LMT-skeleton heuristics for several new classes of optimal triangulations

Given a planar point set, we consider three classes of optimal triangulations: (1) the minimum weight triangulation with angular constraints (constraints on the minimum angle and the maximum angle in a triangulation), (2) the angular balanced triangulation which minimizes the sum of the ratios of the maximum angle to the minimum angle for each triangle and (3) the area balanced triangulation which minimizes the variance of the areas of triangles in the triangulation. With appropriate definition of local optimality for each class, a simple unified method is established for the computation of the subgraphs of optimal triangulations. Computational experiments demonstrate that the method successfully identifies large portion of edges of the optimal triangulations of each class for all problem instances tested, and hence optimal triangulations for each class can be obtained from them by applying dynamic programming.