WatTravl

Campus buildings are the place phone navigation stops working: GPS cannot tell which floor you are on, and the official maps end at the front door. Getting from one room to another in a building you do not know means asking someone. This routes it properly.

There was no data to start from, so the graph is hand-built. Each floor is a set of hallway nodes keyed by room number, with edge weights for walking distance, and the floors are stitched together at the points where you can actually change level. Dijkstra then runs over the whole building at once rather than per floor, so a route that crosses three floors is one search instead of three stitched results.

Staircases and elevators are modelled as separate transition sets covering the same floor changes. That started as a convenience and turned into the feature I am most glad we built: routing by elevator only is what makes the app usable for someone who cannot take stairs, and it fell out of the graph design almost for free.

On the phone the map screen showed one floor at a time: an amber bar with a back arrow and two pickers for building and floor, a Refresh button, and the floor plan filling the rest of the screen with pan and zoom. The route drew over it as blue arrowed lines, a red marker at your start and a cyan one at your destination, with green markers where the path carried on to another floor and a toast reading "Head to Floor 3".

The demo above reconstructs that screen faithfully, with one deliberate omission: it runs over a synthetic building rather than the real plans. Publishing a complete interior map of two university buildings — every corridor, stairwell and room — is not something a portfolio piece needs to do, and the floor plans are the University's material rather than mine. The chrome, the route rendering and the pathfinding are the parts worth showing, and none of them depend on the real geometry.