Smart Curbs

Measuring urban street activity through the eyes of a public transit network anchored this project with the MIT Senseable City Lab. The work focused on embedded systems, computer vision, and urban data collection, outfitting buses in Paris with edge-computing sensors to classify pedestrian and traffic patterns. Because of strict GDPR regulations, the system had to process all imagery locally on the device, ensuring only anonymized metadata ever left the vehicle.

The hardware architecture relied on Raspberry Pi microcomputers running compressed neural networks for real-time image classification. I developed a fault-tolerant microservice pipeline to handle the entire data lifecycle, from edge classification to local storage and network synchronization. Since the sensors operated unattended, I integrated power management protocols that automatically woke the devices when the bus started and put them to sleep during inactivity. I also implemented remote SSH access to monitor the fleet and push software updates over the air.

Prototyping the system required strapping the sensor to a bicycle and riding through Paris to simulate bus routes, which exposed critical edge cases around Wi-Fi handoffs and battery quirks. During integration with the RATP public transport authority, a rigid municipal API rejected negative coordinate values, causing failures for specific geographic areas. Rather than waiting for an upstream fix, I built a coordinate-transformation workaround directly into the microservice pipeline, ensuring the system could reliably submit metadata regardless of the vehicle's location.

The finalized sensors were successfully deployed on public buses, running fully autonomous edge classification over a five-week period. To make the aggregated data actionable, I built a web dashboard that visualized urban activity patterns, traffic density, and pedestrian flow in near real-time. This framework provided researchers with a scalable method for mapping street capacity and infrastructural needs, demonstrating how public transit fleets can double as mobile environmental monitoring stations.