Information for Customizing AI drivers in AMS2

To override AI driver names, personalities and nationalities, user can create a .xml file named with the vehicle class name into the UserData\CustomAIDrivers\ folder of his Automobilista 2 steam folder (usually it is C:\Program Files (x86)\Steam\steamapps\common\Automobilista 2\UserData\CustomAIDrivers\).

For reference, all vehicle class names can be found in the spoiler below:

Spoiler: Full list of vehicle class names

ARC_Cam.xml
Carrera Cup.xml
Cat620R.xml
Cat_Academy.xml
Cat_Superlight.xml
Cat_Supersport.xml
CopaClassicB.xml
CopaClassicFL.xml
CopaFusca.xml
CopaTruck.xml
CopaUno.xml
DPI.xml
F-3.xml
F-Classic_Gen1.xml
F-Classic_Gen2.xml
F-Classic_Gen3.xml
F-Classic_Gen4.xml
F-Dirt.xml
F-Junior.xml
F-Inter.xml
F-Hitech_Gen1.xml
F-Hitech_Gen2.xml
F-Reiza.xml
F-Retro_Gen1.xml
F-Retro_Gen2.xml
F-Retro_Gen3.xml
F-Trainer.xml
F-Trainer_A.xml
F-Ultimate.xml
F-Ultimate_Gen2.xml
F-USA_Gen1.xml
F-USA_Gen2.xml
F-USA_Gen3.xml
F-USA_2023.xml
F-V10_Gen1.xml
F-V10_Gen2.xml
F-V12.xml
F-Vee.xml
F-Vintage_Gen1.xml
F-Vintage_Gen2.xml
G40Cup.xml
G55Supercup.xml
GT1.xml
GT3.xml
GT3_Gen2.xml
GT4.xml
GT5.xml
GTE.xml
GTOpen.xml
Group A.xml
Group C.xml
Hot Cars.xml
Hypercars.xml
Kart125cc.xml
KartGX390.xml
KartRental.xml
KartShifter.xml
Kartcross.xml
LancerCup.xml
LMDh.xml
MiniChallenge.xml
Montana.xml
OldStock.xml
Opala79.xml
Opala86.xml
P1.xml
P1Gen2.xml
P2.xml
P3.xml
P4.xml
Procar.xml
RX.xml
SprintRace.xml
StockCar99.xml
StockCarV8.xml
StockCarV8_2020.xml
StockCarV8_2021.xml
StockCarV8_2022.xml
StockCarV8_2023.xml
StockCarV8_2024.xml
Street.xml
Supercars.xml
SuperKart.xml
SuperV8.xml
TC60S.xml
TC60S2.xml
TC70S.xml
TSICup.xml

For example, if he wants to override AI drivers for the Formula Classic Gen3 vehicle class, he should create a F-Classic_Gen3.xml file inside \UserData\CustomAIDrivers\ folder, like this:

C:\Program Files (x86)\Steam\steamapps\common\Automobilista 2\UserData\CustomAIDrivers\F-Classic_Gen3.xml

Example of F-Classic_Gen3.xml file content:

Code:

<?xml version="1.0" encoding="UTF-8"?>
<custom_ai_drivers>
    <driver livery_name="Ministry Motorsport #51">
        <name>John Smith</name>
        <country>USA</country>
        <race_skill>0.75</race_skill>
        <qualifying_skill>0.823</qualifying_skill>
        <aggression>0.3</aggression>
        <defending>0.51</defending>
        <stamina>0.998</stamina>
        <consistency>0.2</consistency>
        <start_reactions>0.36</start_reactions>
        <wet_skill>0.25</wet_skill>
        <tyre_management>0.85</tyre_management>
        <fuel_management>0.0</fuel_management>
        <blue_flag_conceding>0.53</blue_flag_conceding>
        <weather_tyre_changes>0.12</weather_tyre_changes>
        <avoidance_of_mistakes>0.39</avoidance_of_mistakes>
        <avoidance_of_forced_mistakes>0.69</avoidance_of_forced_mistakes>
        <vehicle_reliability>0.847</vehicle_reliability>
    </driver>
    <driver livery_name="ABS Racing #14">
        <name>William Shakespeare</name>
        <country>GBR</country>
        <race_skill>0.12</race_skill>
        <qualifying_skill>0.0</qualifying_skill>
        <consistency>1.0</consistency>
    </driver>
</custom_ai_drivers>

