APEXOS // F1 Telemetry Hub

An AI-driven F1 command station leveraging linear regression and sigmoid probability curves to deliver real-time tire cliff predictions and optimal pit window strategy directives. (183 characters)

Comment

Inspiration

Formula 1 is a sport won and lost in milliseconds, where a single strategic miscalculation can ruin an entire weekend. While watching live race feeds, we realized that fans and amateur analysts rarely get a glimpse into the actual mathematical machinery running on an elite pit wall. We wanted to bridge that gap. Inspired by the high-stakes, data-driven environment of the Mercedes-AMG Petronas F1 Team, we set out to build APEXOS—an ultra-premium, high-fidelity real-time telemetry analytics platform designed to give engineers and strategy enthusiasts the ultimate race-day edge.

What it does

APEXOS acts as a cognitive pit-wall command station. It ingests historical and simulated live F1 telemetry datasets to deliver critical race-altering insights through a premium dashboard layout:

  • Driver Command Station: Tracks real-time cockpit feeds including live gear, speed (up to 318 KM/H), state of charge (SOC), a 10-segment engine RPM rev bar, and dynamic DRS status badges.
  • APEX Cognitive Strategy System: Uses predictive models to calculate tire cliff degradation points and optimum pit window entry directives.
  • Strategy Control Room: Simulates live pit stop undercut and overcut success probabilities on the fly.
  • Track Radar Feed: Renders an interactive <CircuitVisualizer /> mapping SVG vector paths for iconic tracks (like Monaco and Silverstone) complete with a live pulsing car locator dot tracking micro-sectors.

How we built it

We engineered APEXOS using a robust full-stack architecture optimized for low-latency calculations:

  • Frontend: Built with React and Vite, utilizing a custom carbon-fiber bento grid layout, neon status markers, and fluid SVG data visualizations for a true aerospace/motorsport aesthetic.
  • Backend: Powered by a high-performance FastAPI (Python) HTTP server that parses multi-gigabyte F1 telemetry structures.
  • Database & Caching: Compiled structural CSV records into a localized SQLite database instance (apexos_cache.db) for lightning-fast query execution.
  • The Mathematics: 1. Pace Degradation: Modeled using a filtered linear regression over stint lap times: $$y = m \cdot x + c$$ (Where \(y\) is lap time, \(x\) is lap number, and the first/last laps are automatically scrubbed as anomalies).
    1. Consistency Index: Derived from pace standard deviation (\(\sigma\)) mapped to a 0–100 scale: $$\sigma = \sqrt{\frac{1}{N - 1} \sum_{i = 1}^{N} (y_i - \bar{y})^2}$$
    2. Undercut Probability: Handled via a track-specific sigmoid distribution curve calculating compound grip advantage against historical pit lane variance margins.

Challenges we ran into

  • Telemetry Anomaly Scrubbing: Initial linear regression curves for tire wear were completely skewed by out-laps and in-laps where cars travel under pit-lane speed limits. We had to implement backend filtering logic to programmatically isolate pure green-flag push laps.
  • Dynamic SVG Path Mapping: Translating disparate circuit coordinate pairs into smooth, responsive, scaling SVG paths for the <CircuitVisualizer /> required rigorous coordinate normalization so that the pulsing car locator dot tracked accurately across complex geometry like Monaco's Fairmont Hairpin.
  • Data Scale & Latency: Simulating live telemetry streams from millions of historical lap records threatened to bottleneck the UI thread, forcing us to establish aggressive caching layers in SQLite.

Accomplishments that we're proud of

  • The UI Fidelity: Achieving a beautiful, cohesive, ultra-premium dark theme dashboard that feels like an authentic Mercedes pit-wall terminal rather than just another generic charts page.
  • Predictive Precision: Creating an engine that successfully flags an immediate "HIGHLY RECOMMENDED" pit directive card the moment undercut mathematical probability breaches \(\ge 80\%\).
  • Robust Backend Rigor: Writing clean, testable FastAPI endpoint specifications (/api/metadata, /api/driver-pace, /api/ai-lap-prediction) verified successfully via automated sanity testing scripts.

What we learned

  • Domain-Specific Constraints: We learned that sports analytics—especially motorsport—requires deep domain knowledge; you cannot simply throw standard machine learning models at data without accounting for circumstantial variables like traffic, track evolution, and tire compounds.
  • Component Optimization: We gained a profound respect for optimizing raw SVG rendering paths in React to prevent UI stutters during fast state updates.

What's next for APEXOS // F1 Telemetry Hub

  • Live WebSockets Integration: Transitioning from simulated live telemetry loops to a live WebSockets stream hooked directly up to active live-timing APIs during real-world Grand Prix weekends.
  • Weather Simulation Upgrades: Introducing dynamic rain/track-temperature coefficients into the sigmoid undercut formula to adapt strategies instantly to changing weather fronts.
  • Multi-Driver Comparative Overlays: Allowing engineers to overlay telemetry traces (Throttle, Brake, Steering angle) of both team drivers simultaneously on the same visual track path.

Built With

Share this project:

Updates

Submission history