This paper describes a parallel architecture and a navigation strategy for multiple robots equipped with simple computational units and communication devices. In particular, with the parallel paradigm, the aim is to consider such a collection of computational units as a cluster capable of performing parallel computations at a level similar to a classical massive parallel machine, albeit at a fraction of their cost. Furthermore, the autonomous navigation strategy discussed allows each robot to navigate in dynamic environments with little external guidance. Indeed, using data from its sensors the robot is able locate itself and detect possible changes in the environment. For this reason the Intelligent Building paradigm. on which the project is based, is presented. Moreover, the use of a particular form of neural network is presented to allow each single robot to detect changes in the environment. The initial experimental results from the MOBINET project presented here indicate that the technique is pratical and robust.
展开▼