So, livery_name determines to what livery each driver will be assigned. In the example above we assigned drivers to 2 liveries of the Formula Classic Gen3 vehicle class. You can assign drivers to as many liveries as you want.

livery_name is the name of the livery (case sensitive) as seen in the vehicle selection screen in the game, for example livery_name="Ministry Motorsport #51" will assign the driver to the livery from this screenshot:

View attachment 17943

So, for each vehicle class that you want to assign AI drivers to, you must create a .xml file. For example, now we want to override 3 AI drivers of the Chevrolet Cruze Stock Car 2019 vehicle class. The class filename is StockCarV8.xml (as you can see in the spoiler above) so you create a file StockCarV8.xml in \UserData\CustomAIDrivers\ folder.

Example of StockCarV8.xml file content:

Code:

<?xml version="1.0" encoding="UTF-8"?>
<custom_ai_drivers>
    <driver livery_name="Eurofarma RC - R.Mauricio #90">
        <name>Lorem Ipsum</name>
    </driver>
    <driver livery_name="Full Time Sports - R.Barrichello #111">
        <stamina>1.0</stamina>
    </driver>
    <driver livery_name="Eurofarma RC - D.Serra #29">
        <race_skill>1.0</race_skill>
    </driver>
</custom_ai_drivers>

Now, to race with a custom grid with only the drivers that you have overriden, you simply start a event in AMS2 with the desired opponent classes and select a number of AI opponents less or equal to the total number of AI drivers that you have overriden for those classes.

Notes:

• The valid personality values range is between 0 and 1 (inclusive), except vehicle_reliability (see below). All AI personality parameters are calibrated so that 0.5 is an realistic-ish average driver's value, 1 is mapped to a realistic-ish high extreme, and 0 is mapped to a realistic-ish low extreme. Its a sane range so 0 and 1 are valid and reasonable for the extremes though you would expect the spread to be weighted towards 0.5. Bell curve stuff. Thats the expectation, but of course always test the values of your custom AI drivers values to see if they are matching the result you would expect, and adjust it accordingly for your custom sets.
  
• Any missing or invalid values are replaced with the original values of the AI driver that would be picked for that car in that event.
• For now this feature is disabled in multiplayer mode.
  
• What each personality value does:
  • name: Driver name.
  • country: 3 letter country code. This is used for displaying the country flag.
    
  • race_skill: Race session driver skill. It is mapped into a smaller range based on the "Opponent Skill Level" slider setting. Example:
    • For example, at 90% "Opponent Skill Level" slider setting: it builds a range from 85% to 95% (just example, the exact numbers vary for each vehicle model), so a 1.0 race skill driver would be 95% skill and a 0.0 race skill driver would be 85% skill in this example.
  • qualifying_skill: Same as above but for practice/qualifying sessions. Its completely independent from the race_skill. One detail is that a lower qualifying_skill value increases the likelihood of AI programmed mistakes during qualifying hotlaps.
    
  • aggression: Driver aggression. It is scaled by the "Opponent Aggression" setting:
    • At Low "Opponent Aggression" setting, the 0.0-1.0 aggression value is mapped into a 0.0-0.8 range.
    • At Medium "Opponent Aggression" setting, the 0-1 aggression value is mapped into a 0.2-1.0 range.
      
    • At High "Opponent Aggression" setting, the 0-1 aggression value is mapped into a 0.6-1.0 range.
    • At Max "Opponent Aggression" setting, all drivers have 1.0 aggression.
  • defending: How much the driver will try to defend his position. Is also scaled by the "Opponent Aggression" slider setting. One detail is that a lower defending value increases the likelihood of AI programmed mistakes when under pressure.
    
  • stamina: Lower stamina value means the driver loses more of his skill during the session, and also makes the driver to become tired earlier (which increases the likelihood of AI programmed mistakes)
    
  • consistency: Lower consistency value means the driver skill can be randomly reduced more (basically, the lower the consistency value of a driver, the more skill he can lose from the consistency logic). The randomness is biased towards not losing too much skill, so there is a higher probability for the drivers to lose few skill than to lose much skill. Its has a per-weekend effect (determined upon loading of a track) and also a per-lap effect (determined every new lap or so). One detail is that a lower consistency value slightly increases the likelihood of AI programmed mistakes during the session.
    
