jcbb-simulation.tar

  analyze data association algorithms (Nearest Neighbour and JCBB) in SLAM. Data association is basically a combinatorial optimization problem whose goal is to find the “best” set of associations between the observations and predictions according to a particular criterion (e.g. individual compatibility). Form this point of view, near- est neighbour (NN, a.k.a. ML) is simply a “greedy” algorithm in which the “compatibility” of individual associations with each other is not examined. On the contrary, JCBB, by considering both the individual and joint compatibility, is able to yield much better results. JCBB uses the branch and bound technique in order to prune the search tree