Desire-Driven Reasoning Considering Personalized Care Preferences
研究护理机器人如何根据用户的长期和短期偏好，从多个选项中推理出满足抽象生理欲望（如口渴）的服务，并通过仿真和真实实验验证模型可行性。
Care robots have been developed to address the shortage of caregivers in hospitals and homes. However, providing care services considering abstract physiological desires (e.g., hunger, thirst, hot, and cold) rather than specific commands (e.g., serve a cup of water) has not been fully investigated. The key to such services that satisfy abstract requests, e.g., thirst, is to evaluate the available options (e.g., milk, tea, biscuit, and Coke) with the consideration of the care recipient’s preferences. In this study, we argue that both general taste (long-term care preferences) and sequential influence [short-term care preferences (STCPs)] should be considered for a robot to provide satisfying services. In this article, a long STCP model (LSTCPM) is introduced. The model can be initialized and updated incorporating the feedback from the care recipients. Then, a virtual agent-based simulation system is presented, which provides a way to evaluate and fine-tune the LSTCPM prior to its deployment to actual care robots. Finally, experiments were conducted in a real household domain using our care robot KUT-PCR to demonstrate the feasibility of the proposed approach.