Realizing Space-oriented Control in Smart Buildings via Word Embeddings

TL;DR

This work shifts smart-building control from device-centric to space-centric paradigms by mapping natural language inputs to building APIs through word embeddings and a vector-search pipeline. It introduces a chat-based GUI and an embedding-driven backend to route NL requests to appropriate space-level controls, leveraging ontologies to improve interoperability. A university-based prototype demonstrates feasibility and portability, using Elasticsearch as a vector store and JSON-formatted API metadata for flexible API invocation. The approach promises scalable, goal-oriented control across buildings by standardizing spatial representations and integrating API metadata within building schemas.

Abstract

This paper presents a novel framework for implementing space-oriented control systems in smart buildings. In contrast to conventional device-oriented approaches, which often suffer from issues related to development efficiency and portability, our framework adopts a space-oriented paradigm that leverages natural language processing and word embedding techniques. The proposed framework features a chat-based graphical user interface (GUI) that converts natural language inputs into actionable OpenAI API calls, thereby enabling intuitive space level (e.g., room) control within smart environments. To support efficient embedding-based search and metadata retrieval, the framework integrates a vector database powered by Elasticsearch. This ensures the accurate identification and invocation of appropriate smart building APIs. A prototype implementation has been tested in a smart building environment at the University of Tokyo, demonstrating the feasibility of the approach.

Figures (4)

• Figure 1: Application Control Flow.
• Figure 2: Existing models VS our proposal
• Figure 3: Chat-based GUI
• Figure 4: Chat-based GUI