F1 2011 Official AI Thread

[Carl Williamson]

Playing on Legend @ monza and the AI just won't overtake you no matter what, they will get a good run on you going into certainturns but will instead decide to just stay behind you for the entire race. It's bizarre. Had similar issues on other tracks, they just won't overtake you unless you make some major error. It kinda ruins the offline portion of this game.

 

[TGApples]

Guys, these complex data calculations may well apply to the player cars but there is no evidence of Codemasters trying to replicate on lap activity of the AI in this way. I suspect if you take a look at the laptimes files for each track (eg in Australia open up Codemasters\F1 2011\tracks\circuits\melbourne\route_0\laptimes.xml) it becomes obvious the AI are actually on a virtual rail system with 20 pre-determined points on the track , it appears the cars have been synched against the Ferrari car (Fer) and each sector oint has a time in secs for perfect grip level, level 2 grip etc. I think there is nothing wrong with this model it is real simple and probably used for the entire AI core lap time structure. I think you are both giving CM too much credit for detailing how the AI may apply throttle etc. Mistakes look like they are induced against a multiplier set against a standard index. Again, nothing wrong with this but as they use a more complex system for calculating for mapping AI car track location and lap time performance when the car is running close to the player cars I suspect this is causing the discrepancies between real and sim lap times.

Standard racing game AI works thusly:

1) A line is drawn. This is the racing line.
2) At nodes along this line a maximum/suggested speed is set.
3) The AI will do it's best to stay as close to this line as possible and as close to said speeds as possible while avoiding other objects. Sometimes it'll use the same physics as the player, sometimes it won't. If it doesn't, it'll be made to look like it is. Sometimes not with amazing results.

Sounds like how Codies have done it .

 

[daytona64]

TG, yes it appears that simple.

You mentioned earlier you would look at modding the speeds on the tracks based on those groups. Would you kindly do so based on the increase/decreases of +3 +2 +1 0 -1 for the 5 groups you listed under.

Also, do you have any idea where the simulated (accelerated time etc) Qualifying times for the AI can be sped up by 1-2 seconds per lap. I noticed 2 secs at Australia and 1 sec in Valencia (rest seem to be 1.5 seconds out.

If you fancy having a go at that you might offer the first decent mod which covers lap time issues.

Combine that with Cowboy2's aggression settings and increase the loose_parts_settings to 1.9 across all 4 settings - I reckon you will be hailed a hero across the F12011 world.

If you want me to work with you on it PM me I would be pleased to help.

 

[driver1973]

>If something is calculated using constant values it itself must surely be constant, right? Can't see any way out of that.

Only at the certain point of the track. At other points the calculation will be different as the input values will be different.
However, this doesn't mean the input values are not constant. You just has to look then up in an array or matrix or whatever.

>Now if the cars were nice and simple. No aero, no fancy springs, constrained to a 2D plane of the track with a grip that >doesn't vary at all around the corner we'd perhaps have a chance at finding the best line, especially for simple single >corners. As it is, with the grip at every simulation tick dependant on all the previous ticks? No chance.

Why are you insisting that the best line cannot be pre-calculated? Even F1 2011 does this with that driving aid.

>Incorrect. Cold tyres, worn tyres, options/primes, amount of water on the track, track temperature, fuel quantity, damage, >horsepower, brake wear, whether KERS/DRS are available, fuel mix... these things can only be known as the car is >actually going around the track. Even if these we all known see above for why it can't simply be calculated.

As I said this just modifies the car's grip value. Cold tyres, worn tyres: subtract some value from the grip, options/primes: different initial grip and tyre wear, etc. Let's see an example:

Assume that an AI car's default grip value is 200 in Monza. Let's say we have 50% fuel and worn tyres. Let's say this will reduce the grip value to 100. Now, let's see other parameters like brake wear. Let's say the maximum deceleration is 4.5g (predefined value). Break wear, tyre wear and fuel load decrease this to eg. 4g (wings does not count, they are factored in the default grip value).

