Personalizing Multi-modal Human-Robot Interaction Using Adaptive Robot Behavior

Technology favors better life expectancy, changing the population distribution. This change is known as population aging and brings a lack of care professionals for older and impaired people. Social robotics has become a promising alternative to alleviate this problem in recent years, assisting society. Human-robot interaction is a research area related to social robotics that explores designing effective methods for appropriately understanding robots and users. This paper investigates the design of personalized Human-robot Interaction strategies to provide an adapted experience that facilitates vulnerable people’s use of the Mini social robot. The robot uses Random Forest classification to facilitate multi-modal interaction by predicting the most suitable parameters regarding the interaction time, font size, audio volume, answer time, exercise level, and answer channel. The classification problem is solved using a dataset from 240 participants who completed an online survey about their features and interaction modalities, completing activities about audio, visual, and motor skills. The results show the promising classification scores that Random Forest produced for this task and describe the application of the predicted information during the interaction of the Mini robot with an older adult.