基于多智能体协同的动态教学优化系统

Dynamic Teaching Optimization System Based on Multi-Agent Collaboration

摘要: 针对传统智能教学系统缺乏动态适应性、忽视学生心理状态以及师生交互不足等问题，文章提出了一种基于多智能体强化学习的个性化教学系统。该系统通过学生智能体和教师智能体的协同优化，实现教学策略的动态调整。学生智能体基于Q-Learning算法，综合考虑掌握度、动机、疲劳度等多维状态构建智能体状态与动作空间，自主选择最优学习动作；教师智能体融合启发式规则与强化学习，根据学生状态制定个性化的教学节奏、难度和策略；评估智能体负责对学习效果进行深度诊断与可解释反馈。该系统能有效提升学生的知识点掌握度，提高教师教学质量。同时，系统集成DeepSeek大语言模型，实现个性化学习报告的生成，为用户提供沉浸式的交互体验。研究为智能教学系统的设计提供了新的思路，具有良好的理论价值和应用前景。

Abstract: Aiming at the problems of traditional intelligent tutoring systems, such as a lack of dynamic adaptability, neglect of students’ psychological states, and insufficient teacher-student interaction, this paper proposes a personalized teaching system based on multi-agent reinforcement learning. The system realizes the dynamic adjustment of teaching strategies through the collaborative optimization of student agents and teacher agents. The student agent, based on the Q-Learning algorithm, constructs the agent state and action space by comprehensively considering multidimensional states including knowledge mastery, motivation, and fatigue level, and independently selects the optimal learning action; the teacher agent integrates heuristic rules and reinforcement learning to formulate personalized teaching rhythm, difficulty level, and strategies according to students’ states; the evaluation agent is responsible for in-depth diagnosis of learning effects and interpretable feedback. The system can effectively improve students’ knowledge mastery and enhance the quality of teachers’ teaching. Meanwhile, the system integrates the DeepSeek large language model to realize the generation of personalized learning reports and provide users with an immersive, interactive experience. The work of this paper provides a new idea for the design of intelligent tutoring systems, with good theoretical value and application prospects.