Evolving Fuzzy logic Systems for creative personalized Socially Assistive Robots

Abstract

Socially Assistive Robots (SARs) are increasingly used in dementia and elderly care. In order to provide effective assistance, SARs need to be personalized to individual patients and account for stimulating their divergent thinking in creative ways. Rule-based fuzzy logic systems provide effective methods for automated decision-making of SARs. However, expanding and modifying the rules of fuzzy logic systems to account for the evolving needs, preferences, and medical conditions of patients can be tedious and costly. In this paper, we introduce EFS4SAR, a novel Evolving Fuzzy logic System for Socially Assistive Robots that supports autonomous evolution of the fuzzy rules that steer the behavior of the SAR. EFS4SAR combines traditional rule-based fuzzy logic systems with evolutionary algorithms, which model the process of evolution in nature and have shown to result in creative behaviors. We evaluate EFS4SAR via computer simulations on both synthetic and real-world data. The results show that the fuzzy rules evolved over time are not only personalized with respect to the personal preferences and therapeutic needs of the patients, but they also meet the following criteria for creativity of SARs: originality and effectiveness of the therapeutic tasks proposed to the patients. Compared to existing evolving fuzzy systems, EFS4SAR achieves similar effectiveness with higher degree of originality.