Let's say the AI car is travelling at the start/finish line at 300kph with calculated grip value 100. The AI looks ahead and see the first right hand turn. It sees that at the slowest point of the turn the track grip value is -500. Also, the wheels have to turn also and that's further reducing grip, let's say -30 (just look it up in a table). So, the calculation will show 100 + (-500) + (-30) = -430 for the grip value. Grip 0 means the car is on the limit, so grip -430 means it has to break to manage the turn.

At this point the AI looks into a table and finds out at which speed the car grip value will be 530. Let's say it finds that at 90kph. So, it has to be slow down from 300kph to 90kph. That's 210kph difference. With 4g deceleration is can easily calculate the break distance. If the track is wet the grip of the track at this point will not be -500, but eg. -1000. And the deceleration will be eg. 2g not 4g. And that's all. With just changing some values you simulated wet track.

Now, tell me how this cannot be done at real time.

 

[TGApples]

>If something is calculated using constant values it itself must surely be constant, right? Can't see any way out of that.

Only at the certain point of the track. At other points the calculation will be different as the input values will be different.
However, this doesn't mean the input values are not constant. You just has to look then up in an array or matrix or whatever.

>Now if the cars were nice and simple. No aero, no fancy springs, constrained to a 2D plane of the track with a grip that >doesn't vary at all around the corner we'd perhaps have a chance at finding the best line, especially for simple single >corners. As it is, with the grip at every simulation tick dependant on all the previous ticks? No chance.

Why are you insisting that the best line cannot be pre-calculated? Even F1 2011 does this with that driving aid.

No it doesn't. It gives you a good line, sure, but if you follow that line exactly around the track it's not the fastest way round... so it's not the best line.

Let's say the AI car is travelling at the start/finish line at 300kph with calculated grip value 100. The AI looks ahead and see the first right hand turn. It sees that at the slowest point of the turn the track grip value is -500.

You can't possibly know what the properties of the slowest point in the best line are without first knowing the best line.

My main issue here is your insistance on "best" and "perfect" lines. I have no issue with the fact you can find "good" lines, but if that's what you mean that's what you should say.

 

[driver1973]

>You can't possibly know what the properties of the slowest point in the best line are without first knowing the best line.

>My main issue here is your insistance on "best" and "perfect" lines. I have no issue with the fact you can find "good" lines, but if >that's what you mean that's what you should say.

And why the AI cannot be aware several best lines? You can precalculate several possible lines over the track. I do not see why you should do that but whatever. It seems to me you are implying that there are infinite best lines over a track which is clearly not true. IMHO, one best line per track is enough with a reasonably good AI. You just needlessly overcomplicate things.

Keep it simple and it will work.

 

[TGApples]

>You can't possibly know what the properties of the slowest point in the best line are without first knowing the best line.

>My main issue here is your insistance on "best" and "perfect" lines. I have no issue with the fact you can find "good" lines, but if >that's what you mean that's what you should say.

And why the AI cannot be aware several best lines? You can precalculate several possible lines over the track. I do not see why you should do that but whatever. It seems to me you are implying that there are infinite best lines over a track which is clearly not true. IMHO, one best line per track is enough with a reasonably good AI. You just needlessly overcomplicate things.

Keep it simple and it will work.

...

You cannot find the best line or perfect line without hideous amounts computational work. These two words both imply that there is no possible way one can make the taking of the corner better. There cannot be several best lines (or rather it's massively improbable). The best line is the best line. It is precise, definable, and bloody tricky to find. Nearly best is not the same as best. If you drop a millionth of a second over the best line you are no longer on the best line.

You can find good lines, however I dispute that finding a good line in real time is going to be easy. Far easier to take a line that a competant human has taken through the curve then modify it to suit current car performance. This is what is currently done AFAIK.

I can't work out if you're trolling or just failing to understand the english language.

 

[driver1973]

>I can't work out if you're trolling or just failing to understand the english language.

