文献来源：

Lei Song;Chuheng Zhang;Li Zhao;Jiang Bian.Pre-Trained Large Language Models for Industrial Control[J].2023/WXFX_2023-Pre-Trained_Large_Language_Models_for_Industrial_Control.pdf

文章摘要：

For industrial control, developing high-performance controllers with few samples and low technical debt is appealing. Recently, foundation models are shown to be powerful in dealing with various problems with only few (or no) demonstrations, owing to the rich prior knowledge obtained from pre-training with the Internet-scale corpus. To explore the potential of foundation models in industrial control, we design mechanisms to select demonstrations and generate the prompt for foundation models, and then execute the action given by the foundation models. We take controlling HVAC (Heating, Ventilation, and Air Conditioning) for buildings via GPT-4(one of the first-tier foundation models) as an example, and conduct a series of experiments to answer the following questions: 1) How well can GPT-4 control HVAC? 2) How well can GPT-4 generalize to different scenarios for HVAC control? 3) How do different designs affect the performance? In general, we found GPT-4 achieves a performance comparable to RL methods but with fewer samples and lower technical debt, indicating the potential of directly applying foundation models to industrial control tasks.

分享内容：

声明：本网站内容版权归许鹏教授课题组所有。未经本课题组授权不得转载、摘编或利用其它方式使用上述作品。