Geometric Modeling for Shape Classes Amitabha Mukerjee Dept of Computer Science IIT Kanpur http://www.cse.iitk.ac.in/~amit/ Representations from [Requicha ACM Surveys 1980] 2 Parametric design vs Conceptual Design Conceptual Variation approximated using a finite set of parameters Modeling Fixed Geometries 4 Mathematical Structures • Vectors, orthonormal bases – distances and norms – Angles • Transformations • Motions, boolean operations 5 6 Representing Geometrical Objects • As Primitives • Spatial decomposition • Boolean (Constructive) operations – Continuous constructions: Extrusion / Sweep • Boundary based modeling 7 Boolean operations 8 Intersection of solids not a solid 9 Boundary is not unique specifier • Depends on the embedding space – A boundary on a sphere may represent either side – May need additional neighbourhood information 10 Curves and Surfaces 11 • Implicit equations – Line: p = u.p1 + (1-u). p2 12 • Plane: (p-p0).n = 0 • If n = {a,b,c} and p0.n = -d, we have ax+by+cz+d=0 13 3D Solids : B-rep 14 Algorithms • Point membership classification – 2D planar shapes – 3D ?? • Line – Shape intersection • Solid boolean operations 15 Variational Shape Classes 16 Familiar Shapes 17 Familiar Shapes 18 Generating Variational Shapes 19 Generating Variational Shapes kilian-mitra-07 : Geometric-modeling-shape-interpolation, 20 Shape Classes for Conceptual Design 21 Design = Search in Ill-structured spaces From Goel [VSRD 99] Applications to Conceptual Design 1.Geometric Parametrization 2.Formulation of cumulative objective 3.Parameter Search and optimization 23 Constraints on Shape A Complete Faucet Driving Parameter Set : { Wo , Ho , Lo , 1 , 2 } Sub-parts: Inlet Outlet Cock Algorithms • Boolean operations on probabilistic sets – Point membership classification? • Output also in terms of probability density function • Boolean operations on objects and classes • Function evaluation 25 Generating Variational Shapes 143 “functionality“ - mathematical function “aesthetics” - User interaction Final Population of Faucets Names of instances of faucets shown are given as , [ (A , B); (B , C); (C , D) ] User Assigned Fitness Table A B C D E F 3 4 4 4 4 4 Conclusion • Computational processes are moving from deterministic to probabilistic • Geometric modeling will also need to move more in this direction, which is also cognitively viable. • Need structures for modeling ambiguous shapes • Many algorithmic challenges even for unique shapes, output for shape classes will also be probabilistic 28