RoadSense AI

logo

The inspiration for RoadSense came from observing how reactive road maintenance currently is. Potholes and infrastructure damage are often only fixed after vehicles are damaged or accidents occur. We realized that public transport vehicles, like buses and garbage trucks, cover the entire city grid every single day. We wanted to leverage these existing fleets to create a solution that "sees" road damage before it becomes a danger, avoiding the high costs of dedicated survey vehicles.

RoadSense is designed as an IoT device that can be retrofitted onto heavy vehicles to continuously monitor road quality. The concept relies on "Sensor Fusion"-combining physical sensation with computer vision. The device will use a vibration sensor to establish a baseline of "smoothness" and detect abnormal jolts. When a significant impact is detected, it will trigger a camera to capture the road surface immediately. The onboard computer will then filter out false positives (like speed bumps) and validate the image using an AI Vision API. If a hazard is confirmed, the system will automatically generate a report for authorities with the exact GPS location and severity level.

To bring this to life, we intend to build the prototype using a Raspberry Pi 5 as the edge computing unit, chosen for its ability to handle local data processing efficiently. It will be connected to a standard camera module and a motion sensor. Our technical approach focuses on efficiency: rather than streaming video 24/7, which is costly and data-heavy, our script will only activate the vision system when the vibration sensor detects a specific threshold of impact. This "wake-on-shock" logic is what makes the device scalable for large fleets.

We anticipate that our main challenge will be "vibration noise." Since heavy vehicles naturally vibrate, we will need to carefully calibrate the sensor sensitivity to distinguish between a pothole (a sharp jolt) and a normal road feature like a speed bump. We also plan to address network limitations by ensuring the device processes as much as possible offline, only uploading the small, validated incident reports to the cloud.

We are proud of designing a system that transforms passive vehicles into active smart city agents. By using an "Edge-to-Cloud" architecture, we have outlined a solution that is low-cost enough to be deployed on hundreds of buses, yet smart enough to filter data autonomously. Our goal is to prove that you don't need expensive, specialized equipment to maintain modern infrastructure—just smart sensors and existing fleets.