Maybe it is you who do not understand the language. I always wrote about precalculated lines. Also, I think it is trivial for a mathematician to write an algorithm which calculates the best line for a track and you just have to store the data. But it looks like this simple solution eludes you. And if you think something cannot be done because you can't do it, you immediately states it is impossible to do.

You did not give any technical reason why a simplified simulation which I described can't work. You just want to desperately find excuses. Aren't you a CM employee who want to defend CM's lame AI at all cost?

 

[Jace11]

Looked at some mods that are said to improve AI behaviour (not database aggresion mods). Testing on Abu Dhabi circuit by swapping some in and out. One turn in particular causes default vanilla AI to spin out quite often, it is the 2nd part of chicane after long back straight. Strarted looking at how to correct this and at mods that are said to improve AI performance and make overtakes more often etc.

Some mods say they do this.. but be careful, I think some are just changing values randomly... I noticed one mod that claimed to make AI more aggressive but it contained (I think) lots of safety car AI changes. LOL.

I think the prefixes give big clues as to what the values mean and then if you look at the prefixes the strings make some sense... and you can see what certain sections are for...

Trying to understand AI files: driver config, in REDis what I think it means...

<hvOvTimeThresholdAtMinSpeedDelta float="0.4" /> hv = heavy / heat maybe? Ov = overheat here not Overtake
<hvOvMaxInterestingOvertakeTime float="4.0" />
<hvOvMinInterestingOvertakeTime float="0.5" />
<hvOvMinInterestingHeat float="0.55" />
<hvOvPeakHeatReductionProportion float="0.95" />
<hvOvSkirtHeatReductionProportion float="0.9" />
<rmCurvatureMultiplier float="1.6" /> rm = random or race main maybe?
<rmLowSpeedMinGrip float="0.0" />
<rmHighSpeedMinGrip float="0.0" />
<rmHighSpeedPeakSlipAngle float="1.0" />
<rmHighSpeedPeakSlipAngle float="1.0" />
<rmMaxSteeringDelta float="500.0" />
<rmMinLookAheadDistance float="4.0" />
<rmMaxLookAheadDistance float="25.0" />
<rmLookaheadMinSpeed float="10.0" />
<rmLookaheadMaxSpeed float="50.0" />
<rmLookaheadMinAngle float="0.0" />
<rmLookaheadMaxAngle float="0.02" />
<scTargetSeparationTime float="0.5" /> sc= safety car
<scInLapLeadSeparationTime float="15.0" />
<scBrakingPerformance float="0.8" />
<scCorneringPerformance float="0.85" />
<scMaxThrottle float="0.48" />
<scMaxSpeedThroughIncidentZone float="50.0" />
<scLengthOfOIncidentZone float="100.0" />
<stSlideFactor float="0.0" /> st = unknown
<stProportionRabbitWhenInAir float="0.0" />
<rearWingMaxCurvature float="0.01" /> DRS use by AI
<rearWingMinSpeed float="25.0" />
<rearWingMinDistanceToNextCorner float="200.0" />
<kersRandomVarianceRange float="0.05" /> KERS
<thBrakeLineSpeedDrop float="0.25" /> throttle?
<brFollowSpeedBrake float="0.5" /> br=brake
<tgOffsetMergeFactor float="2.5" />
<tgOffsetSpeedFactor float="0.9" />
<tgOffsetMaxForSpeed float="4.0" />
<tgNonOptimalSpeedFactor float="0.6" /> tg=? time gap? not sure? maybe tailgate ie close pursuit behaviour looking at the values?
<tgClampOffsetLeft bool="true" />
<tgClampOffsetRight bool="true" />
<tgUndersteerMaxOffset float="2.0" />
<asRejectDistance float="15.0" /> as=? no idea but maybe also an "alongside" entry..