  • start_reactions: Lower start_reactions value means the driver will take more time to react to the race green flag and is more likely to make race start programmed mistakes at the moment of the race green flag (like losing rear grip with some smoke).
    
  • wet_skill: How good he is on a wet track. Controls how much he will slow down in curves as the track gets wet (lower wet_skill value means he will slow down more), and how likely he will make programmed mistakes related to wetness, like losing grip in puddles or in wet surface (lower wet_skill value increases the likelihood of those mistakes). For example, if a high race skill driver does a curve at 100km/h when its dry, and he has high wet skill, then at wet track he will do that curve at say 90km/h, while if he had low wet skill he would do it at say 85km/h. So if a low race skill driver does that curve at 95km/h in dry, and he has high wet skill, then he might do it at say 88km/h in the wet. Another way to explain it is: it controls how much percent slower a driver will be in a fully wet corner compared to his own speed in the dry in that same corner. For example: in a given class all drivers with 1.0 wet skill go 10% slower on a fully wet corner compared to their own speed on that same corner fully dry. And all drivers with 0.0 wet skill go 20% slower on a fully wet corner compared to their own speed on that same corner fully dry. (note those numbers are just examples, the actual percentage varies per class).
    
  • tyre_management: How good he is in preventing tyre wear (higher values means he will have less tyre wear and consequently he will be able do keep doing good laptimes for a longer period and pit later due to that). This doesn't change the behavior of the driver (i.e. he won't drive differently to try to save tires), just the tyre wear.
  • fuel_management: For now this parameter works in oval tracks only (but in future it will be extended to all track types); The higher the value, the more the AI will try to save fuel in some strategic situations instead of pushing, like when it sees it can maybe save a pitstop, it will try to stay in draft and coasting instead of overtaking while at the same time trying to not lose much distance to the race leader. It doesnt mean high value is better than low value, its just a characteristic of the driver.
    
  • blue_flag_conceding: Drivers with high blue_flag_conceding will work harder to concede the position when under blue flag.
  • weather_tyre_changes: Drivers with high weather_tyre_changes are more likely to make pitstops for changing tyres when the track wetness state changes. It doesnt mean 1.0 is better than 0.0, its just a characteristic of the driver, not really a skill. Sometimes staying longer on the track with the wrong tyres is better, sometimes its worse.
    
  • avoidance_of_mistakes: Drivers with lower avoidance_of_mistakes value are more likely to make AI programmed mistakes during the session in general (like understeer, oversteer, recoverable and non-recoverable mistakes).
  • avoidance_of_forced_mistakes: Drivers with 1.0 value for avoidance_of_forced_mistakes wont have their chances of mistakes increased when under pressure (when defending position). Drivers with lower value for avoidance_of_forced_mistakes will have their chances of mistakes increased when under pressure (compared to their chances of mistakes when not under pressure).
  • vehicle_reliability: In short this is a ratio between the lowest and highest possible reliability for the car/class in question. Read instructions on how to set this in its own section below.

New in update V1.4.0.0:

Two new custom AI parameters were added: fuel_management and vehicle_reliability. The former is explained above, and the later is explained in detail below:

Vehicle Reliability
To make sense of vehicle_reliability, you need to be aware of the reliability range we have set for each class. We list the ranges as minimum and maximum mean minutes between failures below. If you're working on a mean minutes between failures basis you need to do some math to turn it into the final reliability for a driver in the class in question - best to do in a spreadsheet or other tool when working on a set.

Gather minimum (y) and maximum (z) mean minutes between failures for the class you're working on (see below), then consider your driver/team target mean minutes between failures (x) that you may have derived from some real race statistics for example. You can use this formula to calculate vehicle_reliability:

Code:

sqrt((x-y) / (z-y))

If we had a minimum of 25 mins, maximum of 700 mins, and our target was 500 mins between failures, result of that formula and thus our vehicle reliability would be 0.839. You simply insert it to a driver's personality values:

Code:

<vehicle_reliability>0.839</vehicle_reliability>

Alternatively if you wish to just set a reliability and see what happens, know that values above 0.6 are generally better for each class - relative to how reliable or unreliable the class should be. But for maximum chaos you can of course give low reliabilities a try.

