How to play Choose your journey, roads and traffic. We run two rush hours together. Round length grows to fit your Google journey. Follow the green route: one is today’s traffic; the other replaces your chosen share of background car trips with pods. Your own ride stays the same in “Change the city”. In “Compare my ride”, only your vehicle changes; background traffic stays identical.
Google’s no-traffic estimate excludes traffic conditions; it is not a guarantee of an empty road. The live or forecast traffic estimate is shown separately. Game time starts at your simulated departure, including waiting to enter. Until you arrive, the cards show elapsed time, not an ETA.
Win by moving more people with shorter trips and fewer stops. Your city score belongs to the pod scenario. Each arriving passenger earns up to 100 points, adjusted for journey time and stops. Personal bests compare the exact same setup on this device.
What this model measures Inside detailed neighborhoods, each vehicle chooses gaps, brakes and turns independently. Local bodies and mirrors have collision shapes with 3.81 cm minimum clearance. Elsewhere, finite-storage city links use travel clocks and FIFO service queues; they do not simulate mirror clearance. Signals alternate approaches; vehicles wait for a green and a clear exit. A conservative single-vehicle junction reservation can exaggerate intersection queues.
“At red lights” counts time your vehicle is stopped directly at a red signal. “In queues” includes entry waiting and other stopped time, including queues behind signals. Slower moving time from turns, potholes and traffic also adds to the gap from the empty-road reference.
Road space is all on-road vehicle body and mirror area divided by passengers moving faster than 0.3 m/s. It includes vehicles on city links and in local streets, and excludes headway, pedestrian area and entry queues. When nobody moves, it is unavailable. Background bus occupancy is configurable. Your bus persona represents only your chosen party; no scheduled stops, boarding wait or other passengers are added to that persona trip.
Pod 1’s body is 2.635 × 1.3 m with two seats. Mirrors, turning radius and speed are assumptions. Car trips with 3–5 people split into enough pods; people and destinations are preserved. A smaller footprint can help, but the result may be unchanged or worse. This is not evidence that Pod 1 solves city congestion.
Detailed agents are capped at 160 across all neighborhoods per scenario. Excess demand stays on upstream city links. A run that hits the detail budget is flagged: that extra delay is computational, not evidence of road capacity. Old neighborhoods drain after the rider moves; agents are never deleted to recenter the view. One ownership ledger audits future, entry, city-link, detailed and arrived passengers. Exits reserve downstream storage, so full links send queues backward. Road widths, link storage, service rates, demand and behaviour are illustrative. All signals share a clock. The city view follows actual Google road corridors. The junction lab, signal timing and road conditions are illustrative, not surveyed street geometry.
Download run data ↓ Real maps & a whole city The Google Maps view shows the real basemap and Google route polylines. Distance and traffic-aware travel clocks come from Google Routes API at the selected departure time and stay fixed for the round. Motorcycles and scooters use two-wheeler routes. Pod 1, autos and the bus persona use driving routes as proxies; this is not a bus timetable or a validated Pod 1 ETA. Road widths, potholes, divider polygons and signal timing need an independent road model or manual setup. Google’s traffic estimates and our simulated times must stay separately labelled.
This game runs paired agent simulations in a Web Worker. This pilot simulates detailed neighborhoods on your rider’s route, with aggregated road queues elsewhere. Fourteen corridors in both directions connect ten landmarks, with a direct Google route for your rider. Google’s traffic delay is a shared baseline. Synthetic road damage, signals and queues add more delay and may double-count real congestion. The comparison illustrates mechanisms; its savings are not a validated prediction for Bengaluru. A calibrated city-scale version can run a SUMO simulation on a server and stream the visible area. Millions of detailed vehicles are not a realistic browser target.
Google map overlays ↗ · Traffic-aware routes ↗ · SUMO mesoscopic model ↗