Using A.I. and Machine Learning
to Build a Better Routing App
A Research Project
Presented by

Christopher Moravec
Based in Portland, Oregon, USA.
presenter Dawn on a bike at pedalpalooza presenter Dawn pedaling backwards on a dual bike presenter Christopher's bike in snow
an xkcd comic of a cheap gps
Creating A Functional Navigation App for Cyclists
bike habit loop cartoon
Creating A Better Navigation App
for Flooded or a Private roads
bike habit loop cartoon bike habit loop cartoon
Aren't there enough navigation apps out there?
The Defaults: Google Maps & Waze
google maps logo waze logo
Esri's ArcGIS Navigator
esri navigator logo
Navigator
... but not for seasonal, private or "other" roads
closed road for snow

NPS

Why is this important?
Some Portland Facts:
  • 7.2% of commuters go by bike.
  • Highest percentage of bike commuters for a large US city
  • 17,000+ workers in Portland choose to cycle
  • 350+ miles of bikeways
  • -Portland Bureau of Transportation, 2016

Some National Facts:
  • #1 most distracting task: Programming navigation devices
  • 14% of all police-reported motor vehicle crashes were “distraction affected” in 2015
  • 4.57 million people were seriously injured in motor vehicle crashes in 2017
  • -Fast Company, 2018

Our Goals:
  • A navigation app specifically for cyclists
  • Help make city cycling easier and more inclusive
  • An app to reduce distractions
  • Leverage new or alternate technologies
Routing
The available Data is decent!
OpenStreetMap
openstreetmap logo
The power of OpenStreetMap is that it's open!
  1. Add roads, submit changes
  2. Less infrastructure required
  3. Everyone gets better data
What do we have so far?
Cycling - Roads Rated As
  • Calm
  • Intermediate
  • Fast
Basic Routing using OpenStreetMap data
Private Road Warnings

Flickr
Flooded Road Warnings
Voice Prompts and Interactivity
How do we leverage this data?
Private Road Warnings
no trespassing sign
Leveraging OSM private flags
**Note: We'd like to integrate phone numbers

Flickr
Flooded Road Warnings
massively flooded road
Combine low spots and intersections with waterways
and current weather to warn users of flooding potential
Voice Prompts and Interactivity
mounted cell phone
Keeping the app hands free and simple
App Demo!
What's Next?
Live(ish) integration with OSM
Submit Changes within the app
Live(ish) two-way integration with OSM
Sustainable Data Pipeline
AI/ML for preferred route selection
Thank you.
presenters dawn and christopher in front of dinosaur bones
christopher@dymaptic.com
@dymaptic