Probabilistic-based scheduling for runtime goal sequence of agents

Goal selection is very important in agent research. Many existing solutions are based on the static plans or predefined goal structures. This paper studies the problem and proposes a novel approach to scheduling goal sequence in goal-oriented agent systems. The proposed approach provides a unified probabilistic framework for building the Goal-oriented agents and a run-time policy for scheduling the goal sequence to enable agents to have a more engaging experience in a dynamic environment. Statistical analysis of the experiment on an interactive educational game shows that the proposed approach improves the effectiveness of the assistant agents in the game and promotes diffusion of the curiosity and challenge levels.