AoECR: AI-ization of Elderly Care Robot

TL;DR

This paper introduces AoECR, a universal AI architecture for elderly care robots that enables secure, personalized, and humanized interactions, demonstrated through a new dataset and zero-shot generalization in diverse scenarios.

Contribution

The paper presents a novel universal AI architecture, AoECR, including a specialized dataset and a fine-tuned large language model for elderly care robots, with enhanced security and personalization features.

Findings

Demonstrated zero-shot generalization across scenarios

Enabled understanding of patient instructions

Ensured secure control command execution

Abstract

Autonomous interaction is crucial for the effective use of elderly care robots. However, developing universal AI architectures is extremely challenging due to the diversity in robot configurations and a lack of dataset. We proposed a universal architecture for the AI-ization of elderly care robots, called AoECR. Specifically, based on a nursing bed, we developed a patient-nurse interaction dataset tailored for elderly care scenarios and fine-tuned a large language model to enable it to perform nursing manipulations. Additionally, the inference process included a self-check chain to ensure the security of control commands. An expert optimization process further enhanced the humanization and personalization of the interactive responses. The physical experiment demonstrated that the AoECR exhibited zero-shot generalization capabilities across diverse scenarios, understood patients' instructions, implemented secure control commands, and delivered humanized and personalized interactive responses. In general, our research provides a valuable dataset reference and AI-ization solutions for elderly care robots.