SAR Interaction by Means of AI-Configurable Adaptation Models
Term Paper by Konstantin Herold submitted in February 2023.
Abstract
Due to the expected increase in the number of elderly people in the coming years, the burden on caregivers and family members is increasing. To alleviate this, social assistance robots (SAR) are suitable, which are introduced and classified at the beginning of the work. This work shows that different limitations may arise due to age-related factors in the use of SAR, which may impair human-robot interaction (HRI). After a basic introduction to software variability, feature models, and rule-based artificial intelligence (AI), researched user requirements are summarized. In order for the SAR to adapt the interaction to the user, it is developed that feature models are first needed, which can store both the user’s limitations and the possible interaction variants offered by the SAR. To find the interaction variant that is suitable for a specific user, both a rule set that logically connects the two models mentioned above and an algorithm that allows the feature model of the interactions to be configured based on their own limitations are needed. Based on the developed requirements, two feature models as well as a rule set could be derived in this work, which were then checked for correctness in the form of a prototype.