Same box.
Better fit by shape.
Watch two robots pack the same synthetic parts in a single-layer experiment. One reserves rectangular envelopes. The other nests the actual shapes. Project packing uses bounding boxes.
80 × 60 mm floor · 10 mm parts · 8 identical parts available · Single-layer experiment
A recess hidden by a rectangle.
The grid reserves each part’s rectangular envelope. The outlined space around each L remains unavailable to its neighbors.
Let the shapes fit together.
Nesting checks occupied material and allowed rotations, so another part can use the recess.
What about your parts?
Bring your part shapes, container dimensions and packing constraints. Let’s discuss how to evaluate your packaging in OptiPacker.
The gain starts with geometry.
Representing the L-shaped recess gives the extra space. For the asymmetric L part, the bounding-box grid fits 4 and shape-aware nesting fits 7. Jev fit 7 in each recorded trial, the same as the geometry search. The better fit comes from the part shape, not from Jev.
The recorded experiment compares geometric search with Jev candidate selection.
Read the scope and limits
Six synthetic fixtures. Identical parts in a single 10 mm layer, a 10 mm placement lattice and in-plane quarter turns. The half-turn fixture permits 0° and 180°; the clearance fixture requires 10 mm wall and part gaps.
Jev ranked candidate layouts from the geometry search. A geometric check validated the returned placements.
All 30 main layouts and 132 follow-up control layouts passed bounds, rotation, overlap and clearance checks. The demo computes grid and geometry in your browser and replays one recorded Jev trial per fixture. Jev timings below average three trials of jev-1.13.0 and include API latency and local search.
The robot illustrates the final placement order. The experiment measures single-layer layouts, with no gripper-collision or loading-path validation. Uploaded meshes, mixed parts, full 3D packing and production performance need further evaluation.
Download the public measurements and layouts · About TypeSafe
| Fixture | Grid | Geometry | Jev |
|---|---|---|---|
| Small L | 6 | 8 | 8 / 8 / 8 |
| Thin L | 2 | 5 | 5 / 5 / 5 |
| Asymmetric L | 4 | 7 | 7 / 7 / 7 |
| Half-turn L | 2 | 6 | 6 / 6 / 6 |
| L with clearance | 4 | 6 | 6 / 6 / 6 |
| Rectangle control | 6 | 6 | 6 / 6 / 6 |
| Grid | Geometry | Jev mean |
|---|---|---|
| 0.08 ms | 11.73 ms | 1,850.86 ms |