≪ Back to Projects

A.I. for Cycling Navigation

Portland is an awesome city and one way they prove that is by making their bike infrastructure data available to the public. We combined the Portland bike data, OpenStreetMaps, and hours of manual labor to create something amazing.

Our goal with this project is to inspire members of the Portland community (and beyond) to bike more often while feeling empowered. Our data begins focused on Portland, OR as that is where we (the dymaptic team) work and cycle. We’ve seen the city grow over the past several years and would like to offer safer routing options that learn with the cyclist and are always ready to make improvements.

bike handle bars in greyscale

This project is currently separated into two parts: the A.I. interface and the navigation. Our eventual goal is to create a voice-based interface that will update a chosen route via voice command while cyclists are on the go.

Our current focus, is on the navigation piece. With the help of good friends, we were able to transform and improve the OpenStreetMaps data and the City of Portland bike data. The results turned out better than expected. While cleaning the data, we discovered weird bugs where bikes were being routed into heavy traffic because the routing mechanism (or really the data) was thinking like a car and not like a bike. In our cycling navigation app, we intended to remove these bugs.

We have separated the roads into three levels or types, consisting of seven total categories.

  • Calm
  • Intermediate
  • Fast

Our Calm routes give highest prefernce towards off street paths, trails, and greenways. It provides some protected lanes, with shared lanes and neighborhood streets offered when no others are available.

Intermediate routes give preference to protected lanes, off street trails, and greenways first. Shared lanes, designated bike lanes, and neighborhood streets appear second.

Our final designation is Fast, prioritizing shared lanes, designated bike lanes, neighborhood streets, and protected bike lanes that provide the shortest distance between the starting location and the destination.

Depending on the rider’s level of comfort, more or less intense routes are suggested.

What’s next? We have a lot more work to do on the data, including riding many of Portland’s streets. We hope to bring in elevation data next and allow users to choose routing through intense uphill sections or avoiding hills all together.

Please view out A.I. Cycling Navigation: The Data and Beyond for a more visual account of this project.

If you find this interesting, we’d love your help beta testing the cycling navigation app.

≪ Back to Projects