Inspiration
We were inspired by the challenge race officials (like the FIA in F1) face in enforcing "Parc Fermé" rules. These rules state that cars generally cannot be modified between qualifying and the race. However, teams might try to make subtle, illegal changes and disguise them as normal wear and tear from driving. We wanted to create a high-tech, automated system to help catch these non-compliant modifications accurately.
What it does
The Multi-Spectral Parc Fermé Sentinel acts like a super-smart, incorruptible scrutineer. It uses special cameras (multi-spectral) to scan a race car at three key times: before qualifying (T0), right after qualifying (T1), and just before the race (T2). Its core AI brain is trained to understand "Legal Aging Patterns" – the normal scuffs, scratches, and temperature changes expected from racing. By comparing the car's state at T1 and T2, the system can tell the difference between this normal wear and "Suspicious Changes" like parts being illegally replaced, adjusted, or repaired. It generates a detailed report for officials, flagging potential rule violations with high confidence.
How we built it (The Plan)
This is our plan to build the Sentinel:
- Data Capture: Use multi-spectral cameras (capturing visible light, thermal, potentially UV or others) to create detailed 3D scans of the car at T0, T1, and T2.
- AI Training Data: Collect a massive dataset showing known legal wear patterns from historical F1 sessions across different tracks, weather conditions, and components.
- Build the "Legal Aging" Model: Train a predictive AI model (likely using techniques like convolutional neural networks or transformers adapted for 3D/multi-spectral data) on the historical data. This model learns to predict how a component should look at T2 based on its state at T1 and the session's activity.
- Develop the Plausibility Engine: Create the core logic that compares the actual car state at T2 with the AI's predicted state. Significant deviations (like "Reverse Aging" or asymmetrical changes) are flagged as suspicious.
- Tools: We plan to use Python, AI frameworks like TensorFlow/PyTorch, and libraries for image/3D scan analysis (like OpenCV, PCL).
Challenges we ran into (Anticipated)
We anticipate several challenges:
- Massive Data Needs: Getting enough diverse, high-quality data showing legal wear across all car parts, tracks, and conditions will be difficult.
- Accounting for Variables: Ensuring the AI model can handle variations in lighting, track conditions, driver style, and normal repair allowances.
- Accuracy & Trust: Tuning the AI to be highly accurate, avoiding false positives and false negatives, to gain the trust of teams and officials.
- Integration: Seamlessly integrating multi-spectral scanning hardware into the demanding F1 garage and grid environment.
Accomplishments that we're proud of (Conceptual)
We are proud of the core innovative concepts behind this idea:
- The novel approach of modeling legal aging patterns with AI to more accurately detect illegal deviations.
- The three-stage timeline analysis (T0, T1, T2) providing a richer understanding of the car's state changes.
- The multi-spectral dimension, allowing detection beyond just visible changes (e.g., thermal inconsistencies).
- Designing a system that could significantly enhance the fairness and integrity of motorsport competitions.
What we learned
Developing this idea highlighted:
- The sophistication required to accurately distinguish between legitimate wear and intentional modifications.
- The immense potential for AI in understanding and predicting physical degradation processes.
- The importance of a robust baseline model (legal aging) for effective anomaly detection.
- How combining multiple sensor types (multi-spectral) can provide much richer insights than visual inspection alone.
What's next for The Multi-Spectral Parc Fermé Sentinel
- Secure Data Partnerships: Collaborate with motorsport teams or governing bodies to access historical data for training the "Legal Aging" model.
- Develop Sensor Rig: Design and build a prototype multi-spectral scanning system suitable for the Parc Fermé environment.
- Train Initial AI Model: Begin training the predictive model on available or simulated wear data.
- Build Core Engine: Implement the Plausibility Analysis Engine comparing T1/T2 scans against the model's predictions.
- Validate & Refine: Test the system rigorously against known legal and simulated illegal changes, iterating on the AI model and detection logic.
Built With
- cloud-platforms
- cloud-storage
- cnn
- google-cloud
- multi-spectral-cameras
- opencv
- pcl
- python
- pytorch
- tensorflow
- thermal
- transformers
- uv