Inspiration
Alex had just started her internship in a new city. With only a 30-minute lunch break, she wanted to go for a walk but did not know the area. She did not know where the parks were, which streets were pleasant to walk on, or how to plan a route that would fit her limited time.
That experience highlighted a simple but common problem. Finding a good walk in an unfamiliar place is surprisingly hard. Most mapping apps optimize for speed or distance, not for how a walk feels. We wanted to change that.
That is what inspired Wandr.
What it does
Wandr generates customized walking routes from natural language descriptions.
Users describe the kind of walk they want, such as how long they have or how far they want to go, and Wandr generates a looped walking route starting and ending at their location. The route is displayed directly on a map, making it easy to explore a new area without planning ahead.
The goal is to let users focus on walking, not on planning.
How we built it
Wandr is built as a web app with a lightweight backend and an interactive map-based frontend.
At a high level:
- The frontend collects the user’s location and prompt
- The backend determines a suitable route length, extracted from the prompt
- A circular walking route is generated around the user’s location
- The route is snapped to real walkable paths using OpenStreetMap-based routing
- The final route is returned and rendered on the map
To generate looped routes, we compute a set of waypoints arranged roughly in a circle around the user’s starting point. These include north, east, south, west, and a return to the start. These waypoints are passed to OSRM (Open Source Routing Machine) using the walking profile, which produces a realistic path that follows actual roads and footpaths rather than straight lines.
We also implemented robust fallback behavior so that if routing fails, the app still returns a usable route instead of breaking.
Challenges we ran into
The biggest challenge was route generation.
Mapping libraries are excellent at finding the shortest path between two points, but they are not designed to generate pleasant circular walks of a specific length. Getting a loop that starts and ends at the same point, matches a target distance, and uses walkable paths required experimentation and iteration.
Accomplishments that we're proud of
- Building a full end-to-end system from user prompt to route visualization
- Successfully generating looped walking routes instead of simple point-to-point navigation
- Creating a resilient backend that gracefully handles routing failures
What we learned
- Routing problems are more complex than simple navigation
- OpenStreetMap and OSRM are powerful tools but require careful handling
- Even lightweight natural language processing can significantly improve user experience
- Building a strong MVP is about making good tradeoffs under constraints
What's next for Wandr
Next, we plan to deepen Wandr’s personalization by improving both the natural language processing and routing logic. Planned features include:
- Customization for elevation gain
- Preferences for trails versus main roads
- Scenery and greenery optimization
- Avoiding busy streets
- Multiple route options ranked by how well they match user preferences
- Exporting generated routes to Google Maps so users can follow turn-by-turn walking directions
Log in or sign up for Devpost to join the conversation.