Adaptive grids: an image-based approach to generate navigation meshes

We propose adaptive grids, an image-based approach for constructing navigation meshes, which are used for path planning. A cellular navigation mesh, called an adaptive grid, is constructed from a top-view range image of a three-dimensional urban model. A navigation graph can then be extracted from this adaptive grid for path planning. We compare our approach with two popular navigation mesh-generation approaches and obtain promising results in terms of path accuracy and memory cost.