ORDRIS Embroidery
Optimisation des trajets de point : une petite géométrie
Févr. 2026
An embroidery design is, to the machine, an ordered list of stitches: thousands of needle penetrations grouped into colour blocks, with the needle moving between them. Between two stitches that are not adjacent, the machine either drags a thread across the back of the fabric — a jump — or cuts and re-secures it — a trim. Both cost time, both cost thread, and a careless ordering produces a back full of loose jumps that must be picked out by hand. The geometry of the path is not cosmetic; it is most of the cost.
The problem has the flavour of a travelling salesman: visit every region in an order that minimises the total distance moved with the needle up. But it is constrained in ways the textbook version is not. Stitches within a block must keep their internal order to render correctly; colours impose a partial sequence because a thread change is expensive and you want to finish a colour before leaving it; and a long jump over a delicate area is worse than a slightly longer path that stays on stable fabric.
So the optimisation is not a single clever algorithm but a layered one. Order the colour blocks to minimise thread changes; within the constraints, order the regions of each colour to shorten the up-needle travel; decide jump-versus-trim per gap from the jump length and the fabric. Each layer respects the constraints the one above has fixed. A pure shortest-path solver, blind to these rules, produces a mathematically tidy route that an embroiderer cannot run.
The gains look small per garment and are enormous in aggregate. Shave a few seconds and a metre of thread from one design, then run it ten thousand times across a production order, and the saving is measured in hours of machine time and spools of thread — and in the trims that no longer have to be cleaned up by hand. A small geometry, repeated at scale, is where the economics of the workshop actually live.
This is the kind of problem we like: unglamorous, bounded, and consequential. The win is not a breakthrough but a careful ordering of known costs — which, multiplied by volume, is what makes the difference between a tool that demos well and one that pays for itself on the floor.