Production-grade warehouse picking simulator and route optimizer. 300 × 200 m floor plan, 4 thermal compartments, 24 aisles with U-shape picking, real French SKU catalog, anti-cold doors with time penalties, 4 algorithms compared in real time, multi-operator wave picking, separate forklift replenishment system, 8 stretch wrappers, 12 analysis tabs.
- Overview — order summary, distance + zone donut.
- Warehouse Map — Folium dark map, warehouse-only (no world background), zone tints, doors highlighted, route arrows, mini-map. Toggle for forklift overlay.
- Step-by-step — color-coded picking sequence with cart fill, leg distance, door delays.
- Economic gains — annual projection at 60 orders/day × 250 days.
- Zone analysis — SKUs / distance / time per thermal zone.
- Route comparison — 4-way side-by-side mini-maps + cumulative curve.
- 3D view — picker route at floor level, forklift route at rack height.
- Animation — Play/Pause walker with live cart weight + volume.
- Forklift system — replenishment dispatch, OR-Tools per-forklift TSP, FROZEN > FRESH > AMBIENT > HEAVY priority.
- Filmeuses — 8 stretch wrappers, queue assignment, utilization gauges, breakdown simulation.
- Heatmap — pick-frequency heatmap aggregated from history.
- History — last 50 orders, per-operator performance, cumulative savings, optimizer adoption rate.
300 m wide →
┌─────────────────────────────────────────────────────┐ ▲
│ FROZEN -18°C 100×60 5 aisles F1..F5 │ │
200m │ ───❄──[anti-cold door +12s]───────── │ │
│ FRESH 0..4°C 100×80 5 aisles R1..R5 │ │
│ ───❄──[anti-cold door +8s]────────── │ │ AMBIENT 200×200
│ HEAVY ambient 100×60 4 aisles H1..H4 │ │ 10 aisles AMB1..AMB10
│ + DOCK-H │ │
└─────────────────────────────────────────────────────┘ ▼
3 docks → →
↑ 1100+ slots total
↑ U-shape per aisle (odd L / even R)
- Each aisle has odd positions on the LEFT side, even positions on the RIGHT side. Pickers walk DOWN the odd column then UP the even column — never doubling back inside an aisle.
- One cross-aisle per zone for skipping ahead.
- Two anti-cold doors at zone boundaries with measured delays (8 s and 12 s per pass).
- 4 dispatch docks (3 on the right edge, 1 in the heavy zone).
- 8 stretch wrappers (filmeuses), 2 per zone.
45 real products from 4 categories:
| Zone | Examples |
|---|---|
| FROZEN | Picard Frites, Marie Lasagnes, Ben & Jerry's Cookie Dough, Findus Épinards… |
| FRESH | Danone Activia, Président Beurre, Yoplait Nature, Lactel Lait, Chavroux… |
| AMBIENT | Panzani Farfalle, Barilla Spaghetti, Nutella, Coca-Cola, Heinz Ketchup, Ricard… |
| HEAVY | Evian 6×1.5L, Ariel 3.3L, Lesieur Olive 5L, Lotus PQ x12, Heineken 24×25cl… |
Each SKU carries brand, name, weight (kg), volume (L), fragility flag and priority.
| # | Algorithm | What it does |
|---|---|---|
| 1 | Naive | Walks the order in input sequence (baseline) |
| 2 | Nearest Neighbor | Greedy, ~80 % optimal in milliseconds |
| 3 | OR-Tools TSP | GUIDED_LOCAL_SEARCH, 10 s budget, distance-optimal |
| 4 | OR-Tools + constraints | Same engine, but enforces heavy → ambient → fresh → frozen LAST ordering, U-shape per aisle, and a time matrix that includes door delays |
The constrained algorithm pays a small distance overhead vs raw TSP but typically cuts cold-chain exposure by 60-80 %.
| Layer | Tool |
|---|---|
| Web UI | Streamlit (12-tab layout) |
| 2D map | Folium + streamlit-folium (CRS=Simple, no world tiles) |
| 3D / animation / charts | Plotly |
| Graph model | NetworkX |
| Optimization | Google OR-Tools (GUIDED_LOCAL_SEARCH) |
| Data | pandas, numpy |
| Persistence | JSON file (last 50 orders) |
| Exports | reportlab (PDF + QR), openpyxl, qrcode |
https://picking-route-optimizer.streamlit.app/
git clone <this-repo>
cd picking-route-optimizer
python -m venv .venv
# Windows: .venv\Scripts\activate
# macOS/Linux: source .venv/bin/activate
pip install -r requirements.txt
streamlit run app.pyOpens at http://localhost:8501 and auto-loads a 25-SKU sample order.
picking-route-optimizer/
├── app.py # 12-tab Streamlit UI
├── requirements.txt
├── README.md
├── core/
│ ├── warehouse.py # SKU, Order, Warehouse domain
│ ├── catalog.py # (data/catalog.py) French product catalog
│ ├── doors.py # Anti-cold door detection + time penalties
│ ├── graph.py # Manhattan distance + time matrix
│ ├── optimizer.py # 4 algorithms (incl. constrained + U-shape)
│ ├── cart.py # 250 kg / 1.5 m³ cart, multi-trip splitter
│ ├── wave_picking.py # N operators in parallel
│ ├── deadline.py # SLA check + operator-count recommendation
│ ├── incidents.py # Blocked aisle / missing SKU / wrapper down
│ ├── forklift.py # Forklift dispatch + OR-Tools TSP
│ ├── wrappers.py # 8 stretch wrappers + queue assignment
│ ├── history.py # JSON history + analytics
│ ├── metrics.py # Distance, time, cold exposure, doors
│ └── exporters.py # CSV / Excel / PDF (with QR)
├── visualization/
│ ├── map_view.py # Folium warehouse map (warehouse-only)
│ ├── view_3d.py # Plotly 3D rack scene
│ ├── animation.py # Walker animation
│ ├── heatmap.py # Pick-frequency heatmap
│ ├── charts.py # Bar / donut / side-by-side / wrapper gauges
│ └── graph_view.py # NetworkX node-link diagram
├── data/
│ ├── warehouse_config.py # Geometry, zones, doors, docks, wrappers
│ ├── catalog.py # 45 real French products
│ ├── operators.py # 20 named pickers + 4 forklift drivers
│ └── sample_orders.py # Random + 4 presets
└── assets/
└── style.css # Dark theme polish
Iyad Belkadi