Enforcing behavior with contracts

Abstract

Contracts have been introduced earlier as a way of modeling a collection of agents that work within the limits set by the contract. We have analyzed the question of when an agent or a coalition of agents can reach a stated goal, despite potentially hostile behavior by the other agents. In this paper, we extend the model so that we can also study whether a coalition of agents can enforce a certain temporal behavior when executing a contract. We show how to reduce this question to the question of whether a given goal can be achieved. We introduce a generalization of the action system notation that allows both angelic and demonic scheduling of actions. This allows us to model concurrent systems and interactive systems in the same framework, and show that one can be seen as the dual of the other. We analyze enforcement of temporal behavior in the case of action systems, and show that these provide for simpler proof obligations than what we get in the general case. Finally, we give three illustrative examples of how to model and analyze interactive and concurrent systems with this approach.