Abstract: In college physics experiment teaching, issues such as tight scheduling of experiment time, students' having difficulties in comprehending experimental principles and mastering operational essentials, and teachers' inability to provide effective personalized guidance persist. To address these challenges, this study develops an artificial intelligence agent-assisted system to enhance experimental teaching effectiveness. The paper proposes AI agents across different platforms and conducted comparative analyses of their differences. The systems application effectiveness is based on acousto-optic diffraction experiments and sound velocity measurement in liquids. The experiment involves a control group (traditional teaching methods) and an experimental group (AI agent-assisted). Results show that in the high-score range (30-40 points), the experimental group shows a 12.2% improvement in pre-experiment quiz scores compared to the control group. During the experimental operation phase, the AI agent provides personalized guidance, resulting in significantly lower error rates and an average 10% reduction in experiment completion time for the experimental group. This research on AI agent development for university physics experiments and its exploration of enhancing experimental teaching through intelligent agents offers valuable references for AI agent development in other courses.

Key words: AI agent, college physics experiments, assisted instruction