BetterRoutes
A bike map that prioritize saftey and personal preference over traditional google map's fastest route
Demo video
What it does
BetterRoutes is a San Francisco bike map that plans the safest route, not just the fastest. It runs its own router on 211,221 street segments scored with the city's real data: 3,558 bike-injury crashes, SFMTA's survey of every bike lane, every highway and traffic signal, and the elevation of every street. It offers Fastest, Safest and Safest + bike lanes options, cutting time in dangerous areas by 70% across 10 test trips, and tells riders exactly what each route avoided and what that cost in time. It also makes the ride practical and enjoyable: it can avoid steep hills and red lights, build a custom route from a one-sentence request to its AI assistant, show photos of places along the way, preview the ride in 3D, and guide riders turn by turn.
Inspiration
I am a cyclist myself, I bike with my friends very often. The main Issue that completely inspired for me to create this project in the first place is due to the fact that google maps is very unreliable in terms bike lanes. I've had experience where google maps took me to a very dangerous, full of drug users and homeless in san Francisco, along with a portion of the ride that me and my friends had to bike in the car lane for 15 minutes. Those are all dangers associated with biking routes from google maps, which is why I trained my model to combat this, now my model will avoid dangerous areas and prioritize protective bike lanes that stretches for 140 miles over the unprotected or no lane that stretches 300 miles.
How we built it
Used Google maps for the main map, trained with many sources such as openmaps. AI feature with OpenAI API key, 3js for 3d tours.
Challenges we ran into
Sometimes the map malfunctions and have many bugs such as going into dangerous areas despite me choosing "safe route", those bugs are all fixed now.
What's next
Learn form users experience and add more features, expand to more cities.
Screenshots
0 comments
Sign in to upvote this project or leave the team a comment.