Configuration Reference

Configuration Reference

Static configuration lives in lua/ge/extensions/taxiDriver/config.lua. Values are loaded once by the main GE extension. Runtime state does not belong in this module.

Supported languages

en, de, fr, es, it, pl, uk, ru, zh-CN

Runtime configuration

Key	Value	Unit / meaning
minRideDistance	1,000	metres
maxRideDistance	25,000	metres
minPickupDistance	400	metres
maxPickupDistance	3,500	metres
arrivalRadius	14	metres
passengerPickupRadius	10	metres; logical passenger pickup only
maxArrivalSpeedKmh	0.5	km/h wheel-speed sensor tolerance; AI targets zero
averageCitySpeedKmh	40	ETA reference speed
minimumDrivability	0.7	route/road cutoff
hudUpdateInterval	0.2	real seconds
boardingDuration	3	simulation seconds
alightingDuration	3	simulation seconds
completedDuration	4	simulation seconds
stopWaitingDuration	10	simulation seconds
forcedExitDuration	5	simulation seconds

Sanitized user-setting ranges

AI Driver also stores a master enabled flag. It defaults to false; the UI requires an experimental-feature confirmation before writing true.

Setting	Default	Range
uiScalePercent	100%	80–180% in 10% steps
appVolume	0.65	0–1
dynamicZoomIntensity	100%	0–200%
overspeedWarningKmh	10 km/h	0–30 km/h
economyMultiplier	1.00×	0.25–5.00×
deliveryOrderSharePercent	50%	0–100%
unlimitedRouteDistance	false	Boolean; removes the 25 km upper bound
aiDriver.aggressionPercent	40%	30–100%
aiDriver.followingTimeGap	2.3 s	1–4 s
aiDriver.minimumFollowingDistance	4 m	2–10 m
aiDriver.brakingDeceleration	3.5 m/s²	2–8 m/s²
aiDriver.trafficWaitSeconds	3 s	1–10 s
aiDriver.preset	balanced	novice, cautious, balanced, assertive, racer, custom
aiDriver.obeySpeedLimits	true	Boolean
aiDriver.laneDiscipline	true	Boolean
aiDriver.strictGpsRoute	false	Boolean; independent from the preset
aiDebugLogging	false	Boolean; streams dedicated AI JSONL diagnostics
randomEvents.<event>.enabled	Event-specific	Boolean
randomEvents.<event>.chancePercent	Event-specific	0–100%
randomEvents.policeCheck.preloadConfirmed	false	Must be true before police preload is allowed
fleet.enabled	true	Boolean
fleet.aiPreset	standard	careful, standard, or fast
fleet.ownerSharePercent	35%	10–90%
fleet.hiringFee	$75	$0–1,000
fleet.wagePerTenMinutes	$12	$0–250
fleet.maxDrivers	6	1–12
fleet.worldLabelDistance	400 m	50–1,000 m in 50 m steps
fleet.minimumJobDistanceKm	1.5 km	0.5–20 km
fleet.maximumJobDistanceKm	8 km	minimum–50 km
fleet.passengerJobs / deliveryJobs	true / true	At least one remains enabled
lanManualAddress	Empty	Optional private LAN IPv4 address; normalized and bind-tested before use
externalMapEnabled	true	Boolean
externalTerrainEnabled	true	Boolean
externalMapQuality	balanced	eco, balanced, or smooth
godMode	false	Boolean
debugLogging	true	Boolean

routePlanner.isDistanceAllowed() receives no maximum when unlimited routes are enabled. Minimum passenger/delivery distances and drivability checks are unchanged.

AI supervisor configuration

The native adapter has a small set of fixed safety rules:

Value	Current behavior
Player observer interval	100 ms
Curved-path samples	12
Observer jerk limit	2.5 m/s³
Emergency threshold	TTC below 0.55 s or required deceleration at least 8.5 m/s²
Pickup radius	7 m plus complete stop
Premature Route Done retries	3
Horn sequence	600 ms on, 200 ms off, 600 ms on

The old predictive planner, local spatial graph, custom bypass, reverse escape and world visualization constants are retained only in historical release notes and are not part of the active 4.0.0 RC runtime. See AI Driver Engine 0.39 and AI Driver and Recovery.

