Hierarchical Organization of Robots
The sort of situation we have in mind would be a team of robots cleaning the floor and at the same time monitoring inside a museum during the night (such robots, at least as single systems, exist). Each of them can sense and act in a variety of ways and work independently. Each robot has a map of the museum but is "responsible" for a particular area, e.g. room.
The robots know where they are and can locate objects in the map. However, over time some functions will fail in some robots, so an organisational structure would be advantageous in order to avoid having to repair each of the robots once they fail or once their functionality is reduced. If one robot fails, its corresponding "area of responsibility" would not be covered properly. On the other hand its fellow robots might have "spare time" which they could use to help out. As long as the group can compensate for deficiencies of single robots and they can fulfil their job in the given amount of time (during the night) then the group of robots can function autonomously. The next day (or once a week) the robots can be repaired during routine checkup, but their night shift job would not be interrupted.
There are distinct literatures on the simulation of hierarchical organization of robots and the simulation of emerging or alternative structures of social organizations. The robotics literature does not consider purposive development of organizational structures by robots while the emergence of such structures in simulation models of simple agents is the meat and drink of computational organization theory.
The representation of the environment employed in this paper is a digit-string framework from a qualitative model previously shown by Moss (1998) to map into several real-world domains including environmentally sensitive critical-incident management and the management of complex design and implementation projects of which an actual example is the development of space-launch vehicles. Consequently, we claim that the models and results reported here are not restricted to simple, test problems but generalize naturally to real environments.
Generated with CERN WebMaker