한국어 음성인식 기반의 비정형 빈피킹 시스템

Abstract

This study proposes a system for performing bin-picking tasks in unstructured environments using Korean speech recognition. In practice, unstructured bin-picking environments utilizing robots face challenges such as hardware performance limitations, AI model uncertainties, and non-experts’ operational inefficiencies. To address these issues, a method was designed to control bin-picking tasks through Korean voice commands, allowing users to monitor the system status or modify commands in real time during operation. In particular, a voice feedback system based on Large Language Model (LLM) was introduced to enable intuitive and efficient error correction and task optimization. By integrating Korean speech recognition with a bin-picking system through the optimization of multiple AI models, this system aims to enhance worker productivity and convenience while maximizing operational efficiency in unstructured environments.