Using a LLM-Based Conversational Agent in the Social Robot Mini

Natural Language Processing has witnessed significant growth in recent years. In particular, conversational agents have improved significantly thanks to the proliferation of the Large Language Models (LLM). Conversational agents have already been integrated with smartphones, smart speakers, or social robots (SRs). Unlike the mentioned electronic devices, a social robot allows more active and closer user engagement due to the presence of a physical object with a lifelike appearance that is able to express emotions. Therefore, SRs represent an appealing platform for deploying a conversational agent. In the field of social robotics, the ability of robots to interact with humans has traditionally been limited by their verbal skills. Until recently, robots could only understand a limited set of human utterances using specific rules, and the utterances of the robots were pre-defined sentences crafted offline. These restrictions, on many occasions, lead to repetitive interactions, which could cause users to lose interest during prolonged engagement with the robot. In this paper, we propose to integrate into our social robot Mini a conversational agent based on LLM. We present a new robot skill that can maintain a natural and seamless conversation with the user on any desired topic. The obtained results show a high usability of the skill and a high-quality interaction.