A paraconsistent system of autonomous agents for Brazilian bank check treatment

This work is part of the Multicheck Project that defines the architecture of autonomous agents for the automatic treatment of handwritten Brazilian bank checks. The competence of these agents is implemented in two layers. The first corresponds to pattern recognition algorithms directly applied to image segments. The second one corresponds to reasoning mechanisms applied to the information from the first layer, either to validate or to interpret it. The interpretation process also involves information obtained from other agents. This information can present inconsistencies. This problem is treated properly and naturally through the concepts and operators of paraconsistent logic. This paper focuses on the second layer, on task distribution problems and on communication between agents. The first layer information was obtained through a simulated database