If you wish to go below 0.0 or above 1.0, that can be done too. But know that if you go negative, the formula changes to this:

Code:

-((y/x)-1)

When negative, vehicle reliability is used as a divider for the minimum mean time so we need this extra step.

Note that mean minutes between failures is simply a way to quantify reliability. In the end there can be a lot of variance and generally speaking if you want someone to last for 60 minutes consistently, better make them much more reliable than that.

Current failure time ranges
Very Low (5-400):

• Copa Truck
• F-Junior
• Group A
• Karting 125CC, GX390, Rental, Shifter & Super Kart
  
• M1 Procar
• Vintage Touring Cars T2

Low (25-700):

• F-Vintage Gen 1-2
• F-Retro Gen 1-3
• F-Classic Gen 1-4
• F-HiTech Gen1-2

Medium (50-1000):

• Aussie Racing Camaro
• Caterham 620R, Academy, Superlight & Supersport
  
• Copa Classic B & FL
• Copa Fusca
  
• Copa Uno
• F-3
  
• F-Dirt
• F-Inter
• F-USA Gen 1 & Gen3
• F-V12
• F-Vee
• Ginetta G40 Cup
• GT5
  
• Hot Cars
• Kart Cross
  
• Lancer Cup
  
• Old Stock Race
  
• P2, P3 & P4
• RX
  
• Sprint Race
  
• Stock Car Brasil 1979, 1986 & 1999
• Street Cars (Chevrolet Camaro SS)
  
• Vintage Touring Cars T1

High (100-1500):

• F-V10 Gen 1-2
• F-USA Gen 2
• Group C
• GT Classic

Very High (300-2000):

• Copa Montana
• DPi
  
• F-Reiza
• F-Ultimate Gen 1-2
• F-USA Gen 4
  
• Ginetta G55 Supercup
  
• GT1
• GT3 Gen1-2
  
• GT4
  
• GTE
• GT Open
• LMDh
  
• P1 Gen1-2
  
• Porsche Cup
  
• Stock Car Brasil 2019-2024
• Street Cars - Hyper & Super
  
• Super V8

New in update V1.3.5.0:

Starting on AMS2 V1.3.5.0, we implemented forced and unforced mistakes logic for AI, with support for user overriding the likelihood of forced and unforced mistakes personality values for AI drivers (check above to learn how to override the new avoidance_of_mistakes and avoidance_of_forced_mistakes personality values).

The AI may make understeer or oversteer programmed mistakes in the following situations: on race reaction to green flag (depending on his start_reactions personality value), on puddles (depending his wet_skill personality value), on brakezones, on mid corners and on throttle appliance in corner exit, depending on his avoidance_of_mistakes personality values. The types of mistakes also depend on if the car is FWD or RWD.

Some mistakes are programmed to be recoverable (i.e. minor mistakes, wont cause AI to spin or go off track, are more likely to occur) and some mistakes are programmed to be non-recoverable (i.e. major mistakes like spin or go off track, are less likely to occur).

The likelyhood of a mistake on a given moment depends on several factors, like if he has poor related personality traits like the avoidance_of_mistakes personality value and the consistency value, or is suffering pressure (even more if the pressure occurs near the end of race or if he has poor defending or poor avoidance_of_forced_mistakes personality value), or track is wet and he has poor wet_skill, or he has slicks in wet, or he is in qualifying hotlap and has poor qualifying_skill value, or he is tired (due to poor stamina for example), or his tyres are worn. Mistakes will be less likely if that part of road has walls near (to simulate him being more careful to mistakes on tracks with walls near road).


