家电科技 ›› 2022, Vol. 0 ›› Issue (5): 16-21.doi: 10.19784/j.cnki.issn1672-0172.2022.05.001

• 论文 • 上一篇    下一篇

智能家居系统的Multi-Agent建模研究

曲宗峰   

  1. 中国家用电器研究院 北京 100176
  • 出版日期:2022-10-01 发布日期:2022-11-01
  • 作者简介:曲宗峰,教授级高级工程师,硕士,天津大学在读博士。研究方向:人工智能、机器学习。地址:北京市亦庄经济技术开发区博兴八路3号。E-mail:quzf@cheari.com。

Research on smart home system of Multi-Agent modeling

QU Zongfeng   

  1. China Household Electric Appliance Research Institute Beijing 100176
  • Online:2022-10-01 Published:2022-11-01

摘要: 以智能家居系统作为研究对象,通过Multi-Agent理论建立了相关模型,并采用价值分解网络(Value Decomposition Networks,VDN)作为模型算法对Q函数进行了优化分析。以此提出了建立智能家居中各智能体规则库和知识库,并系统式开放建立各智能体Agent的BDI(信念-需求-意图)集合的建议,从哲学逻辑到收益建模对智能家居使用效用进行量化分析及优化。

关键词: 多智能体系统, 智能家居, 场景, 强化学习, Q学习

Abstract: Takes the smart home system as the research object, establishes the relevant model through the Multi-Agent theory, and uses the Value Decomposition Networks (VDN) as the model algorithm to optimize the Q function. Based on this, it is proposed to establish the rules database and knowledge database of each Agent in the smart home, and systematically open the BDI (Belief-Demand-Iintention) set of each Agent, so as to quantitatively analyze and optimize the use utility of the smart home from philosophical logic to income modeling.

Key words: Multi-Agent system, Smart home, Scene, Reinforcement learning, Q-learning

中图分类号: