Design a Green Score for Routes

Problem Statement: A Green Score for Routes

A green score for routes estimates the carbon emissions (or energy) of driving a given route in a given vehicle, and uses that estimate to recommend the greener of several ways to get somewhere. When a driver asks for directions from A to B, a routing engine returns a few candidate routes; this system scores each one's emissions and surfaces the lowest-carbon option — Google Maps eco-routing does exactly this, nudging you to the route that burns less fuel even if it is a minute slower. The same engine rolls trips up into fleet emissions reports and a per-driver eco-score.

Crucially, this is not a routing problem. We do not compute turn-by-turn directions — a Maps/routing engine does that and hands us candidate routes. Our job is the emissions model and the scoring, and that splits into three problems.

First, the emissions model: estimate the CO2/energy for a route. A route is a sequence of road segments, each with features (length, grade/slope, speed profile from traffic, surface). A vehicle profile (powertrain — gas, hybrid, or electric — mass, aerodynamics, efficiency) determines how much energy each segment costs. Route emission = the sum over segments. This must be deterministic and auditable — the same route and vehicle must always produce the same number, because that number feeds carbon accounting and offsets.

Second, the green-vs-fast scoring: the fastest route is not always the greenest. We score the trade-off (CO2 vs ETA) and recommend the greener route only when its time penalty is small enough — a tunable threshold.

Third, fleet reporting and offsets: aggregate trip emissions into an immutable ledger for sustainability reporting and carbon-offset purchasing, plus a gamified eco-score that rewards smooth, efficient driving.

The scale is a read-path enrichment on a routing platform: potentially ~100,000 route-scoring requests/second at peak (each route ~tens to hundreds of segments), plus ~10M fleet trips/day flowing into the emissions ledger. The hard parts are a deterministic, calibrated emissions model, a fair green-vs-fast scoring rule, and an auditable carbon ledger — an estimation-and-accounting problem wearing a maps costume.