An Architecture for Universal Knowledge-based Agent

Typical approach to agent-oriented programming results in slight limitations of agent’s usability. One agent is typically usable only in single environment type and for a small, strictly given set of tasks. We introduce a new term Universal Knowledge-based Agent (UKA) and propose an architecture and needed set of algorithms, providing complete methodology for its implementation. One single implementation of this agent is usable in any environment and for various practical tasks, with modifications only in its hardware interface, thanks to environmentally-independent knowledge representation and general motivational mechanism inspired by a model from psychology. Convenient formalism for symbolic knowledge representation and reasoning is described in detail in this thesis, together with significant modifications of classical approaches.