How are Autopilots implemented in Flight Simulators

[Guest]

There is one observation that bothers me in some Flight Simulators:The autopilot does not manage to keep the localizer centered during ILS approaches under cross-wind conditions, I see this in X-Plane, MS-FS 2002 and FLYII (not in the final FLY2K, if I remember correctly).In real life, I only owe a PPL, and have never flown an ILS approach under cross -wind conditions, so I can't tell. But from all I know, the autopilot should apply rudder (-trim) to keep the needle centered, independent of any wind conditions. How can it be that all (or some) Flight Simulators use a wrong implementation?Or am I wrong?Thanks for any hint on that matter.Harald

[Guest Fly II Pilot]

Harald,I didn't know that X-Plane and MSFS2002 had the same problem with the wind, I thought that was just a Fly!II problem.I'm glad you asked this question because what I found out is that it seems to be just a PMDG problem. I just shot two ILS approaches to the same place (SBGR) with the same wind conditions (33kts x-wind), one with PMDG's 757 and the other one with the default Hawker 800 (Paregrine). With the 757 I landed 100 yards to the left of the runway and all the time I had a significant x-track error while the Hawker nailed it all the way through. Give it a try for yourself.Safe landings!!!Alejandro AmigorenaCheshire, CTFly! II Beta Team MemberAthlon XP 1800ABIT KR7A-RAID768Mb RAMMSI GeForce4 Ti4600 8xAGP 128MBSB Audigy GamerCH Flight Yoke USBCH Pro Pedals USBhttp://home.earthlink.net/~finsdad/FlyII/p..._aamigorena.jpg

[michal 44]

>But from all I know, the autopilot should apply rudder (-trim)>to keep the needle centered, The AP (if works correctly) will try to counter the wind with ailerons rather than the rudder. In other words the AP will do a classical crab into the wind, not the side-slip.Michael J.

[Roger Mazengarb 0]

G'day Michael,>>But from all I know, the autopilot should apply rudder>(-trim)>>to keep the needle centered, >>The AP (if works correctly) will try to counter the wind with>ailerons rather than the rudder. In other words the AP will do>a classical crab into the wind, not the side-slip.Whilst it's true that the aileron channel of the auto pilot is used to turn the aircraft under cruise condition this is not the case under all conditions. When either Loc, GS auto or GS man mode is selected SOME aileron channel signal is also fed to the rudder via the yaw damper coupler. Any such signal adds to or subtracts from the gyro signal in the coupler and the rudder will move accordingly. This use of rudder by the auto pilot provides tighter coupling on the ILS approach. ( this is how they did it on real airplanes B707 :-) ) Log Defect" aircraft auto land places the aircraft 10 metres to left of runway centreline"Rectification" auto land NOT fitted to this aircraft."Cheers,Roger @YSSY

[Guest]

Thanks for your answers.I have thought a bit more about my question, and still are not quite certain about my ideas:Assume, there is cross-wind from the left and the needle is initially centered and the aircraft aligned along the localizer. Ailerons and rudder centered.The aircraft will start drifting to the right and the needle will start deviating to the left.If needle deviation is the only input to the autopilot, then the autopilot may counteract this by applying rudder, aileron, or a combination of both.If it applies rudder alone, the aircraft will turn left and should eventually intercept the localizer again, resulting in a zero needle deflection and the autopilot,-assuming it is based on a proportional controller only-, will center the rudder again. This would result in the aircraft performing a pure crab, but remaining on the extended runway centerline (with some oscillations, which should be damped by the autopilot controller).Basically, the same should happen if the autopilot applies only aileron, because of the bank/roll moment a banking to the left will result in a change of heading to the left.Also, for obvious reasons, the same should happen, when the autopilot applies a coordinated combination of aileron and rudder.The only case, which might be consistent with my observation, i.e. ending up on a track parallel to the extended centerline, would be when the autopilot enters a slip by applying a combination of left aileron and right rudder with an amount proportional to needle deflection. In that case there would be a stable condition with the needle to the left, a slip to the left and a track parallel, but offset to the centerline.But this undesired behaviour might be overcome if the autopilot controller would consist not only of a proportional element, but would also include an integrator to counteract the steady needle deflection.Apparently, in the real world, there are autopilots with different degree of sophistication, ranging up to autoland, including autoflare and automatic decrab just before touchdown.I am just wondering, how much of this is included in flight simulators.Harald

[michal 44]

>the case under all conditions. When either Loc, GS auto or GS>man mode is selected I wonder how many aircraft in the world do it this way. Certainly none of the general aviation ones. If we are talking about heavy transport category the only one I know something about is 767. 767 will crab first and ONLY in *extreme* wind conditions will use rudder.So I probably wouldn't be too far off the mark saying that in 97% of crossiwnd approaches with AP - rudder plays no part during the automated part of the flight.Michael J.

[Roger Mazengarb 0]

G'day Michael, On the 747 when the AP is in Land mode, with at least 2 switches in command, the rollout logic engages at 5 ft radio altitude and at touchdown the rudder is commanded to maintain the runway centreline. The localiser signal is the primary guidance command to maintain a straight rollout until the AP is disengaged. This rollout logic was not incorporated in the 747 SP. The 747 has a turn co-ordination feature whereby bank angle info is fed from the INS system (the primary source of pitch and roll) to the yaw damper coupler to give rudder to counteract adverse yaw. When the flaps are selected down the signal is amplified to give the increased rudder necessary due to the low speed.The operation of the AP in modern aircraft is very complex and Auto Land functions complicate the issue.I think it a fair bet to say that 97% of GA aircraft don't have AP with auto land feature.Cheers,Roger @YSSY

[michal 44]

Roger,Points well taken however I personally don't think it is fair to the original poster to bring autoland into the mix since it has little to do with the spirit/intent of the original question and it only muddles the issue. Michael J.

