Scholarlink

This chapter introduces two frameworks aimed at achieving the ultimate objective of this monograph: the development of personalised robot-assisted cognitive training (RACT) for Persons with Dementia (PwDs). The first framework employs Reinforcement Learning to learn the robot’s assistive behaviour that best matches the users’ preferences (adaptive) and the concept of Personas based on which therapists can initialise the robot’s behaviour in an easy way by generating patient-persona’ profiles (adaptable). Simulation-based evaluations demonstrate the system’s adaptability to different patient profiles, however, the approach did not fully address all the requirements defined by the therapists. Hence, the second framework called aCtive leARning agEnt aSsiStive bEhaviouR (CARESSER) addresses these limitations. It employs Learning from Demonstration through Inverse Reinforcement Learning to learn a patient-specific policy. CARESSER enables therapists to customise the robot’s assistive behaviour (adaptable) and allows the robot to learn in situ personalised patient-specific policies (adaptive) in an autonomous fashion. An experimental study was conducted in a healthcare facility in which N = 22 patients were requested to solve cognitive exercises with the assistance of a robot endowed with CARESSER. The results highlight that the robot managed to keep the patients’ performance stable throughout the sessions, and its assistance eventually matched the therapist’s preferences.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Learning from the Experts: Learning-Based Approach for Robot Personalisation

  • 章节

  • 发表日期：

    2025年

查看全文
  • Antonio Andriella,
  • Carme Torras,
  • Guillem Alenyà

摘要

This chapter introduces two frameworks aimed at achieving the ultimate objective of this monograph: the development of personalised robot-assisted cognitive training (RACT) for Persons with Dementia (PwDs). The first framework employs Reinforcement Learning to learn the robot’s assistive behaviour that best matches the users’ preferences (adaptive) and the concept of Personas based on which therapists can initialise the robot’s behaviour in an easy way by generating patient-persona’ profiles (adaptable). Simulation-based evaluations demonstrate the system’s adaptability to different patient profiles, however, the approach did not fully address all the requirements defined by the therapists. Hence, the second framework called aCtive leARning agEnt aSsiStive bEhaviouR (CARESSER) addresses these limitations. It employs Learning from Demonstration through Inverse Reinforcement Learning to learn a patient-specific policy. CARESSER enables therapists to customise the robot’s assistive behaviour (adaptable) and allows the robot to learn in situ personalised patient-specific policies (adaptive) in an autonomous fashion. An experimental study was conducted in a healthcare facility in which N = 22 patients were requested to solve cognitive exercises with the assistance of a robot endowed with CARESSER. The results highlight that the robot managed to keep the patients’ performance stable throughout the sessions, and its assistance eventually matched the therapist’s preferences.