Key points are not available for this paper at this time.
Currently, most service robots typically receive and execute commands in a passive manner, which is unsuitable for more meaningful Human-Robot Interaction (HRI). In this study, a Human-Robot Empathy Decision-Making Model (HREDM) of service robots is developed for personal assistance services. HREDM contains the perception, cognition, and decision-making that enables the robot to understand emotion of users and respond appropriately with behaviors that appease or encourage them. First, the SE-ResNet (Squeeze-and-Excitation-Residual Neural Network) is used to recognize and understand users’ facial emotions. Then, a reinforcement learning model is constructed, which enables the robot to actively learn and interact with users by training on their interaction preferences. The proposed mechanism is used to assess the relationship between the robot’s behaviors and the users’ emotions and to make decisions to influence the users positively. The experiment results demonstrate that the proposed model allows the robot to actively learn, analyze, and make decisions based on identified emotions, leading to appropriate calming behaviors. Further, it attained a score of 3.7 in a satisfaction assessment with volunteers.
Zhu et al. (Mon,) studied this question.