破解物流专业本科科研瓶颈：AI赋能强化学习模型开发的教学探索

Breaking the Bottleneck in Undergraduate Research of Logistics Major: Teaching Exploration on AI-Enabled Reinforcement Learning Model Development

摘要: 针对物流专业本科生日益增长的深造需求与科研能力薄弱的矛盾，本文以本科创新实践课程为载体，开展“AI辅助开发强化学习模型”教学实践研究，旨在破解学生编程能力不足的科研瓶颈。研究选取主流AI辅助编程工具，以应急物资调配强化学习环境开发为核心任务，构建多维度评估体系验证工具有效性。结果表明，工具在标准化开发环节适配率达100%，其中豆包对场景化需求适配最优；同时明确学生与AI的分工边界——学生聚焦需求拆解、文献提取与公式转化，AI负责代码生成与框架调用。实践还暴露学生英文文献信息提取效率低等问题，据此提出强化专业英文训练、设计“数学–代码”转化模块的改进方向。研究可为非计算机专业本科科研能力培养提供参考，助力智慧物流领域创新型人才输送。

Abstract: Aiming at the contradiction between the growing postgraduate study needs of undergraduate students majoring in logistics and their weak research capabilities, this paper conducts a teaching practice study on “AI-assisted reinforcement learning model development” based on the undergraduate innovative practice course. The purpose is to address the research bottlenecks caused by students’ insufficient programming skills and limited effective research time. The study selects state-of-the-art AI-assisted programming tools, takes the development of a reinforcement learning environment for emergency material allocation as the core task, and constructs a multi-dimensional evaluation system to verify the effectiveness of the tools. The results show that the adaptation rate of the tools in standardized development reaches 100%, among which Doubao has the best adaptation to scenario-based requirements. At the same time, the division of labor boundary between students and AI is clarified: students focus on requirement decomposition, literature extraction, and formula conversion, while AI is responsible for code generation and framework invocation. The practice also reveals problems such as students’ low efficiency in extracting information from English literature and insufficient rigor in converting mathematical formulas into code. Accordingly, improvement directions are proposed, including strengthening professional English training and designing a mathematics-to-code transformation module. This study can provide a reference for cultivating the research capacity of non-computer science undergraduates and sup-port the cultivation of innovative talents in the field of intelligent logistics.