FleetIO

A browser visual simulation of ten orange warehouse AGVs. React, Three.js and a deterministic Web Worker render and coordinate a 16 × 46 metre warehouse. This is a demonstration of MAPF-style coordination; it does not claim optimal routing or implement a robotics backend.

Open the hosted prototype.

Run locally

node --version # Node 22.12 or newer
npm ci
npm run dev

Original downloads in Assets/ stay untouched. Prepared assets in public/models/ are committed for deployment. See public/models/ATTRIBUTION.md for sources and material changes. If the original download folders are present, node scripts/prepare-assets.mjs rebuilds runtime copies; cloning this repository does not require that step.

npm run build
npm run lint
npm test

Demonstrate

Open Live, select a scenario and press Run. Guided demonstration advances only when required simulation outcomes occur. Individual scenarios can be restarted independently. Pause preserves the run, Stop preserves the final view, Play after Stop starts afresh, and Reset restores the selected fixture. Speed changes simulated time.

Select a robot in the scene or roster to inspect its task, battery, payload and route. Use Overview, Top and Follow camera controls; drag to orbit and scroll to zoom. Map expands the warehouse. Search locates robots, tasks, stations and JECs. Analytics uses measured simulation records.

For a short walkthrough, start Junction conflict, then Robot failure and Payload recovery from Settings. Normal logistics runs for about 12.4 simulated minutes; Guided demonstration takes about 20.1 simulated minutes. At 2×, real time is approximately half those durations on a responsive browser.

Payload recovery waits for an explicit Recover robot action. A failed robot retains custody until recovered. Manual controls also expose robot failure, JEC outage/restoration, passage blockage/clearing and degraded communications.

Docks and presentation motion

Each of the ten robots has a physical home dock at its staging bay. Click a dock to locate its bay and inspect its occupant; Map includes a directory of all ten home docks and both chargers. The inspector also provides Locate Home Dock. Status comes from the robot's footprint and state: Ready, Occupied, Charging or Fault. Home docks show parking; only Charge 1 and Charge 2 replenish simulated battery. Hardware remains outside robot turning footprints and navigation paths.

Controls use brief hover, press and focus transitions. Panels, drawers and new events have restrained entrances; battery/progress bars and utilization update smoothly from measured values. Settings → Reduce Motion disables these effects and makes camera changes immediate. The preference survives reloads, and the operating system's reduced-motion preference takes precedence. Charging lights pulse only while an actual robot is charging and the simulation is running.

Search supports Arrow Up/Down and Enter, Escape dismisses results, and Ctrl/Cmd+K focuses search. Dragging the warehouse preserves selection; clicking a robot or dock selects it.

Architecture

• src/types.ts defines common snapshots and commands.
• src/simulation/layout.ts defines navigation and rack/station footprints.
• src/simulation/core.ts owns robot motion, tasks, arbitration and scenario outcomes.
• src/simulation/worker.ts provides the fixed-step clock and snapshots.
• src/simulation/engine.ts maintains one worker and rejects obsolete generations/sequences.
• src/components/WarehouseCanvas.tsx interpolates received poses without advancing simulation logic.

Reset clears bounded event/metric histories and increases snapshot generation. The worker uses 60 Hz simulation steps and 20 Hz snapshots. Reservations gate entry, occupied resources retain ownership, and peer arbitration uses the same policy during a JEC outage.

Verification and limits

See GATES_3D.md for acceptance and recorded release evidence. Scenario success must follow actual tasks, motion and coordination. Browser performance depends on the GPU and resolution; the verification report records measured results rather than treating a successful build as visual proof.

Production release

See production setup, existing-project deployment, demo guide, limitations, and release notes.

Run npm run verify:build after building and npm run package:release to generate the static ZIP and SHA-256 manifest in release/. The Git source remains the input for the existing Vercel project.