How it works

System overview

1. City digital twin: We stream TomTom Route Monitoring telemetry for key corridors (Audi Forum to Hauptbahnhof, Klinikum to Nordbahnhof) and fuse it with municipal charging assets.
2. Demand inference: Corridor delay, travel time variance, and land-use priors feed a demand surface rendered as 3D extrusions in deck.gl.
3. Fleet agent: A constrained heuristic assigns AV blocks to hotspots, guaranteeing minimum coverage and reserving capacity for first/last-mile connections.
4. Energy co-optimisation: Charging windows align with time-of-use pricing (avoiding peak demand hours); the model can recommend V2G buffer opportunities when integrated with utility grid signals.

Reward design

The orchestrator balances served demand, delay reduction, energy cost, and neighbourhood equity. KPIs remain non-negative: we report reliability (0 to 100), energy per ride (kWh), grid stress (0 to 100), and a composite reward score. Human operators can trace each recommendation back to the data that triggered it.

From demo to deployment

• Extend corridor coverage with additional TomTom monitor IDs, micromobility feeds, and city sensors.
• Plug in grid signals (day-ahead prices, CO₂ intensity, feeder limits) for more granular charging optimisation.
• Layer in civic constraints: pedestrian zones, school buffers, accessible vehicle targets, and policy-defined service floors.
• Run city-in-the-loop pilots starting with Audi’s factory shuttle, then scale across public transport interchanges.