Road memory: a pothole map verified across multiple trips

The problem

The municipality only learns about a pothole from a complaint or an accident, and maintenance follows complaints rather than danger. Automatic detection attempts stumble at the same point: a single vibration reading can't tell a pothole from hard braking, a regular speed bump, or driving over a stone. The result is a map full of false alarms that nobody trusts, so the municipality goes back to complaints.

How it works

1. SensingThe accelerometer reads the vehicle's vibration at a high rate, along with location and speed.
2. FilteringExcluding vibrations caused by braking, turning, and speed changes.
3. The signatureEvery event is saved with its intensity, time profile, and passing speed.
4. AssemblyNearby events from different trips are grouped into a single point.
5. Confidence scoreThe confidence score is calculated from the number of confirmations versus passes without detection.
6. Priority rankingIntensity, confidence, rate of deterioration, and traffic density set the maintenance priorities.

Project features

• Confirmation by repetition: a point detected on seven independent trips is real, while a one-off detection may be incidental.
• A confidence score for each point that rises with every confirming pass and falls with every contradicting one.
• Tracking deterioration: a pothole getting worse month after month gets higher priority than a stable one.
• It tells a regular speed bump from a pothole, since a bump's signature is symmetrical and repeats for all vehicles at the same intensity.
• It detects a pothole's disappearance automatically: repeated passes without detection mean it was repaired.

Expected challenges

These are the main obstacles to expect during the build, laid out clearly so you know them before you start.

• Telling a pothole from braking and speed bumps is the heart of the project; without it the map is worthless.
• Speed affects the signature: the same pothole causes different vibration at 40 than at 80.
• Suspension differs from vehicle to vehicle, so readings must be normalized per vehicle.
• Positioning accuracy is several meters while potholes are smaller, so grouping needs a carefully chosen radius.
• Proving the idea needs several vehicles and several trips; one trip isn't enough to prove anything.

The consultation gives you a solution for each one.