Orders are generated on storage nodes only (never on cross-aisles).
Disjoint: If checked, generated orders do not have any item in common.

The solver-neutral PickBench contracts. Export the instance so another solver can work on it, then import that solver's solution to replay and check it here. Save Config above is the full workbench state; these two files are the portable problem and answer.

Runs the picking policies and routing methods against the orders currently in the warehouse. Walk is total floor travel; span is the wall-clock time once the pickers work in parallel, which is what the zoned policies are chosen for. The exact method proves the shortest possible route, so the gap column says what each heuristic really costs.

Runs one policy on the settings above and shows the shift it produces: which picker does which tour, when, and where the time goes. Idle is unused capacity measured against the whole shift — a picker who finishes early is idle for the rest of it, which is how an unbalanced zoning shows up.

Experiments automatically build a warehouse, create orders, and generate picking routes. Results are saved with performance statistics.

Load and run pre-configured experiment scenarios

🏭 Tiny Warehouse (3×3, 2 orders) 🏭 Small Warehouse (5×10, 20 orders) 🏭 Medium Warehouse (10×20, 50 orders) 🏭 Large Warehouse (15×30, 100 orders)

Clear All Experiments