An Implementation for Abductive Logic Agents

This paper presents the distributed implementation of ALIAS, an architecture composed of several cooperating intelligent agents. This system is particularly suited to solve problems in cases where knowledge about the problem domain is incomplete and agents may need to form reasonable hypotheses. In ALIAS agents are equipped with hypothetical reasoning capabilities, performed by means of abduction: if the knowledge available to a logic agent is insufficient to solve a query, the agent could abduce new hypotheses. Each agent is characterized by a local knowledge base represented by an abductive logic program. Agents might differ in their knowledge bases, but must agree on assumed hypotheses. That global knowledge base is dynamically created and managed by means of a shared tuple space. The prototype, developed using Java and Prolog, can run on a TCP/IP network of computers. In the paper, we also discuss some experimental results to evaluate prototype efficiency.