Skip to content

Tutorials

Start with the existing getting-started tutorial, then use the live pathfinder to change endpoints and obstacles while the same core query runs in WebAssembly. Together they are the basic pathfinding learning path; this page does not introduce a competing beginner sequence.

Flow field steering

The flow field steering tutorial builds one dense distance product for a shared goal, then moves independent agents by reading public distance labels rather than retained directions. Its compiled native model and embedded WebAssembly view demonstrate deterministic next-step selection, goal rebuilds, and explicit at-goal and unreachable states.

Use this path after basic pathfinding when many independent agents share a destination and retaining complete paths would duplicate guidance.

Colony simulation

The colony composition tutorial follows one fixed tick from queued wall edits through topology, bounded planning, movement, and DeltaFrame presentation. Its article-sized live view uses the same model as the full standalone demo.

The wider colony family shows how focused agent movement grows into that complete fixed-step composition:

  1. path_agents.cc isolates goal assignment, bounded planning, movement, and blocked-path handling.
  2. colony_2d.cc adds queued construction, dirty-driven topology, scheduling, and DeltaFrame presentation.
  3. The web_colony host compiles the same model for the live colony while keeping interpolation in the browser.
  4. The native model check exercises the composition without a renderer.

Use this family when you want to understand the boundary between simulation state, retained paths, and presentation state.

Colony diagnostics

The colony diagnostics tutorial compiles that same model in a separately gated host. It captures real path, queued-phase, timing, allocation, and lifecycle-flow snapshots on the recording thread, then hands read-only values to Dear ImGui and accessible DOM status.

Use it after the colony composition tutorial when you need to distinguish instrumentation counters from lifecycle conservation and from pathfinding flow fields.

Procedural sparse streaming

The procedural sparse-stream tutorial holds a large bounded world to 32 resident pages. Its shared native and Wasm model demonstrates deterministic chunk generation, an ordered 5×5 camera window update, LRU eviction, and agents that retry indeterminate routes.

Use it after basic pathfinding and the residency guide when a complete dense world would exceed the available memory budget.

Pathfinding strategies

The strategy family connects API call shape to measured workload shape:

  1. pathfinding_strategies_model.cc runs plain A*, exact route caching, weighted batches, and shared-goal distance fields on one map.
  2. The interactive comparison exposes the paths and reuse counters without treating browser timing as a benchmark.
  3. The strategy comparison article explains the decision boundary and links to the benchmark evidence.

Use the pathfinding decision guide for a compact branch-by-branch choice after working through the comparison.

Congestion pricing

The congestion family keeps the supported recipe, evidence, and experimental lab visibly separate:

  1. congestion_pricing.cc is the focused public-API recipe for a separate surcharge field and affected-route replanning.
  2. The web_congestion model layers screened policies over the colony model and retains a native self-check.
  3. The congestion pricing guide states which results are supported, rejected, or experimental.
  4. Congestion Lab makes those variants explorable.

Congestion pricing changes route cost. Reservations, collision avoidance, and local steering remain separate coordination concerns.

Advanced labs and integrations

These experiences are useful next steps, but do not need dedicated tutorials yet:

  • Congestion Lab compares pricing policies.
  • Traffic Lab shows a large deterministic crowd workload.
  • Tower explores one routed three-dimensional world.
  • WebGPU demonstrates the optional GPU transport boundary, not GPU pathfinding.

The examples catalog groups every compiled model, lab, recipe, and optional integration by how it is best used.