Skip to content

simulation-model.md

Latest commit

 

History

History
76 lines (51 loc) · 2.71 KB

simulation-model.md

File metadata and controls

76 lines (51 loc) · 2.71 KB

Simulation Model

Waremax uses a discrete-event simulation (DES) kernel. The goal is to scale to many robots without stepping time at millisecond resolution.

Time Model

  • Global clock advances to the next scheduled event.
  • Each event updates state and schedules downstream events.
  • A small motion tick is optional for models that require periodic updates, but the default is fully event-driven.

Event Types (Baseline)

  • Order arrival
  • Putaway or replenishment arrival
  • Task assignment
  • Robot departs node
  • Robot arrives node
  • Station service start
  • Station service end
  • Inventory decrement/increment
  • Charging start/end (if enabled)

Robot Motion

Robots traverse edges in the map graph.

  • Travel time = edge_length / robot_speed.
  • If an edge is at capacity, the robot waits at the current node.
  • Waiting may trigger rerouting based on a policy threshold.

Traffic Model

Each edge and node has a capacity. The Traffic Manager decides whether a robot may enter an edge at a given time.

Supported traffic policies (configured via traffic.policy in scenario config):

  • wait_at_node: Robot waits at current node until edge is available. Simplest policy with no rerouting.
  • reroute_on_wait: If waiting exceeds reroute_after_s threshold, recompute path to avoid congestion.
  • reservation: Reserve edges over a short time horizon (reservation_horizon_s). Prevents conflicts proactively.

Routing

Routing returns a path and estimated travel time. Configuration is via routing in scenario config.

Algorithm options (routing.algorithm):

  • dijkstra: Standard shortest-path. Lower overhead, suitable for most cases.
  • astar: A* with Euclidean heuristic. Faster for large sparse graphs.

Congestion awareness (routing.congestion_aware):

  • When enabled, edge costs include a congestion penalty based on current occupancy.
  • congestion_weight controls penalty strength (0.0 = ignore, 1.0 = heavy penalty).

Route caching (routing.cache_routes):

  • When enabled, caches routes between frequently used node pairs.
  • Cache is invalidated when map constraints change.

Stations

Stations are modeled as queues with service times.

  • Queue capacity can be finite or unbounded.
  • Service time can be modeled as base + per_item or as a distribution.
  • Concurrency defines the number of parallel service slots.

Inventory

Inventory is represented as counts in bins.

  • Picks decrement quantities at the source bin.
  • Putaway increments quantities at the destination bin.
  • If a pick is not available, the task can backorder or trigger replenishment based on policy.

Determinism

  • Random processes must be seeded from the scenario config.
  • Identical inputs and seeds should produce identical outputs.