New in update V1.3.4.0:

Starting on AMS2 V1.3.4.0, now you can optionally override AI driver values for specific tracks. See example below:

Code:

<?xml version="1.0" encoding="UTF-8"?>
<custom_ai_drivers>
    <driver livery_name="Ministry Motorsport #51">
        <name>John Smith</name>
        <country>USA</country>
        <race_skill>0.75</race_skill>
        <qualifying_skill>0.823</qualifying_skill>
        <aggression>0.3</aggression>
        <defending>0.51</defending>
        <stamina>0.998</stamina>
        <consistency>0.2</consistency>
        <start_reactions>0.36</start_reactions>
        <wet_skill>0.25</wet_skill>
        <tyre_management>0.85</tyre_management>
        <fuel_management>0.0</fuel_management>
        <blue_flag_conceding>0.53</blue_flag_conceding>
        <weather_tyre_changes>0.12</weather_tyre_changes>
        <avoidance_of_mistakes>0.39</avoidance_of_mistakes>
        <avoidance_of_forced_mistakes>0.69</avoidance_of_forced_mistakes>
        <vehicle_reliability>0.847</vehicle_reliability>
    </driver>
    <driver livery_name="Ministry Motorsport #51" tracks="Azure_Circuit_2021,Long_Beach,Salvador_Street_Circuit">
        <race_skill>0.99</race_skill>
        <qualifying_skill>0.998</qualifying_skill>
        <consistency>1.0</consistency>
        <wet_skill>0.95</wet_skill>
        <blue_flag_conceding>0.75</blue_flag_conceding>
        <stamina>1.0</stamina>
    </driver>
    <driver livery_name="Ministry Motorsport #51" tracks="Jacarepagua_OVAL">
        <race_skill>0.5</race_skill>
        <tyre_management>0.4</tyre_management>
        <defending>0.0</defending>
        <stamina>0.9</stamina>
        <aggression>0.0</aggression>
        <avoidance_of_mistakes>0.1</avoidance_of_mistakes>
        <avoidance_of_forced_mistakes>0.9</avoidance_of_forced_mistakes>
    </driver>
    <driver livery_name="ABS Racing #14">
        <name>William Shakespeare</name>
        <country>GBR</country>
        <race_skill>0.12</race_skill>
        <qualifying_skill>0.0</qualifying_skill>
        <consistency>1.0</consistency>
    </driver>
</custom_ai_drivers>

In the above example, you can see there are 3 driver entries for the same livery (Ministry Motorsport #51).

The first entry doesn't contain a tracks list, so it means those are the base values of the driver of that livery on any tracks. For example, on any tracks, his race_skill will be 0.75.

The second entry contains a tracks list with comma separated track names ("Azure_Circuit_2021,Long_Beach,Salvador_Street_Circuit"), so it means the driver of that livery has some specific personality values to those 3 tracks. So on Azure, Long Beach and Salvador, his race_skill will be 0.99 instead of 0.75, for example. The fields that arent present will still use the base value of that driver: for example, aggression isnt present, so it will use his base value for aggression on any tracks which is 0.3.

The third entry contains a tracks list with only one track name ("Jacarepagua_OVAL"), so it means the driver of that livery has some specific personality values to this track. So on Jacarepagua_OVAL track, his race_skill will be 0.5 instead of 0.75, for example. The fields that arent present will still use the base value of that driver: for example, qualifying_skill isnt present, so it will use his base value for qualifying_skill on any tracks which is 0.823.

You can add as many track overrides as you want, for as many drivers (liveries) as you want.

The full list of track names is in the spoiler below:

