Delaunay triangulations have been extensively used in many applications and, in particular, in surface reconstruction and mesh generation. Although they enjoy many nice properties, Delaunay triangulations have limitations: they are difficult to compute in high dimensions, may contain flat simplices, and are strongly tied to Euclidean spaces. The lectures will show how to overcome these limitations. Two main applications will be considered: the reconstruction of submanifolds embedded in high dimensional spaces and the triangulation of Riemannian manifolds. 1. Good Delaunay triangulations. 1.1. Delaunay triangulation of nets : combinatorial bounds, random subsets, algorithmic complexity. 1.2. Thick triangulations. 1.3. Stability and protection of Delaunay complexes. 2. Sampling theory for geometric objects 2.1. Sets of positive reach. Sampling conditions and quality of approximation. 2.2. Distance functions and homotopic shape reconstruction. 2.3. Reconstruction of submanifolds of $\R^d$. 3. Triangulation of manifolds 3.1. Anisotropic triangulations. 3.2. Delaunay triangulation of Riemannian manifolds. 3.3. Local criteria for homeomorphism.