friendliness’ of a particular interface.
Another study of the effect of physical location on the acquisition of flying skills also found that trainees had considerably more difficulty in learning simulated flying skills with the side stick positioned controller than with a centrally located one (Bo-Kuen 2002).
Finally we should note that the original Airbus evaluation compared a side-stick FBW implementation with a yoke and traditional flight controls system (Airbus 1998). So it seems that Airbus initial design evaluation was constrained to an pre-selected FBW candidate solution.
A key cognitive engineering flaw
The decision to use passive controllers unfortunately introduced a cognitive engineering design error in the form of the absence of tactile feedback to the pilots. This in turn transferred the behavioural cue from the tactile (hands on stick) to the visual channel of the pilots, with the design implementation further translating the cue from an analog kinaesthetic to visual symbolic form, which in turn requires higher levels of cognitive processing. This visual cue then competes for attentional resources in the visual channel and requires additional ‘mental bandwidth’ (e.g. symbology requires abstract processing) to handle. The net result an ineffective coordination mechanism under high stress conditions.
The problem is not where the controller is located per se but how effective is the total human machine interface in coordinating crew behaviour. Under circumstances of high crew workload and stress it appears that Airbus has built an error trap into their aircraft’s human machine interface. If you then asked me why this occurred I’d answer that it was probably because this very well intentioned change was driven by engineers, without the involvement of psychologists or human factors specialists.
This post is part of the Airbus aircraft family and system safety thread.
Airbus, A319/A320/A321 Flightdeck and systems briefing for pilots, Airbus Industries, STL 945.7136/97, 1998.
ATSB, Aviation Occurence Report 200505311, Crosswind Landing Event Melbourne Airport, Vic. – 26 October 2005 HS-TNA, Airbus A340-642, 2007.
Bo-Kuen, Cho., Impact of subject related factors and position of flight control stick on acquisition of simulated flying skills using a flight simulator, Doctoral dissertation, Louisiana State University and Agricultural and Mechanical College, 2002.
Corps, S.G., Airbus A320 side stick and fly by wire an update, Society of Automotive Engineers (SAE) Paper 861801, 1988.
NTSB, US NTSB Factual Report FTW96LA269 1996.
Summers, L.G., Shannon, J.H., White, T.R., Shiner, R.J., Fly By Wire Sidestick Controller Evaluation, Aersopace Technology Conference and Exposition, Long Beach, California, 1987.
1. This does assume that the only reason a pilot would need combined force/displacement feedback from the controls is to prevent inputing an unsafe command.
2. Thereby allowing pilots to cross their legs and use a table tray, the positive effect of increased comfort upon crew efficiency and fatigue management should not be underestimated.
3. Note that all the usually claimed benefits of increased precision and reliability through elimination of mechanical gearing, clutches and so on also accrue from the decision to use a passive controller, which lends itself to an electronic implementation. Again side or central location is irrelevant.
4. Violation is used in the sense of James Reasons taxonomy of human error. Violations are the deliberate disregard for rules and procedures. A routine violation is one that occurs regularly (such as ‘comfort’ dual inputs) and is generally tolerated while an exceptional one is one that is both unexpected for that operator and not (or should not be) tolerated by management (such as ‘instinctive’ dual inputs).