[Guest Dean]

>The autopilot does not manage to keep the localizer centered>during ILS approaches under cross-wind conditions, I see this>in X-Plane, MS-FS 2002 and FLYII (not in the final FLY2K, if I>remember correctly).I could be mistaken but I seem to remember we tested for this fix in one of the last patches for Fly II.

[Guest]

Hello all Im speaking from a GA backround so correct me if Im wrong.Autopilots do have diffrent levels of sophophistication.Some are mearly wing levelers while others will fly a GPS approach to the numbers, including altitudes over specific fixes. Not to positive about the Boings but....Most GA aircraft have only aileron and elevator manipulation.No rudder input with autopilot. The CDI on the Localizer/Glidslope represents a straight line, it really makes no diffrence which direction your heading, it just shows your represention to the straight line.If you approach on the center line and say have a xwind from the left the aircraft will point left while drifting to the right, ie wind correction, only this is not wind correction its the aircraft "windvane"ing into the wind like that metal rooster on the barn.With the proper wind correction the plane should fly the line(approach) nice and straight, the only issue is, you have to kick the rudders to straighten the plane out before touchdown otherwise you have massive sideloading on the gear. The other approach is to fight the wind vane effect with right rudder while banking the plane left to counteract the drift. This is a hard manouver to fly when you are in a heavy xwind as you are really fighting and making constant small adjustments, however this is the technique the FAA prefers.To the point: I would not want my autopilot to try the second method,(slipping) as this indicates that the autopilot would be programmed to cross-control the aircraft. Do the big boeings do this? Im not sure, it would seem wierd though.Most likely the autopilot is using the elevators and ailerons to get the plane on the line regardless of heading, it will use the correction needed to keep the plane on the line even if the approach is HDG 000 and the plane is flying HDG 350. As long as the plane is on the line it will try to keep it there with heading adjustments not rudder adjustments. Id like know if Im wrong though, as they saying goes a good pilot is always learning.

[Guest]

>>The autopilot does not manage to keep the localizer>centered>>during ILS approaches under cross-wind conditions, I see>this>>in X-Plane, MS-FS 2002 and FLYII (not in the final FLY2K, if>I>>remember correctly).>>I could be mistaken but I seem to remember we tested for this>fix in one of the last patches for Fly II. >I have the latest (December 2002) version of FLY II, and the problem exists in this version. Also, I checked in FLY2k again, and it is there as well (I used the King Air).Harald

[Guest]

I do not object to anything that has been written in this thread.However, it does not help me with my question.Basically, for the ability to keep the needle (CDI) centered on approach, it is immaterial whether the autopilot controls aileron or rudder. The only case which would be consistent with my observations (in XPLANE as well) is that the autopilot enters a slip with a purely proportional (to CDI deflection) control law.This again, is in contradiction to the plausible assumption that real-world GA-Aircraft (low-end, including King Air) use autopilots, which affect aileron/elevator only.Of course, in a simulation, you can do many things, and my question was just: How do you think (or know), it is implemented in these flight simulators? Harald

[michal 44]

>However, it does not help me with my question.and what question was that ??? The AP are simulated in software - through some sort of feedback-system. Are you asking for lines of code, psuedo-code, design logic ? Check with people who did it. You bet that some implementation are "wrong". For example FLY! never had corectly working Flight Director and its operation is closely tied to autopilot. A non-working FD on a sophisticated biz jet like a Hawker is really a major drawback.So you have to address specific teams, individuals, grab them by their neck, strap them to chairs in a basement, apply torture and question how come they failed to design something correctly or never managed to fix the problem. :-lol>it is immaterial whether the autopilot controls>aileron or rudder. I am afraid it is. As a pilot you should know that it is diffcult to change airplane's heading by more than say 10 deg with rudder alone. Aileron (roll axis) does it much better therefore AP are hooked to ailerons.I don't have XPlane, stopped using FLY sometime ago but I know that something like 767PIC (MSFS add-on) or Flight1's Meridian (another MSFS add-on) or even Bill Grabowski's freeware ERJ-145 have very good autpilots that are close enough to how things work in real life. If I come across an aircraft/panel which otherwise is great but has poor autopilot or flight director - it is out of my HD in 1 minute. My "hangar" has very few aircraft but they meet my quality criteria.Michael J.

[Guest]

Sorry, it was not my intention to offend anybody by my post or to blame him for not giving the proper answer to an unclear question.I don't expect pseudocode or other implementation details.Basically, I would like to know which inputs are provided to the autopilot simulation, especially if it is based on needle deviation only, and what sort of control law is applied (proportional, PI, PID, or other). Also, which control surfaces (rudder, aileron, elevator,...) are affected by the autopilot would be of interest.I am just wondering, which sort of implementation is consistent with my observations in FLY and X-Plane.When I wrote "it is immaterial whether rudder or aileron is used", I was well aware that this would have an impact on aircraft behavior, but in both cases, the trajectory would be a pure crab "exactly" on the localizer path (no offset).(By the way, in manually flown ILS approaches, the fine corrections for heading are mostly done by rudder input only).Any volontary comment is welcome. (I do not want to put pressure or apply torture)Harald