Maroto-Gómez Marcos, Castro-González Álvaro, Malfaz María, Salichs Miguel Ángel
Systems Engineering and Automation, University Carlos III of Madrid, Butarque 15, 28911 Leganés, Madrid Spain.
Complex Intell Systems. 2023 May 29:1-19. doi: 10.1007/s40747-023-01077-5.
The decisions made by social robots while they fulfill their tasks have a strong influence on their performance. In these contexts, autonomous social robots must exhibit adaptive and social-based behavior to make appropriate decisions and operate correctly in complex and dynamic scenarios. This paper presents a Decision-Making System for social robots working on long-term interactions like cognitive stimulation or entertainment. The Decision-making System employs the robot's sensors, user information, and a biologically inspired module to replicate how human behavior emerges in the robot. Besides, the system personalizes the interaction to maintain the users' engagement while adapting to their features and preferences, overcoming possible interaction limitations. The system evaluation was in terms of usability, performance metrics, and user perceptions. We used the Mini social robot as the device where we integrated the architecture and carried out the experimentation. The usability evaluation consisted of 30 participants interacting with the autonomous robot in 30 min sessions. Then, 19 participants evaluated their perceptions of robot attributes of the Godspeed questionnaire by playing with the robot in 30 min sessions. The participants rated the Decision-making System with excellent usability (81.08 out of 100 points), perceiving the robot as intelligent (4.28 out of 5), animated (4.07 out of 5), and likable (4.16 out of 5). However, they also rated Mini as unsafe (security perceived as 3.15 out of 5), probably because users could not influence the robot's decisions.
社交机器人在执行任务时所做的决策对其性能有很大影响。在这些情况下，自主社交机器人必须展现出适应性和基于社交的行为，以便在复杂和动态的场景中做出恰当决策并正确运行。本文提出了一种用于社交机器人的决策系统，该系统适用于诸如认知刺激或娱乐等长期交互场景。该决策系统利用机器人的传感器、用户信息以及一个受生物启发的模块，来模拟人类行为在机器人身上的呈现方式。此外，该系统还能使交互个性化，以保持用户的参与度，同时适应他们的特征和偏好，克服可能存在的交互限制。系统评估涉及可用性、性能指标和用户感知。我们使用迷你社交机器人作为集成该架构并进行实验的设备。可用性评估包括30名参与者在30分钟的时段内与自主机器人进行交互。然后，19名参与者通过在30分钟的时段内与机器人玩耍，评估他们对“加速”问卷中机器人属性的看法。参与者对决策系统的可用性评价很高（满分100分，得81.08分），认为机器人智能（满分5分，得4.28分）、有活力（满分5分，得4.07分）且讨人喜欢（满分5分，得4.16分）。然而，他们也认为迷你机器人不安全（安全感知得分为3.15分，满分5分），这可能是因为用户无法影响机器人的决策。