<asOffsetMergeFactorGentle float="2.5" />
<asOffsetMergeFactorExtreme float="5.0" />
<asTargetBuffer float="-2.0" />
<asWideLimitsSpeed float="999.0" />
<asImminentCollisionTime float="0.4" />
<asImminentCollisionDistance float="0.25" />
<asCarInFrontYeildDistance float="9.0" />
<rtRetirementSpeed float="60.0" /> rt=retirement behaviour
<rtRetirementResetTimer float="12.0" />
<rtStuckTime float="10.0" />
<rtStuckDistanceSq float="4.0" />
<rtPlayerDistanceSq float="10000.0" />
<rtStoppedSpeed float="0.5" />
<rtModerateDamageThreshold float="0.5" />
<damagedMaxThrottle float="0.6" />
<damagedCurvatureModifier float="0.6" />
<dfMinDraftSeperation float="1.25" /> df = slipsteaming effects? not defending as I first thought
<dfBrakeZoneLimit float="300.0" />
<dfTrackCurvatureLimit float="0.005" />
<dfBrakeSpeedUpperLimit float="84.9376" />
<dfMaxDraftSpeedDifference float="10.0" />
<drMinDraftSpeed float="35.0" />
<rcResetAllowed bool="false" /> rc=recovery (after spin or offline etc)
<rcRecoveryTimeOut float="10.0" />
<rcMergeFactor float="1.0" />
<rcCollisionTimeOut float="0.2" />
<rcTrackViewRecheckEnabled bool="false" />
<rcResetTrackViewTimeOut float="1.0" />
<rcWaitForAIWhenRecovering bool="true" />
<rcRecoverToTargetNearDistance float="10.0" />
<rcRecoverToTargetFarDistance float="45.0" />
<rcRecoverySpeedLimit float="10.0" />
<rcMinTimeBetweenRecTypeChanges float="0.0" />
<rcRayCastingEnabled bool="true" />
<rcVehicleAvoidEnabled bool="true" />
<rcSceneryAvoidanceCoefficient float="6.0" />
<rcSceneryForceReductionCoefficient float="1.0" />
<rcSceneryForceMinimumThreshold float="0.1" />
<rcForceToTargetCoefficient float="3.0" />
<rcForceForVelocityCoefficient float="0.5" />
<rcCollisionDetectionRange float="1.0" />
<rcSceneryCollisionDetectionRange float="15.0" />
<rcVehicleCollisionDetectionRange float="10.0" />
<rcVehicleAfterTimeMax float="1.0" />
<rcVehicleAfterTimeFactor float="1.0" />
<rcAvoidVehicleFactorCoefficient float="8.0" />
<rcAvoidVehicleExponentialCoefficient float="3.0" />
<rcThresholdSteeringForceToMove float="0.05" />
<rcThresholdDotDirectionBlocked float="0.5" />
<rcThresholdForceLength float="5.0" />
<rcReactToPlayerOnly bool="false" />
<raGenerateRandomOffsets bool="true" /> ra= race alone? race? race attacking?
<raOffsetMergeFactor float="1.0" />
<raOffsetMergeFactorCornering float="0.5" />
<raOffsetMergeFactorAlone float="0.2" />
<raAloneDistanceOn float="70.0" />
<raAloneDistanceOff float="60.0" />
<raOffsetGenerateTime float="5.0" />
<ovOffsetMergeFactor float="4.0" /> ov=OVERTAKING BEHAVIOUR!!!! important to test these I think...
<ovTime float="5.0" />
<ovBrakeOvertakeTime float="0.0" />
<ovRejectDistance float="400.0" />
<ovSpeedDifference float="0.1" />
<ovFollowDistance float="3.0" />
<ovDecisionTimeOut float="0.0" />
<ovMinFireTime float="0.25f" />
<ovMinCornerCurvature float="0.001" />
<ovMaxCornerCurvature float="0.04" />
<ovOffsetMergeFactorCorner float="0.5" />
<ovSlowSpeed float="10.0f" />
<ovWideLimitsSpeed float="7.5" />
<ovMinMetresSeparationAwayFromCorner float="10.0" />
<ovSeparationDistanceAtBrakeLine float="10.0" />
<ovSecondsBeforeBrakeZoneToReact float="2.5" />
<vlDropoffStartTime float="0.5" />
<vlDropoffEndTime float="1.0" />
<vlViewAngle float="1.57" />
<blOffsetMergeFactor float="2.0" /> bl=BLOCKING
<blMaxBlockeeDeceleration float="5.0" />
<blMaxTrackCurvature float="0.005" />
<blMinIntersectionTime float="1.0" />
<blMaxIntersectionTime float="4.0" />
