APXGP · Use case 02

Every tyre call is adata problem.

Run a fictional pit wall at Austin. Follow telemetry from acquisition through quality checks to a tyre strategy, challenge the forecast, then approve the call. Every decision leaves an inspectable record.

Synthetic vendor demonstration. 2024-era Austin tyre selection and rules, not a replay or a 2026 rules simulation. No live F1 feed or trained AI model.

01
Acquire
Synthetic vehicle + trackside feeds
02
Validate
Tyre stream current; quality gate open
03
Compare
Legal plans + degradation sensitivity
04
Authorize
Human approval required

SYNTHETIC FEED · 10 Hz SIMULATIONSession t+0.0 s · tyre age of data 0.0 s · 0 missed tyre frames

APXGP.

Race strategy

RS–02 / Decision support

Simulation live

United States Grand Prix · Car 7

Circuit of the Americas, Austin

RAIN RISK · 30 MIN 18%

Speed95 km/hLateral G2.2Tyre surface est.96.6°

West / stadium · DryCentral / esses · DryEast / hairpin · Dry

12 km/h

Select a corner for estimated apex conditions, not a second live sensor. Circuit outline is sourced; corner speeds, timing loops and pit lane are schematic. Rain scenarios move west to east.

01 / Scenario desk

Change the conditions. Watch the call.

For a vendor walkthrough: inspect the baseline, inject rain or graining, test a feed dropout, then review and export the decision. Scenarios are scripted; rain arrival is compressed for the demo.

Signal correlation

A braking event, across channels.

Speed falls as brake pressure rises. Throttle closes on entry, then returns on exit. Pause the simulation to inspect a timestamp.

Collecting the first two simulated seconds of trace data.

Speed · km/h · left axisThrottle · % · right axisDashed: brake · bar · right axis

1 Hz display trace from the 10 Hz model. Lap position drives these synthetic channels; this is a correlation demonstration, not an instrumented car or a vehicle-dynamics validation.

Sensor-style channels: Vehicle bus: speed, throttle, brake pressure and steering. Thermal: surface IR, brake temperature and pressure. All synthetic, with simulated-time timestamps.
Derived estimates: Wear, carcass temperature, fuel, grip and strategy are calculated. Wear is not directly measured by a tyre sensor. Estimates inherit the quality of their inputs.
External context: Weather forecasts, race-control calls and radio reports are scripted scenarios. Opponent gaps are illustrative; no competitor private telemetry is available.

02 / Signals

Every reading has a source.

All values are generated locally at 10 simulated Hz. Sensor-style channels, model estimates and operator reports are labeled separately. This is not a hardware throughput benchmark.

Trackside / environment

Synthetic weather + scenario forecast

Air temperature: 27.2°C
Track temperature: 36.0°C
Humidity: 48%
Wind: 12 / 270°km/h
Rubbered-in index: 74/ 100

10 min

20 min

13%

30 min

18%

Rain probability

West / stadium: 0.02/ 1
Central / esses: 0.01/ 1
East / hairpin: 0.01/ 1

Car / chassis

Synthetic channels + derived estimates

224km/h

Throttle: 90%
Brake pressure: 0bar
Steering: -7.8°
Lateral G: 1.15g
Longitudinal G: 0.72g
Ride height F / R: 19.5 / 44.3mm
Fuel estimate: 76.0kg
ERS estimate: 59%

Car / tyres

Synthetic IR / pressure; derived wear

87°C

91 / 87 / 84

Carcass est.: 80°C
Pressure: 1.73bar
Wear estimate: 26%
Slip estimate: 0.030

89°C

94 / 89 / 86

Carcass est.: 82°C
Pressure: 1.74bar
Wear estimate: 26%
Slip estimate: 0.030

89°C

93 / 89 / 86

Carcass est.: 82°C
Pressure: 1.74bar
Wear estimate: 26%
Slip estimate: 0.030

91°C

96 / 91 / 88

Carcass est.: 84°C
Pressure: 1.75bar
Wear estimate: 27%
Slip estimate: 0.030

CoolWorking rangeHot

Surface IR and pressure are synthetic sensor-style readings. Carcass temperature, wear and slip are model estimates. Load follows the outside tyre: left-side load rises through the long right at T16–T18.

Brakes + power unit

Synthetic thermal channels

FL: 445°C
FR: 445°C
RL: 375°C
RR: 375°C
Oil temperature: 116.0°C
Water temperature: 100.2°C
Engine mode: Race

Timing / race control

Simulated loop crossings + gap estimates

Last lap: Awaiting full lap
Best observed lap: Awaiting full lap
Last completed S1: Awaiting crossing
Last completed S2: Awaiting crossing
Last completed S3: Awaiting crossing
Gap ahead (est.): +2.80s
Gap behind (est.): +1.60s
Assumed net pit loss: 20s
Laps remaining: 39

Timing resolution: 0.1 s. The opening partial lap is excluded. Gaps are scenario estimates, not a full-field race simulation.

Generated since session start0 frames · 0 numeric values

03 / Interpretation

Evidence in. A decision out.

Candidate plans are compared lap by lap. The fastest forecast leads; the engineer still has the final say.

What is moving the call

Heuristic priority bars, not causal attribution or measured feature importance.

