London Underground

A live tube graph — eleven lines, hundreds of trains, platforms held as a mutex.

Async Tick Spatial FSM

Try it live→ Download DuckDB

Service replay — a full tube day, line by line FULLSCREEN ⤢

Clock tick · 1 min/turn

Update Asynchronous

Lines 11

Topology Spatial graph

Coordination Platform mutex

OVERVIEW

The London Underground as a running simulation. Stations form a directed graph; trains move along it through a lifecycle of depot, dwelling, in-transit and terminus turnaround. Platform occupancy is enforced as a mutex — a departing train releases its current platform and claims the next one atomically, so trains can never overtake or double up. Each turn is one minute of wall-clock time, and passenger volumes shift with each station's demand level.

A large, spatial, asynchronous world that stresses routing, scheduling and contention reasoning. The platform mutex creates real knock-on delay: a held train backs up the ones behind it, exactly as signal congestion does on the real network. Every train and station resolves through foreign keys, so the whole timetable is queryable after the fact.

TRAITS

Async

Independent, event-driven timelines

Tick

Discrete fixed-step time

Spatial

Entities have position & topology

FSM

Entities are finite-state machines

SCHEMA

Linked tables with guaranteed referential integrity.

TABLECOLUMNSDESCRIPTION

station ID, station_name, line, demand_level, position, terminus_out, terminus_in, platform_out_occupied, platform_in_occupied, board_typical, board_rush, alight_typical, alight_rush, trains_served, next_out_id, next_in_id, current_state One row per station-platform: its line, position, demand band, the platform-occupancy locks and the next-station references in each direction.

train ID, line, direction, passengers, stops_made, boarded_total, alighted_total, trips_completed, held_at_signal, current_station_id, current_state One row per train: line, direction, occupancy, cumulative boardings/alightings, turns held at a signal and its current station.

LIVE API

Generated REST endpoints. Also exposed as MCP tools.

POST /scenarios/london-underground/experiments Seed a new world (an experiment)

POST /scenarios/london-underground/experiments/{eid}/run Advance N turns, or request an event

GET /scenarios/london-underground/experiments/{eid}/entities/train Query trains in service

GET /scenarios/london-underground/experiments/{eid}/events Append-only event log

GET /scenarios/london-underground/experiments/{eid}/dataset Download the exported dataset

SEMANTIC LAYER

OSI-compatible definition, emitted with the dataset.

# london-underground.osi.yaml — emitted automatically
semantic_model:
  name: "london-underground"
  source: "duckdb://london-underground.db"
  entities:
    - name: station
      primary_key: id
  dimensions:
    - name: state
      type: categorical
    - name: t
      type: time
  measures:
    - name: row_count
      agg: count
    - name: active
      agg: sum
      filter: "state = 'ACTIVE'"