Why we did this
Like any city, Boston is constantly working to maintain, repair, and upgrade its transportation infrastructure. Contractors and Public Works crews are usually the ones who find potholes and other problems in our roadways. However, that works takes up a lot of time and doesn’t always find issues right away.
At the same time, everybody has a smartphone. This has created the possibility for an army of volunteer pothole-spotters, similar to what we did with BOS:311. But, what if City workers and residents could help identify problems without having to lift a finger?
We developed Street Bump through a series of experiments:
- In 2011, in collaboration with Fabio Carrera and the Red Fish Group, we built a alpha version of the app. This confirmed for us that smartphones were sensitive enough to sense road problems.
- Later that year, we worked with Innocentive on a public good challenge to refine the algorithm used to identify road issues.
- Based on the results of the competition, we built the current version of the app in 2012. We worked with Public Works, IDEO, and the software company Connected Bits.
This version of the app uses the accelerometers and GPS of a smartphone. The app finds potential issues and their locations as a user drives. The data is then collected by the City in a map. The map includes both urgent fixes and areas ripe for investment.
Once a problem is identified in Street Bump, we can add it to the City’s system to schedule a repair. We’re currently working with Boston University researchers to enhance the ability of the app to identify bumps.
Results and lessons learned
We learned about what actually makes Boston’s roads bumpy. Residents most frequently reported problems about potholes, but the biggest cause of bumps is sunk manhole covers. This problem was found four times as much as potholes. Working with utility companies, the City fixed 1,250 of the worst manhole covers. This finding spurred our participation in the Living Labs Global competition.
Our initial analysis found that the app successfully detects many problems while keeping false positives under 10%. We’re working with machine intelligence experts at Boston University to improve the app’s ability to isolate actual defects.