Defeasible Inference in Intelligent Agents of BDI Type

The method with intelligent agents in multi-agent systems (MAS) in comparison with the traditional approaches enables the increase of efficiency in information protection, including their adequacy, failure resistance, destruction resistance, universality, flexibility, etc. [1] The success of such systems is due to a great extent to the distributed way of solving the problems and the search for an approximate solution instead of an optimal one [2]. A specific feature of the distributed decision making in MAS is the necessity every agent to make decisions in lack of complete information concerning the environment and the other agents [2]. The lack of information is overcome using techniques like messages exchange and the use of protocols. The co-operation of the agents has its specific problems connected with the solution of the conflicts occurring among them [3, 4]. The present paper discusses the problems of solutions annulment in multi-agent systems – one of the leading directions for solving conflicts in these systems. At first a short review is made of previous works in the area of MAS. After that a system of notions is introduced facilitating the solutions annulment in the area together with the corresponding logical formalism. Some examples are given of agents. In which the approach considered cannot be applied. The paper ends with some inferences.