Personalization of a Social Health Care Companion

Thumbnail Image
Issue Date
Journal Title
Journal ISSN
Volume Title
The objective of this thesis is to gain insights into the benefits, limitations, and future directions of personalization techniques in social robotics. The study focuses specifically on Lizz, a digital healthcare companion, within the context of exercise recommendation during rehabilitation. The aim is to contribute to the understanding of this field by exploring the various aspects of personalization in social robotics. Three different prototypes are tested to showcase different mechanisms of personalization: a rule-based system without user feedback, a combination of rule-based and reinforcement learning with user feedback, and a combination of rule-based and reinforcement learning with user feedback and other dynamic factors. The prototypes aim to demonstrate the advantages and disadvantages of different personalization approaches and contribute to improved well-being and recovery outcomes. The findings indicate that greater personalization leads to exercise recommendations that better matches individual preferences and needs. However, it is worth noting that higher levels of personalization require additional effort for the model to reach optimal performance. Future studies could explore the generalizability of personalization approaches in different application domains. Additionally, conducting user studies in real-world settings is necessary to gain deeper insights into practical implementation challenges. Keywords: Socially Assistive Robotics; Personalization; User Profiling; Rule-based systems; Reinforcement Learning.
Faculteit der Sociale Wetenschappen