<blMaxSpeedDifference float="20.0" />
<blBrakeLineDistMin float="150.0" />
<blMinValidBlockingMovement float="3.0" />
<blCornerCurvatureToBlockToTrackInside float="0.015" />
<blMinSpeed float="25.0" />
<avOffsetMergeFactor float="5.0" /> av=avoidance behaviour... AI avoidance of stopped, spun player and AI cars?
<avMinSpeedNormal float="0.0" />
<avMinSpeedSlow float="9.0" />
<avNormalTargetSpeed float="200.0" />
<avStoppedTargetSpeed float="30.0" />
<avWrongWayTargetSpeed float="5.0" />
<avBufferZone float="2.0" />
<avBufferZoneBaseSpeed float="5.0" />
<avETAReactionTime float="3.0" />
<avMinRacingSpeed float="4.0" />
<avPercentagePedalPushDistance float="1.2" />
<avSimilarDirectionMaxAngle float="0.5" />
<avPassAngle float="0.15" />
<avPassSlowSpeed float="20.0" />
<avOncomingTraffic bool="false" />
<avFigureOfEightTrack bool="false" />
<avAvoidAlongside bool="true" />
<avSlowSpeed float="40.0" />
<avSeparationSpeedDiff float="10.0" />
<avSideBuffer float="2.0" />
<avSideBufferMinSpeed float="5.0" />
<avSideBufferMaxSpeed float="20.0" />
<lcRecoverSpeed float="10.0" />
<lcRecoverTime float="3.0" />
<rsLandRushStart bool="false" /> rs= rush start (starting line behaviour)
<rsMinimumWaitAtLights float="0.0" />
<rsMaxThrottleAtLights float="1.0" />
<rsStallPercentage float="0.0" />
<rsEndSpeed1 float="30.0" />
<rsEndSpeed2 float="50.0" />
<rsTimeout float="10.0" />
<rsStaggeredIgnoreDriverAtStartGrid bool="false" />
<rsMaxRevDelay float="1.0" /> rs=start behaviour also..
<rsOffsetMergeFactor1 float="0.5" />
<rsOffsetMergeFactor2 float="2.0" />
<crMaxTrackCurvature float="0.01" />
<coYieldMinSpeed float="16.0" /> co = cooperation, team orders? friendly yields?
<coYieldDistToLimit float="1.0" />
<coLetPassThrottleRatio float="0.8" />
<coSpinSpeed float="1.6" />
<coMinSpaceForForkDecision float="5.0" />
<coFriendshipThreshold float="0.5" />
<sqMinSafeSeperation float="0.4" /> sq=squeezing by AI or to AI?
<sqTargetSeperation float="0.55" />
<sqMaxSafeClosingSpeed float="0.75" />
<sqMaxSafeSqueezedCarClosingSpeed float="0.2" />
<sqMaxSafeSpeedDifference float="10.0" />
<sqMaxCornerRadius float="80.0" />
<sqMaxDistanceToCorner float="300.0" />
<trTrackCurveTargetDistance float="200.0" />
<trTrackCurveOffsetRecalc float="0.25" />
<trTrackCurveMinSpeedInRatio float="0.0" />
<trTrackCurveOfflineRatio float="0.6" />
<desiredSpeed float="0.0" />
<minSpeedLinePercentage float="0.0" />
<contactZone float="0.1" />
<contactOffsetMergeFactor float="5.0" />
<resetTrackViewTimeout float="3.0" />
<maxNumObservedClosestForwardVehicles int="6" /> This section deals with AI awareness
<maxNumObservedSoonestForwardVehicles int="3" />
<maxNumObservedClosestRearVehicles int="2" />
<maxNumObservedSoonestRearVehicles int="2" />
<maxObservedViewDistance float="1000.0" />
<maxObservedViewTime float="10.0" />
<atgBrakingZoneTriggerTime float="2.5" />
<atgDefaultDistanceSensitivity float="10.0" />
<atgTimeSeperationFudge float="0.3" />
<alongsideSpaceStraightCurvature float="0.005" /> SIDE by SIDE behaviour / limits
<alongsideSpaceTightCurvature float="0.03" />
<alongsidespaceCurvatureFactor float="1.0" />
<pqSessionRhythmActive bool="true" /> pq= MUST BE PRACTICE -QUALIFYING..
<pqSessionRhythmQualifyingScalar float="1.0" />
<pqSessionRhythmFastScalar float="0.98" />
<pqSessionRhythmNormalScalar float="0.95" />
<pqSessionRhythmSlowScalar float="0.87" />
<raceSessionAdjustmentScalar float="1.0" />


