FleetLab

Coordinated forklifts. Shared factory floors. Interactive 3D simulation replays.

Live demo · Quick start

Overview

FleetLab simulates warehouse fleets sharing tasks, navigating traffic, and recharging. A Python simulator generates reproducible runs; the 3D web viewer lets you explore them.

• Factory floor — Four layouts with working forklifts, visible cargo, and three patrolling workers that planners avoid.
• Task routes — Inspect the selected forklift's current task: solid lines show travel so far, dashed lines show its recorded plan. Finished task markers disappear.
• Operations — Seeded tasks arrive progressively. Track the fleet, charging stations, and task queue with playback and timeline controls.

Routing algorithms

Algorithm	Approach
Coordinated A*	Occupancy-aware A* with per-step cell and edge reservations.
WHCA*	Windowed space-time planning with rotating vehicle priority.
RHCR + PBS	Rolling-horizon planning with conflict-driven priority search.

Switch between recorded runs using the algorithm selector. All three use the same task allocation and charging rules; windowed planners display their current planning horizon.

Quick start

Requires Node.js 22.13+, pnpm 11, and a WebGL-capable browser.

From the repository root:

cd web
pnpm install --frozen-lockfile
pnpm dev

Open the local URL printed by the dev server. Replay data is included; no Python process is needed to use the viewer.

Generate new replays with Python

From the repository root, with Python 3.11+:

python -m venv .venv
source .venv/bin/activate
python -m pip install -e .

python -m adaptive_agent_lab.reporting.fleet \
  --config configs/fleet-demo.json --algorithm coordinated-astar \
  --output web/public/fleet-demo.json

python -m adaptive_agent_lab.reporting.fleet \
  --config configs/fleet-demo.json --algorithm whca \
  --output web/public/fleet-whca.json

python -m adaptive_agent_lab.reporting.fleet \
  --config configs/fleet-demo.json --algorithm rhcr-pbs \
  --output web/public/fleet-rhcr-pbs.json

Set task count, random seed, and simulation horizon in configs/fleet-demo.json. Regenerate all three files with the same configuration to keep the runs comparable.

Reference

1. David Silver. Cooperative Pathfinding. AIIDE, 2005. — WHCA*.
2. Hang Ma et al. Searching with Consistent Prioritization for Multi-Agent Path Finding. AAAI, 2019. — Priority-Based Search (PBS).
3. Jiaoyang Li et al. Lifelong Multi-Agent Path Finding in Large-Scale Warehouses. AAAI, 2021. — Rolling-Horizon Collision Resolution (RHCR).