Exploiting GenAI for Plan Generation in BDI Agents

Extending BDI agents with the ability to autonomously generate plans has long been a goal in cognitive agent engineering, aiming to improve their adaptability in dynamic environments. Recent advances in GenAI offer promising new opportunities in this area by leveraging the natural language understanding, means-end reasoning, and abstraction capabilities of LLMs. This thesis explores the integration of GenAI-driven plan generation into AgentSpeak(L) agents, examining how knowledge can be effectively transferred between the LLM and the BDI agent to support dynamic, runtime plan creation. To this end, a novel framework is proposed that extends the AgentSpeak(L) reasoning cycle with generative capabilities, enabling agents to synthesize plans on-the-fly. The design and implementation of this generative process are discussed, along with the architectural modifications required. The framework is implemented using the JaKtA BDI interpreter, and the quality of the generated plans is systematically evaluated across different types of LLMs.