Fleet AI presets

Fleet workers use fixed profiles rather than the player's six-profile/Custom editor. They apply only native aggression/legal-speed values and the following/braking values used by the reduced-rate traffic guard.

Preset	Native aggression	Native speed mode	Effective no-progress window	Replans
careful	20%	Legal	45 s minimum	3
standard	30%	Legal	45 s minimum	3
fast	52%	Unrestricted	45 s minimum	3

All Fleet workers use native avoidCars="on", driveInLane="on", recover-on-crash, and enabled electrics. Traffic observation runs every 200 ms with six curved-path samples. Route monitoring runs every 250 ms with per-worker phase offsets. A retry starts from the current closest road segment, and an assignment is abandoned after the bounded retry budget.

See Fleet Operations for hiring, wages, assignments, persistence, and worker lifetime.

Offer configuration

Key	Value	Meaning
initialDelay	1.5 s	Delay before first generation job
intervalMin / intervalMax	1.2 / 2.2 s	Delay between completed offers
generationStepInterval	0.02 s	Cooldown between coroutine resumes
semanticScanBatchSize	6	Objects processed before yielding
semanticCandidateAttempts	40	Bounded semantic candidates per stop search
randomRouteAttempts	24	Bounded graph-derived fallback attempts
semanticLongitudinalJitterMax	70 m	Maximum variation along a semantic road segment
randomDistanceExponent	1.15	Random route-distance distribution
recentStopLimit	48	Remembered roadside points
recentStopSeparation	150 m	Preferred separation from recent points
routeDiversityMinimumShare	0.65	Pool share requiring strict endpoint diversity
routeDiversityAttempts	3	Candidate routes checked per normal offer
routeDiversityCellSize	400 m	Spatial route-history cell size
routeDiversityEndpointSeparation	280 m	Minimum endpoint separation under strict checks
recentRouteLimit	20	Remembered accepted/generated route pairs
minVisible / maxVisible	10 / 12	Dispatcher pool target
nextOfferDuration	5 s	Floating offer lifetime
nextOfferRetryMin / Max	5 / 10 s	Retry after expiry/failure
nextOfferProgressThreshold	0.80	Route progress required for floating offers
pickupTimeMultiplier	1.35	Pickup ETA multiplier
pickupTimeGraceSeconds	60 s	Added pickup grace
pickupTimeMinSeconds / Max	120 / 600 s	Pickup budget clamp
multiStopChance	0.24	Random unplanned type chance
multiStopVisibleMin / Max	2 / 3	Requested pool composition
rushVisibleMin / Max	2 / 4	Requested pool composition
multiStopCountMin / Max	2 / 2	Intermediate stop count
multiStopMinimumCandidates	20	Semantic candidates required
multiStopSegmentMin / Max	3,000 / 7,000 m	Normal leg range
multiStopRouteAttempts	18	Attempts per special leg
generationFailureBackoffMin / Max	4 / 7 s	Backoff after repeated failures
rushChance	0.28	Random unplanned type chance
rushBonusMin / Max	0.22 / 0.42	22–42% bonus
rushTimeRatioMin / Max	0.72 / 0.86	Normal ETA fraction

Base fare and rating

Key	Value
baseFare	$3.25
farePerKm	$1.55
farePerMinute	$0.20
ratingBonusThreshold	3.99
maxRatingBonus	0.15
maxTotalPenalty	0.50
ratingPenaltyScale	8
collisionRatingPenalty	0.15
minimumRideRating	1

Pickup penalty

Key	Value
pickupLateBasePenalty	0.05
pickupLatePenaltyPerSecond	0.000333
maxPickupLatePenalty	0.12

