家电科技 ›› 2025, Vol. 0 ›› Issue (zk): 153-158.doi: 10.19784/j.cnki.issn1672-0172.2025.99.032

• 第一部分 优秀论文 • 上一篇    下一篇

智能家居和智慧座舱跨域智能体交互技术的研究探索

王统帅1,3,4, 田云龙1,2,3,4, 杜永杰1,2,3,4   

  1. 1.青岛海尔科技有限公司 山东青岛 266100;
    2.数字家庭网络国家工程研究中心 山东青岛 266100;
    3.山东省智慧家庭人工智能与自然交互研究重点实验室 山东青岛 266100;
    4.青岛市智慧家庭交互与控制工程研究中心 山东青岛 266100
  • 发布日期:2025-12-30
  • 作者简介:王统帅(1983年10月—),男,毕业于北京理工大学企业管理专业,现任青岛海尔科技有限公司项目经理。E-mail:wangtsh@haier.com。

The research and exploration of cross-domain agent interaction technology for smart home and smart cockpit

WANG Tongshuai1,3,4, TIAN Yunlong1,2,3,4, DU Yongjie1,2,3,4   

  1. 1. Qingdao Haier Technology Co., Ltd. Qingdao 266100;
    2. National Engineering Research Center of Digital Home Networking Qingdao 266100;
    3. Shandong Key Laboratory of Artificial Intelligence and Natural Interaction in Smart Home Qingdao 266100;
    4. Qingdao Engineering Research Center of Smart Home Interaction and Control Qingdao 266100
  • Published:2025-12-30

摘要: 伴随着智能家居和智能汽车的快速普及和推广,大部分用户都感受到了智慧生活和出行带来的便捷体验,同时也经常会遇到智慧家庭和智能汽车座舱无法互联互通的问题,导致用户体验不佳。针对当前智慧家庭与智慧座舱系统“数据孤岛、交互割裂、任务协同弱”的问题,提出一种基于MCP协议与大模型的跨域智能体交互技术。该技术将汽车智慧座舱(车载智能体)与智慧家庭智家大脑屏(家居智能体)作为核心交互节点,依托MCP协议实现跨系统设备控制与状态同步,结合大模型的上下文理解、任务拆分与延续能力,构建“场景联动-任务协同-服务定制”的跨域交互体系。从实验室数据来看,该技术的跨域指令响应延迟≤500 ms,任务延续成功率达94.3%,用户场景需求满足率提升至96.5%,较传统智能家居和智慧座舱两个独立系统交互相比,跨域智能体交互效率提升了3.8倍,显著打破了智慧家庭与智慧座舱的交互壁垒,实现用户智慧生活+智慧出行智能交互和无感协同的全场景极致体验。

关键词: 智慧家庭, 智慧座舱, 智能体, MCP协议, 大模型

Abstract: With the rapid popularization and promotion of smart homes and smart cars, most users have felt the convenient experience brought by smart life and travel, and they often encounter the problem that smart homes and smart car cockpits cannot be interconnected, resulting in poor user experience. Aiming at the current problems of data islands, interaction fragmentation, and weak task collaboration “between smart home and smart cockpit systems, proposes a cross-domain agent interaction technology based on MCP protocol and large model. This technology takes the automotive smart cockpit (vehicle agent) and the smart home smart home brain screen (home agent) as the core interaction nodes, relies on the MCP protocol to realize cross-system device control and state synchronization, and combines the context understanding, task splitting and continuation capabilities of the large model to build a cross-domain interaction system of scene linkage-task collaboration-service customization”. Judging from the laboratory data, the cross-domain instruction response delay of this technology is≤500 ms, the task continuation success rate reaches 94.3%, and the user scenario demand satisfaction rate has increased to 96.5%, which is 3.8 times higher than the interaction between the traditional smart home and the smart cockpit two independent systems, significantly breaking the interaction barriers between smart homes and smart cockpits, and realizing the ultimate experience of users’ smart life+smart travel intelligent interaction and non-inductive collaboration.

Key words: Smart home, Smart cockpit, Agent, MCP protocol, Large model

中图分类号: