Rational behaviour planning of cognitive semiotic agent in dynamic environment

The paper presents a general architecture of a cognitive semiotic agent acting in a dynamic environment. A new implementation and integration of the agent planning and learning subsystems are proposed to solve the symbol grounding problem. We suggest a new approach to the description of the semantic level of the sign-based world model component, which is used as a base for agent rational behavior synthesis. A formal definition of the behavior script and its use in generating a rational agent's action plan is proposed. In conclusion, we describe a model experiment that demonstrates the work of a semiotic agent in a game environment.