S.U.N.O.D.
AI-adaptive traffic light system — real-time signal control from live IP camera feeds.
The Problem
Metro Manila intersections run on fixed-timer lights that ignore real-time conditions — MMDA's own semi-adaptive systems still fall back to pre-programmed schedules. Team capstone: build an A
The Approach
Rebuilt the mobile client from a basic map view into a navigation tool comparable to Google Maps/TomTom Orbis. Dark Mode base map (CartoDB Dark Matter) for traffic-color contrast, live GPS st
The Outcome
Working prototype, evaluated by 5 technical experts and 56 end users under ISO/IEC 25010 — 4.68/5 and 4.65/5 respectively ("Excellent" on all measured characteristics: functional suitability,
Critical Challenges
API quota vs. responsiveness
Live typeahead search hitting TomTom's Search API on every keystroke would burn quota fast; added a 300ms debounce so suggestions only fire once typing pauses.
POI name resolution
TomTom's reverse geocoding was returning raw street addresses instead of business names (e.g. "One Way Cafe" showed as a street number). Fixed POI recognition so businesses display their actu
Route disambiguation UI
Needed to show up to 3 alternative routes without visual clutter. Solved with hit-detection: active route rendered thick/colored, alternates thin/grey and tappable to swap, with ETA recalcula
Crash-safe data parsing
TomTom's traffic servers intermittently sent whole numbers where the client expected decimals, crashing the parser. Added explicit .toDouble() coercion as a defensive parsing layer rather tha