Modeled degradation0.084 s/lap
Modeled fuel effect2.28 s/lap
Scenario rain forecast18% / 30 min
Scripted driver report4/5
Scenario temperature trend-0.3 °C / 10 min
Fresh-tyre payback estimateLap 31
Safety-car state / assumed prior22% · Illustrative prior
Undercut threat estimate1.6 s
Estimated gap ahead2.8 s
Pit-exit traffic assumptionClear exit

Scripted driver radio"Balance is good. Grip is there."

Current model candidate

Medium

Box at end of lap

Within 2 s of best 20–28

Sensitivity checks

3 / 3

2.5 s ahead of next-best legal plan

Current tyre wear makes fresh rubber worth the pit loss; medium at lap 24 leads the next-best legal plan by 2.5 s and stays fastest in 3 of 3 degradation checks. Weather and traffic remain assumptions.

Dry-race rule: a second slick compound is still required. Available sets: S ×1 · M ×2 · H ×1 · I ×3 · W ×2.

Ranges vary degradation by ±15%; weather is held to the scenario forecast. Checks count how often this plan stays fastest, not a probability of success.

Legal candidate strategies

Soft→Medium

1:03:36 remaining · 1:03:31–1:03:42

Lap 1M @ 2456

Soft→Medium→Soft

1:03:39 remaining · 1:03:36–1:03:42

Lap 1M @ 18 · S @ 3856

Soft→Hard

1:03:42 remaining · 1:03:39–1:03:46

Lap 1H @ 2356

Soft→Medium→Hard

1:03:50 remaining · 1:03:48–1:03:52

Lap 1M @ 19 · H @ 3956

Race engineer review

Awaiting your decision

04 / Decision record

Show the evidence behind the call.

The export captures the seed, scenario inputs, source quality, candidate forecasts and engineer actions. Session-only data; reset clears it. The last 60 actions and 60 seconds of traces are retained.

t+0.0 s · lap 18
Session opened. Fictional car; 2024-era Austin rules; simulated sources.

What this demonstrates for vendors

Time-aligned signals, visible data quality and an approval trail that can be inspected after the decision. A production integration would need authenticated ingestion, a time-series store, calibrated models and a durable audit log. Those services are not connected in this demo.

The model, explained

How the model decides.

Model the degradation

Hand-tuned curves connect tyre age to lap-time loss. Heat and scripted graining steepen the curve. A thermal cliff penalizes long stints; fuel burn reduces lap time. No model is trained here.

Tyre ageLap-time loss

Find the crossover

A fresh set has to earn back the stop. Compare remaining race time, including an assumed 20 s net pit loss (12 s under the safety car), traffic and tyre warm-up. Net loss is not stationary service time; the simulated tyre change takes 2.6 s.

Fresh-tyre gain−Pit loss=Net gain

Compare the weather signals

A forecast is not a wet-track observation. Scripted rain reaches three zones in sequence. The wet-tyre branch becomes eligible at a modeled surface crossover. Losing the tyre feed blocks approval until it recovers.

Scenario forecastSurface estimateScripted radio

Check legality and sensitivity

Respect the available tyre sets and the two-compound dry-race rule, waived after using intermediates or wets. Compare each plan at 85%, 100% and 115% degradation. The pit window shows nearby laps within 2 s of the fastest plan.

Lower wearBaselineHigher wear

A deterministic, hand-tuned model, not trained AI. Candidate times are conditional on forecast weather; the range varies degradation by ±15% only. It is not a confidence interval. The 20 s / 12 s net pit losses, tyre inventory, weather, gaps and radio are demonstration assumptions. No full-field race, penalties, red flags or calibrated standing-water safety model is implemented.

Reference profile: Formula 1, 2024 United States Grand Prix. 56 laps, 5.513 km; C2 hard, C3 medium, C4 soft. The fictional session does not reproduce the race result.

A longer view

How the tyre call was made, then and now

1950s–1980s
The stopwatch era
Hand-held stopwatches, visual tyre checks and driver feedback shaped the call. Telemetry and radio became more common toward the end of this period.
1990s
Refuelling and the chief strategist
Refuelling returned in 1994. Fuel load, stop count and tyre life became a spreadsheet problem. The three-stop approach at Hungary in 1998 showed what mid-race lap-time arithmetic could do.
2000s
Telemetry and race simulators
Hundreds of channels streamed to the garage. Teams compared possible races in simulators; a four-stop win at Magny-Cours in 2004 challenged conventional strategy.
2010s
The degradation era
Tyre wear and the undercut became central. Factory strategy groups fed predictions to the pit wall using thermal cameras, pressure readings and traffic models.
2020s
AI-assisted strategy
Learned degradation models and weather fusion refresh the forecast throughout the race. The engineer and driver still make the call: a model cannot feel a kerb strike or hear confidence in a voice.

Back to all workflows

Illustrative concept, not client results. Fictional APXGP car 7; simulated data throughout. No official team, tyre supplier or championship affiliation.

The next move is a conversation

Make your next move a better one.

Bring the workflow that slows you down, the decision on your desk, or the idea you want to put to work. We will help you find a sensible next step.

Tell us what you have in mind

A specific challenge. An honest conversation. A clear next step.