Thoughts, input / help needed... hope some of you are willing to try a few edits etc..

 

[TGApples]

Maybe it is you who do not understand the language. I always wrote about precalculated lines.

Ok?

So, a real AI can always make a perfect lap

The best line is clearly constant. It can be pre-calculated

Why are you insisting that the best line cannot be pre-calculated?

one best line per track is enough with a reasonably good AI

Anyway...

Also, I think it is trivial for a mathematician to write an algorithm which calculates the best line for a track and you just have to store the data.

No. No it's really not. I have explained this before. I have an analogue throttle/brake/steering input. Maybe I'll find a nice line, then I'll change the steering intput by 0.0001% at simulation tick 26/90 and I'll go faster by a millionth of a second. A better line. Given all simulation ticks are dependant on all simulations ticks than have gone before, and hence all input variables are dependant on ticks that have gone before, knowing that you have the best line is a mammoth task.

 

[TGApples]

Good job Jace! I suspect co-operation might be for all lapped cars/slow cars, so would probably be a good one to tweak to get rid of their slightly odd behaviour right now.

I too have noticed awful AI at Abu Dhabi. I was going to make some mods after I've finished my current season.

I think the two main ones I'd want to start with are:
1) Don't spin at that corner at Abu Dhabi. Seems to spin there ~25% of all laps.
2) Adjust braking zones on big stops to be later

Hopefully both should be fairly easy. The file you posted looks like a generic AI file, but I guess there are ones for every track?

 

[Jace11]

Yes, the Abu dhabi corner will require a track AI fix. Above is a generic AI file. I'm looking into it just to see if there is any hope of improving AI overtaking behaviour.

 

[daytona64]

Jace - outstanding discovery

<lcRecoverSpeed float="10.0" />
<lcRecoverTime float="3.0" />

lc = launch control (start procedure variables)

 

[Jace11]

Jace - outstanding discovery

<lcRecoverSpeed float="10.0" />
<lcRecoverTime float="3.0" />

lc = launch control (start procedure variables)

Yes maybe, it's right before race start entries so could be...??? Although it may also be "lost control" and it uses the recover string.. hmm tricky one.

 

[Paul Luck]

Yes, the Abu dhabi corner will require a track AI fix. Above is a generic AI file. I'm looking into it just to see if there is any hope of improving AI overtaking behaviour.

These lines look like they control AI braking:
<ovMinMetresSeparationAwayFromCorner float="10.0" />
<ovSeparationDistanceAtBrakeLine float="10.0" />
<ovSecondsBeforeBrakeZoneToReact float="2.5" />

Would be worth testing these I think

 

[daytona64]

Yes, Jace is fixing Abu Dhabi corner at moment, once done he will be having a deep look into this - he is onto this one and I am testing out some changes.....
<ovRejectDistance float="500.0" />
<ovMinMetresSeparationAwayFromCorner float="10.0" />
<ovSeparationDistanceAtBrakeLine float="11.0" />

Why dont you guys have a look at this too...

change from
500-10-11
to
400-8-50
- maybe try
300-6-60
or
250-4-70

If anyone else can have a crack at this. I think we are onto something here.

