Each delivery is scored using f(n) = g(n) + h(n) where g = actual travel cost with traffic weights and h = heuristic distance to destination. This gives O(n log n) performance vs greedy O(n²).

Nearby orders are grouped using greedy nearest-neighbor clustering before vehicle assignment — reducing total distance traveled by batching pickups in the same zone.

Every 2 seconds, new orders trigger a re-prioritization pass. Vehicles are reassigned in real-time using a priority queue sorted by urgency × distance × capacity.