Publicação
Prototyping an LLM-Based AI Agent for Operator Support in Augmented Reality Industrial Quality Inspection
| Resumo: | This paper presents the design, implementation, and initial empirical evaluation of an LLM-based AI agent integrated into an Augmented Reality (AR) environment for quality inspection support in an Industry 5.0 manufacturing context. The fully implemented system combines spatially anchored visual overlays with a conversational AI assistant grounded in a deployed Retrieval-Augmented Generation (RAG) pipeline over Manuel da Conceição Graça, Lda official inspection documentation. A preliminary within-subject user study (N= 6) was conducted to compare two implemented input modalities, virtual keyboard and speech-to-text, across a set of simulated industrial inspection tasks. The objective of the study was to assess the usability of the implemented system with first-time AR users, and to determine which input modality better fits the hands-busy demands of shop-floor inspection. Results show that voice interaction was consistently preferred for ease of use, speed, and comfort, while the virtual keyboard was perceived as more precise but more demanding. Both conditions received positive SUS scores, confirming that the implemented system is usable without domainspecific training. |
|---|---|
| Autores principais: | Latas, Martim |
| Outros Autores: | Marques, Gabriel; Macedo, Patricia; Santos, Pedro Albuquerque; Freitas, Jose; Madeira, Rui Neves |
| Assunto: | Augmented Reality Human-AI Collaboration Industry 5.0 Large Language Models Mixed Reality Artificial Intelligence Computer Science Applications Information Systems Energy Engineering and Power Technology Electrical and Electronic Engineering Instrumentation |
| Ano: | 2026 |
| País: | Portugal |
| Tipo de documento: | documento de conferência |
| Tipo de acesso: | acesso embargado |
| Instituição associada: | Universidade Nova de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório Institucional da UNL |
| Resumo: | This paper presents the design, implementation, and initial empirical evaluation of an LLM-based AI agent integrated into an Augmented Reality (AR) environment for quality inspection support in an Industry 5.0 manufacturing context. The fully implemented system combines spatially anchored visual overlays with a conversational AI assistant grounded in a deployed Retrieval-Augmented Generation (RAG) pipeline over Manuel da Conceição Graça, Lda official inspection documentation. A preliminary within-subject user study (N= 6) was conducted to compare two implemented input modalities, virtual keyboard and speech-to-text, across a set of simulated industrial inspection tasks. The objective of the study was to assess the usability of the implemented system with first-time AR users, and to determine which input modality better fits the hands-busy demands of shop-floor inspection. Results show that voice interaction was consistently preferred for ease of use, speed, and comfort, while the virtual keyboard was perceived as more precise but more demanding. Both conditions received positive SUS scores, confirming that the implemented system is usable without domainspecific training. |
|---|
Atividades financiadas
Não foram encontrados projetos financiados.