Post your results in here.

 

[jowak]

Just keep in mind. To change the AI braking distance just change the "distance_modifier=""" value inside the "ai_vehicle_track.xml". if you want AI breaking later DECREASE this value. Ex: distance_modifier="1.00" to distance_modifier="0.93". It´s gona change the AI laptimes too, they will drive faster, but only in the races, not practice or qualify (simulated).

And to make them faster in a corner the value is "brake_data curve_modifier=""", INCREASE this value.

Try the AI TOOL KIT from 2010,works with F12011 too, very very fast. You don´t need to convert *.xml files, and you can use screenshots of the tracks to know what curves you are changing.

[URL=http://imageshack.us/photo/my-images/801/f12011.png/]

Uploaded with ImageShack.us[/URL]

sry for the english.

 

[TGApples]

That looks like a very useful tool. Can't install it however as I don't have a copy of F1 2010 and it requires that to install. Any chance you could post a .zip of the installation directory?

 

[keanan]

Can the ai toolkit be used to improve 1st lap speed?

 

[Jace11]

I wondered whether it worked or not. I tried to install it tonight but it wouldn't install because I have removed F1 2010. My worry over using this is that if any corner designations have changed it will screw up the results. I will check.. but I think 2011 has many more variables per brake line. I will check...

2010 valencia corner 1 lap 1:
<brake_data type="curve" min_speed="0.0" max_speed="0.0" drift_speed="-5.0" distance_modifier="1.00" hold_line_id="0" hold_distance_modifier="0.00" curve_modifier="1.00" />

2011 valencia corner 1 lap 1:
<brake_data type="curve" min_speed="0.0" max_speed="0.0" drift_speed="-5.0" distance_modifier="15.00" hold_line_id="0" hold_distance_modifier="0.00" curve_modifier="1.00" curve_modifier_left="1.35" curve_modifier_right="1.35" curve_modifier_low_grip="0.90" curve_modifier_left_low_grip="0.90" curve_modifier_right_low_grip="0.90" curve_modifier_med_grip="1.00" curve_modifier_left_med_grip="1.35" curve_modifier_right_med_grip="1.35" />

Yeah I wouldn't use that tool.. although the number of brake lines for valencia is the same in 2010 and 2011 (18 corners / braking zones)... so it may be useful in identifying which corner is which...


I have fixed Abu Dhabi turn 9 btw... need to test it in wet conditions though too just to make sure. AI is still slow at this track generally but it should stop them spinning out at the chicane..

http://www.megaupload.com/?d=L8MNV91Y My Abu Dhabi mod. It is just the turn 9 fix and some team flags (ferrari, redbull etc) that replace the generic national flags. Try it, or just use the corner fix. Goes in F1 2011/tracks/circuits/ Backup your files originals first though...

<ovMinMetresSeparationAwayFromCorner float="10.0" /> <------ Not sure yet, does something for sure...
<ovSeparationDistanceAtBrakeLine float="10.0" /> <----- THIS HAS AN EFFECT on AI overtakes at tight corner (fast and slow it seems). I increased it (set to 50) and I had cars diving in, and some reckless contacts too. But it was at corners - vettel snuck in a small gap I left at abu dhabi chicane: GOOD, but then on Nurburgring Petrov rammed into my side in a high speed corner he had no hope of making cleanly - BAD. Often they do pass cleanly. I think a mid value and some other changes and we may get it working.

 

[TGApples]

Abu Dhabi problem fixed is a good start. I'm currently struggling to read the .xml files (notepad++ doesn't like 'em) though I can write them by hand in raw xml. Was going to have a look at some braking zones this evening.