ABSTRACT
In this chapter, we will consider the problem of self-organizing a spatial map in an unknown environment by using a group of autonomous robots. Generally speaking, map building in an unknown environment presents a
challenging problem in robotics. Some earlier studies have tackled this problem by using exact search algorithms to derive graph-like representations of the environment. An example of this approach is the work by Betke et al. [BRS94] on the piecemeal learning of a robot environment containing convex obstacles. Others [BBHCD96, VBX96] have addressed the problem by modeling an unknown environment with a set of basic geometric primitives such as line segments, circles, regions, landmarks, and/or local maps. In such an approach, incremental learning algorithms, such as Kohonen neural networks and Kalman filters, are often applied [HBBC96, JGC+97, Koh88, KE94]. While the majority of the map building studies deal with the problems of modeling two-dimensional environments, some researchers [BH94] have investigated the use of a self-organizing approach in reconstructing an unknown three-dimensional surface.
