Towards Dynamic Epistemic Learning of Actions for Self-Improving Agents and Multi-agent Systems

摘要

Action learning is an important aspect of self-improving. This paper explores a new approach for the learning of two types of actions, namely precondition-free actions and conditional actions. The corresponding two learning algorithms are designed and implemented using modern logic reasoners. Finally, a simple system of action learning agents is implemented to explore cooperative self-improving multi-agent systems.