Direct computation of minimal rotation for support slimming

To reduce the usage of supporting structure in additive manufacturing, an orientation-driven shape optimizer was developed in our prior work [Hu et al. 2015] which employs a volumetric mesh enclosing the input 3D models as the domain of computation. The orientation of a model is computed indirectly by surface of the volumetric mesh. In this paper, we extend our indirect computation to an approach that the computation is directly based on the information of input models by an algorithm using incremental linear programming and K-means clustering. The performance of this approach is decoupled from the shape similarity between the volumetric mesh and the input model. As a result, representations obtained from simpler volumetric decomposition such as voxels can be adopted as the domain of computation.