Multi-Constrained Optimization for Example-Based Synthesis

Table 1 in the paper presents related content generation approaches based on graph search, constraint graph search and dynamic programming. Column resource constraints/type denotes the number of resource constraints the algorithm can handle. Type refers to the demand, how a constraint might be satis ed/achieved. If a constraint needs to be satis ed exact, then it is referred to an equality constraint. If it is restricted to any discrete value between a lower and upper bound, then its an interval constraint. Key points provide the capability of an approach to support placing concrete elements ('Element') or a class of elements ('Class') on user de ned or domain speci c positions. Examples for such classes are 'l-shaped' and 't-shaped' building parts in the building synthesis application. In the motion synthesis application, classes represent multiple motions segments performing the same action e.g 'jump', which might contain segments, where the character performs the jump on the right or the left leg. The last column 1.5D reports if an approach can handle tree or graph structures that can be solved globally optimal and avoiding element ambiguities at once in a single graph traversal.