A SystemC Model of a Low-Level Inference Engine for Symbolic Reasoning in Cognitive Agents

Abstract

In this paper, a SystemC model of a low-level inference engine is designed to serve as reasoning mechanism of an embedded cognitive agent. In this work the Concurrent Autonomous Agent (CAA) was the cognitive agent architecture used as testbed. The architecture of the CAA comprises three levels, namely, the reactive level, instinctive level and the cognitive level, which runs concurrently. The reactive level consists of behaviours created over a embedded kinematic controller. The instinctive level receives perceptions from and sends the active behaviour to the reactive level. This level uses a Knowledge Based System (KBS) with First Order Logic (FOL) as automatic reasoning mechanism. The cognitive level receives symbolic information from the instinctive level to update its logical world model. It also sends new local goals to instinctive level. The SystemC library, used in this works, extends the standard C language with concurrency, timed events, data types and classes, which makes it suitable for system-level designs. A system-level design of an application specific processor that runs a KBS inference engine is constructed in SystemC, and simulation results shows that it behaves as expected.