![]() ![]() ![]() Therefore, we have developed OGM algorithms that maintain local cell correlations to a varying degree and this degree can be user-defined. We show that mutual information between cells in the full Bayesian posterior is concentrated locally and approaches zero for cells that are distant. While the full Bayesian posterior is intractable, we develop improved approximations and show how two popular techniques, variational inference (to optimize parameters of a chosen model type) and Markov Chain Monte Carlo (MCMC), can be used to relax traditional assumptions. This thesis revisits the assumptions made in traditional OGM and reintroduces cell correlations in the prior, the measurements, and the posterior. Computing the probability of all possible maps is computationally intractable and thus it is typically estimated using Bayesian inference to compute the probability of each cell and assuming that the cells are independent. Occupancy grid mapping (OGM) is a popular mapping technique that discretizes the environment into cells (or voxels) and seeks to estimate the occupancy probability of each cell. These maps, although not perfect, are useful for path planning, exploration, search, and often the objective is to create an accurate map of the environment. A mobile robot equipped with a range sensor can create a map of its environment given the range measurements and corresponding robot pose(s). ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |