Observational Logic for Composition of Agents

摘要

We propose a logic language for agents and their composition in a limited discrete space where they can perform activities in mutual competition. We suppose that each agent has the same activity capability which they try to apply over a given space of resources. This logic is based only on external observations of agent's activities in this scenario, and does not take into consideration the internal knowledge structure of agents. In this framework an event is considered as the simplest kind of agents whose activity is just its presence in a given space/time. The data structure of basic agents in this logic theory are ground atoms of a Herbrand base over a given domain/universe for agents. The composition operations for the agents are based on the complete lattice of agents w.r.t. their observational qualities. Such compositions of agents are parameterized by different kinds of conjunctive/disjunctive activity strategies, and integrated into this propositional language for agents, augmented by statements that are able to reason about agent's activities. Finally, we show how the complex algebraic structure of agents can be fully-embedded in the multi-modal 2-valued propositional logic where we are able to express higher level statements about agents, as, for example, their overlapping in a working space, and we show that such multi-modal framework can be flattened into Constraint Logic Programming. The flattened agent's structures can be conveniently saved in the standard Relational Databases.