Spoiler: Full list of track names

Adelaide_Historic
Adelaide_Modern
Ascurra_Dirt
Ascurra_RX
Azure_Circuit_2021
Bathurst_1983
Bathurst_2020
BrandsHatch_GP
BrandsHatch_Indy
Brasilia_Full
Brasilia_Outer
Buenos_Aires_Circuito_6
Buenos_Aires_Circuito_6T
Buenos_Aires_Circuito_7
Buenos_Aires_Circuito_8
Buenos_Aires_Circuito_9
Buenos_Aires_Circuito_12
Buenos_Aires_Circuito_15
Buskerud_Long
Buskerud_Short
CadwellPark
CampoGrande1
Cascavel2
Cleveland_GP
Barcelona_1991
Circuit_de_Catalunya_GP
Circuit_de_Catalunya_GP_NC
Circuit_de_Catalunya_NATL_NC
Circuit_de_Catalunya_RX
Cordoba_GP
Cordoba_International
Cordoba_NATL
CuritibaInternacional
Curitiba_outer
CurveloLong
CurveloShort
Daytona_Speedway_Tri_Oval
Daytona_Road_Course
Daytona_Nascar_Road_Course
Donington_GP
Donington_Nat
Cascais
Cascais_1988
Cascais_Alternate
Fontana_OVAL
Fontana_SCC
Foz_RX
Galeao_Airport
Gateway_OVAL
Gateway_RC1
Gateway_RC2
GoianiaMixed
GoianiaOuter
GoianiaShort
CopaSaoPauloStage2
GranjaVianaKart101
GranjaVianaKart102
GranjaVianaKart121
Guapore
Hockenheim_GP
Hockenheim_ShortA
Hockenheim_ShortB
Hockenheim_National
Hockenheim_2001
Hockenheim_1988
Hockenheim_1988_short
Hockenheim_1977
Hockenheim_RX
Ibarra2
Ibarra_Reverse
Indianapolis_2022_Oval
Indianapolis_2022_RC
Imola_GP_2018
Imola_GP_1972
Imola_GP_1988
Imola_GP_2001
Interlagos_1991
Interlagos_1993
Interlagos_GP
Interlagos_SCB
Interlagos_Historic
Interlagos_Historic_Outer
InterlagosKart1
InterlagosKart2
InterlagosKart3
Jacarepagua_Historic
Jacarepagua_2005
Jacarepagua_OVAL
Jacarepagua_SCB
Jacarepagua_SHORT
Jerez_GP1988
Jerez_GP2019
Jerez_Standard
Kansai_GP
Kansai_Classic
Kansai_East
Kansai_West
Kyalami_2019
Kyalami_Historic
Laguna_Seca_2020
Le_Mans_24h
Le_Mans_Circuit_Bugatti
Londrina_short
Londrina_long
LondrinaKart1
LondrinaKart2
Long_Beach
Montreal_Historic
Montreal_Historic_1991
Montreal_Modern
Monza_1971_10k
Monza_1971_10knc
Monza_1971
Monza_1991
Monza_1971_Junior
Monza_2020
Monza_2020_Junior
Nordschleife_2020
Nordschleife_2020_24hr
Nurb_GP_2020
Nurb_GP_2020_Veedol
Nurb_GP_2020_Sprint
Nurb_GP_2020_Sprint_S
Nurb_1971_Beton
Nurb_1971_Gesamt
Nurb_1971_Nords
Nurb_1971_Suds
Nurb_2020_RX
Ortona1
Ortona2
Ortona3
Ortona4
OultonPark_International
OultonPark_Classic
OultonPark_Fosters
OultonPark_Island
Road_America_RC
Road_America_RCB
Salvador_Street_Circuit
SantaCruzDoSul
Silverstone_1975
Silverstone_1975_No_Chicane
Silverstone_1991
Silverstone_2001
Silverstone_2001_International
Silverstone_2001_National
Silverstone_GP
Silverstone_Intl_2019
Silverstone_Natl_2019
Snetterton_100
Snetterton_200
Snetterton_300
Spa_Francorchamps_1970 (this is the Chicane version)
Spa_Francorchamps_1970_1000km (this is the No Chicane version)
Spa_Francorchamps_1993
Spa_Francorchamps_2020
Spa_Francorchamps_2022
Spa_Francorchamps_2022_RX
Speedland1
Speedland2
Speedland3
Speedland4
Spielberg_Historic
Spielberg_Vintage
Spielberg_Modern
Spielberg_Short
Taruma_Internacional
Taruma_Chicane
Termas_Rio_Hondo
Tykki_Dirt1
Tykki_Dirt2
Tykki_RX
Tykki_Tarmac
VeloCitta1
VeloCittaTD
VeloCittaClubDay
Velopark_2010
Velopark_2017
Virginia_Full
Virginia_Grand
Virginia_North
Virginia_Patriot
Virginia_South
Watkins_Glen_GP
Watkins_Glen_GPIL
Watkins_Glen_S
Watkins_Glen_SIL

Note: This is the only official post about custom AI drivers, and we keep it updated. Any posts below this one may be outdated or inaccurate.

 

Its the name of the livery that the driver belongs to (case sensitive), as seen in the vehicle selection screen in the game, check this screenshot:

 

The example isnt just for the #51 car. The example contains 5 overriden drivers, 1 for each livery_name specified, only one of them is for the #51 car. You can add drivers for as many liveries as you want.