A topologically-driven strategy to prevent instability in rule-based autonomous agents

Pervasive computing has been found to have a fundamental problem of instability. This problem is due to interaction between distributed but coordinating rule based autonomous agents in pervasive and intelligent environments. It is impossible to predict when a set of rules will lead the system to an unstable state, as it depends on the rules and on the state of the pervasive environment. Clearly, such instability could be a major obstacle to the commercial success of pervasive computing. In this paper we present a method of describing and reasoning about such behaviour, an Interaction Network (IN), together with a strategy based on finding the loops and locking the nodes in a way that minimizes the impact on the other devices and the user. This strategy is tested successfully with different topologies and rules.