Passenger mood and stress

Key	Value
calmPenaltyIgnoreMaximum	0.70
calmPenaltyMultiplierThreshold	50
passengerStressThresholdMinimum	72
passengerStressThresholdCalmBonus	23
passengerMoodMaximumGain	40
passengerMoodPerfectRideGain	40
passengerMoodStressRecovery	0.8
passengerMoodPickupLateLoss	4
passengerMoodPickupLateStepLoss	2
passengerMoodSpeedingBaseLoss	2
passengerMoodCollisionBaseLoss	10
passengerMoodAggressionBaseLoss	4
passengerMoodRushExpiredLoss	6

Difficulty matrix

Speeding

Preset	Fixed tolerance	Ratio tolerance	Grace	Rate	Configured max
Elementary	15 km/h	25%	6 s	0.00025	6%
Easy	11 km/h	20%	4.5 s	0.00045	10%
Standard	8 km/h	15%	3 s	0.0008	15%
Professional	4 km/h	8%	1.5 s	0.0018	25%

Collision

Preset	Damage threshold	Damage scale	Damage factor	Event add	Configured max
Elementary	60	22,000	0.10	0.004	10%
Easy	35	17,000	0.14	0.007	14%
Standard	20	12,000	0.18	0.010	20%
Professional	10	8,000	0.24	0.020	30%

Collision events use a three-second cooldown in every preset.

Harsh manoeuvre

Preset	Longitudinal trigger/release	Lateral trigger/release	Event add	Extra rate / cap	Configured max
Elementary	0.88 / 0.55 g	0.78 / 0.48 g	0.002	0.006 / 0.004	5%
Easy	0.74 / 0.48 g	0.66 / 0.42 g	0.0035	0.010 / 0.006	8%
Standard	0.62 / 0.40 g	0.55 / 0.35 g	0.005	0.015 / 0.010	12%
Professional	0.48 / 0.32 g	0.42 / 0.28 g	0.010	0.025 / 0.018	20%

Aggression events use a three-second cooldown in every preset.

Internally, individual component accumulators allow up to twice the configured component max so a low-mood sensitivity multiplier can be represented. The final combined fare penalty remains capped at 50%.

Early-exit and abandonment reductions

Both passenger-triggered early exit and the extra percentage portion of driver abandonment use:

Preset	Reduction
Elementary	10%
Easy	20%
Standard	30%
Professional	45%

Driver abandonment additionally subtracts one full rating star before clamping to 0–5.

Realistic fuel configuration

Key	Value
fuelInitialLevel	5%
electricInitialLevel	30%
fallbackPricePerUnit	$1.00
fuelRatePerSecond	2 readable units/s
electricPercentRatePerSecond	4 percentage points/s
Gasoline fixed price	$0.93/L
Electric fixed price	$0.50/kWh
dashboardRefreshInterval	1 s

Estimated range uses 10 L/100 km for gasoline, 8.5 L/100 km for diesel, 12.5 L/100 km for kerosene, and 22 kWh/100 km for electric energy.

Cargo delivery configuration

Key	Value	Meaning
visibleMin / visibleMax	5 / 7	Requested delivery count in a full pool
minimumWeightKg / maximumWeightKg	2 / 250 kg	Generated cargo range
weightDistributionExponent	1.55	Bias toward lighter cargo
minimumDistance / maximumDistance	2,000 / 25,000 m	Delivery route range
routeAttempts	28	Route attempts per delivery candidate
diversityAttempts	3	Spatially distinct candidates checked
fareMultiplier	0.82	Pre-weight rate relative to passenger work
weightFareBonusThresholdKg	15 kg	Weight premium threshold
maximumWeightFareBonus	1.50	Maximum +150% premium at 250 kg
collisionDamageThreshold	20	Minimum telemetry damage delta
minimumImpactDamagePercent / maximumImpactDamagePercent	1 / 35%	Package damage per impact
impactDamageScale	4,500	Exponential impact-severity scale
ratingDamageThresholdPercent	5%	Damage threshold before rating loss

Phase IDs

inactive, searching, toFuelStation, toPickup, boarding,
toStop, stopWaiting, toDestination,
passengerStopDemand, passengerForcedExit, driverAbandoning,
alighting, complete, error