Pasar al contenido principal

Resultados relacionados

Resultados totales (Incluyendo duplicados): 3
Encontrada(s) 1 página(s)

RUA. Repositorio Institucional de la Universidad de Alicante
Publicación Otros (Other)

Re-engineering EBO: Advancing Social Robotics for Enhanced Care

  • Eslava, Sergio
  • Torrejón, Alejandro
  • Núñez, Pedro
This paper outlines the redesign process of the social robot EBOv2, an evolution of the EBO robotics platform. The original version of EBO has been used in various applications, from promoting computational thinking in educational settings to providing cognitive therapy for older adults. The redesign of EBOv2 aims to meet today’s needs by incorporating state-of-the-art hardware and a new casing. EBOv2 features differential navigation hardware, an upgraded screen for expressive emotional display, microphones and speakers for personalized user interaction, and optimized components to enhance overall performance. The main goal is to improve the user experience and the robot’s versatility for widespread use in occupational therapy, assistance for older adults, and managing neurodegenerative diseases. This article details the co-creation process of EBOv2, focusing on key hardware design aspects, component selection and integration, aesthetic casing design, and comprehensive system integration. Additionally, a group of occupational therapists evaluated the new physical redesign of the robot. This evaluation involved surveys designed by experts in the field to assess the robot’s functionality, user acceptance, and overall impact on future therapeutic outcomes. The results highlight significant improvements in the acceptance of the new robot platform, underscoring its potential for broad application in therapeutic and assistive contexts.
RUA. Repositorio Institucional de la Universidad de Alicante
Publicación Otros (Other)

Discrete Active Inference with Dynamic Scene Graphs

  • Lanillos, Pablo
  • Bustos, Pablo
Active Inference is a cutting-edge theoretical framework that leverages Bayesian inference, control theory, and the free energy principle to explain perception, action, and learning in both biological and artificial systems. This paradigm posits that systems actively infer the causes of their sensory inputs and seek to minimize the surprise between the internal model (learnt by experience) and the actual world state given the observations. However, it usually assumes that the generative model of the world is known—particularly in the discrete domain. Hence, providing grounded and dynamic internal representations of the world can provide the scalability for real world applications, such as robotics. Here, we discuss the potential of integrating dynamic scene graphs, composed of heterogeneous nodes and edges that code parts, objects and individuals in the world with human-shared semantics, with model-based Bayesian inference approaches. We suggest that this type of representation, widely used in Robotics, can be endowed with a kind of dynamics that provides more expressiveness than other solutions such as behaviour trees over fixed ontologies or categories. We present the advantages and the challenges of such a model for robotics through examples framed in real-world use cases.
RUA. Repositorio Institucional de la Universidad de Alicante
Publicación Otros (Other)

Enhancing Cognitive Therapies for Older Adults with EBO: An Autonomous Social Robot System Integrating LLMs and Therapist-in-the-Loop

  • Blanco-Castillo, Antonio
  • Chong Ojeda, Jackie Lee
  • Condón, Alicia
  • Núñez, Pedro
The growing population of older adults necessitates effective cognitive therapies. Social robots have potential as therapeutic tools, providing companionship, entertainment, and support. Traditional methods for cognitive engagement need more in terms of personalization and interaction. This paper presents the development of a system using the social robot EBO for interactive cognitive games and therapies, leveraging Large Language Models (LLMs). The system integrates the narrative generation via Llama on the CORTEX architecture, user response capture with Whisper, and narration through the MeloTTS engine. Tailored for older adults with mild to moderate cognitive impairment, it personalizes interactions based on individual interests, preferences, and cognitive levels. The therapist-in-the-loop approach allows the therapist to participate in the initial game configuration. The system conducts the therapy sessions autonomously, providing a final summary for analysis. This personalized therapy aims to enhance user experience and maximize the therapeutic utility of the system in promoting cognitive and emotional well-being within this demographic group. We evaluated various opensource LLMs and demonstrated their potential effectiveness through a pilot study, showcasing enhanced user experience. This approach could significantly enhance therapies for older adults using social robots as interactive tools.