Thijs Roumen, Conrad Lempert, Ingo Apel, Erik Brendel, Markus Brand, Laurenz Seidel, Lukas Rambold, and Patrick Baudisch

AutoAssembler automatically converts laser-cut 2D cutting plans to 3D models so users can perform parametric manipulations on the models. Unlike our earlier paper assembler3 where users manually reconstruct models. AutoAssembler uses a beam search algorithm to search possible ways of assembling plates. It uses joints on these plates to combine them into assembly candidates. It thereby preferably pursues candidates (1) that have no intersecting plates, (2) that fit into a small bounding box, (3) that use plates whose joints fit together well, (4) that do not add many unpaired joints, (5) that make use of constraints posed by other plates, and (6) that conform to symmetry axes of the plates. This works for models that have at least one edge joint (finger or t-joint). In our technical evaluation, we imported 66 models using autoAssembler. AutoAssembler assembled 79% of those models fully automatically; another 18% of models required on average 2.7 clicks of post-processing, for an overall success rate of 97%.

The search space for the simple VR headset consisting of 9 parts and 33 joints after limiting our search to joints that fit and only exploring orientations of plates that do not lead to an immediate collision. (Labels denote the number of joints that fit at a given position x the number of orientations they fit in). This is obviously intractable, which is why we devise a beam-search algorithm to traverse the space of possible combinations. The key to making this work is a set of heuristics that determine what is a good laser cut assembly

Heuristic 1: give preference to compact candidates: (a) Adding the red plate gives it a much larger bounding box, autoAssembler thus gives (b) this candidate the higher score.(c) The resulting magazine holder.

heuristic 2: Avoid intersections between plates:  (a) The compactness heuristic suggested mounting this bearing (red) on top of this train wagon, but here it intersects with plates mounted on top of the wagon, causing autoAssembler (b) to flip the bearing plate to its correct position. (c) the same heuristic fixes the other bearings too.

heuristic 3: Give preference to unambiguous joints: (a) Assembling this ambiguous joint early on forms little or no constraints on other plates, as a result the top plate here can be assembled in many different ways (b) autoAssembler prefers to greedily connect plates with high probabilities. This adds constraints for other plates, (c) to eventually make this dice tower.

heuristic 4: Minimize the number of unmatched joints: AutoAssembler prioritizes completing joints first as this reduces the search space: (a) inserting the side plate (red) would add four incomplete joints to the model. (b) AutoAssembler therefore rather adds this “divider” plate, which only adds one unmatched joint. (c) leading to this desktop organizer.

heuristic 5: Make use of constraints posed by other plates: (a) This candidate offers too few constraints to orient the red plate correctly. (b) AutoAssembler therefore prioritizes this plate, which completes three joints. (c) This adds constraints that come in handy when eventually inserting the middle piece.

heuristic 6: Favor symmetric candidates: (a) AutoAssembler detects symmetries by having pairs of joints vote for symmetry axes (b) autoAssembler uses the information to prefer similar plates connected to joints on opposite sides of the symmetry axis. (c) resulting in this birdhouse