tailieunhanh - Robot Localization and Map Building Part 5

Tham khảo tài liệu 'robot localization and map building part 5', kỹ thuật - công nghệ, cơ khí - chế tạo máy phục vụ nhu cầu học tập, nghiên cứu và làm việc hiệu quả | 134 Robot Localization and Map Building we focus on this approach. In this context two main proposals can be found in the literature. On the one hand there are some solutions in which the estimate of the maps and trajectories is performed jointly Fenwick et al. 2002 Gil et al. 2007 Thrun Liu 2004 . In this case there is a unique map which is simultaneous built from the observations of the robots. In this way the robots have a global notion of the unexplored areas so that the cooperative exploration can be improved. Moreover in a feature-based SLAM a landmark can be updated by different robots in such a way that the robots do not need to revisit a previously explored area in order to close the loop and reduce its uncertainty. However the maintenance of this global map can be computationally expensive and the initial position of the robots should be known which may not be possible in practice. On the other hand some approaches consider the case in which each robot builds a local map independently Stewart et al. 2003 Zhou Roumeliotis 2006 . Then at some point the robots may decide to fuse their maps into a global one. In Stewart et al. 2003 there is some point where the robots arrange to meet in. At that point the robots can compute their relative positions and fuse their maps. One of the main advantages of using independent local maps as explained in Williams 2001 is that the data association problem is improved. First new observations should be only matched with a reduced number of landmarks in the map. Moreover when these landmarks are fused into a global map a more robust data association can be performed between the local maps. However one of the drawbacks of this approach is dealing with the uncertainty of the local maps built by different robots when merging them. The map fusion problem can be divided into two subproblems the map alignment and the fusion of the data. The first stage consists in computing the transformation between the local maps which have .