Modeling Affordances and Functioning for Personalized Robotic Assistance Alessandro Umbrico , Gabriella Cortellessa , Andrea Orlandini , Amedeo Cesta CNR - Institute of Cognitive Sciences and Technologies Via S. Martino della Battaglia 44,00185, Rome, Italy {alessandro.umbrico,gabriella.cortellessa,andrea.orlandini,amedeo.cesta}@istc.cnr.it Abstract A key aspect of robotic assistants is their ability to contex- tualize their behavior according to different needs of assis- tive scenarios. This work presents an ontology-based knowl- edge representation and reasoning approach supporting the synthesis of personalized behavior of robotic assistants. It introduces an ontological model of health state and function- ing of persons based on the International Classification of Functioning, Disability and Health. Moreover, it borrows the concepts of affordance and function from the literature of robotics and manufacturing and adapts them to robotic (phys- ical and cognitive) assistance domain. Knowledge reasoning mechanisms are developed on top of the resulting ontological model to reason about stimulation capabilities of a robot and health state of a person in order to identify action opportuni- ties and achieve personalized assistance. Experimental tests assess the performance of the proposed approach and its ca- pability of dealing with different profiles and stimuli. 1 Introduction Socially Assistive Robotics (SAR) aims at designing robots capable of continuously assisting users through social in- teraction, supporting their daily living activities (Feil-Seifer and Matari´ c 2005; Tapus, Mataric, and Scassellati 2007). A challenge for SAR is to ensure continuous assistance, facing a large variety of situations and contextualized interactions ranging from, e.g., reminding dietary restrictions and med- ical appointments to monitoring physiological parameters (Mataric, Tapus, and Feil-Seifer 2007). Personalization and adaptability are key features to effectively address specific user needs and achieve good acceptance levels (Moro, Nejat, and Mihailidis 2018; Rossi, Ferland, and Tapus 2017). In our view, personalization is crucial to tailor general assistive capabilities of a robot to the specific needs of a person. Different users may require different types of as- sistance according to specific health conditions. For ex- ample, a user may need a cognitive stimulation or a con- stant monitoring of different physiological parameters, etc. Adaptability is crucial to keep track of the evolving state and behaviors of users. Indeed, health conditions of a patient may change over time and therefore, it is neces- sary to change (i.e. adapt) online the types and charac- teristics of robot assistance. Namely, adaptability allows robots to take into account user feedbacks in order to dy- namically update the user profile (or the robot abilities) and dynamically change the way assistance is carried out, potentially improving its efficacy. Additionally, explain- ability, i.e., the general ability of an artificial agent to ex- plain the rationale behind its choices (Arrieta et al. 2020; Miller 2019), is also crucial for robots interacting with peo- ple. The realization of such SAR systems poses technologi- cal and research design challenges. Our research objective is to realize (autonomous) assis- tive robots endowed with abstract thinking features in or- der to internally represent health needs of an assisted person and contextualize their behaviors by reasoning about their assistive capabilities. To achieve personalization and adap- tation of assistive behaviors we borrow some relevant con- cepts from the literature on robotics and manufacturing and adapt them to SAR. We consider the concept of affordance, widely used in robotics, to enhance flexibility and adapt- ability of robot behaviors (see e.g., (Bozcuo˘ glu et al. 2019; Awaad, Kraetzschmar, and Hertzberg 2015; Yamanobe et al. 2017; Beßler, D. and Porzel, R. and Pomarlan M. and Beetz, M. and Malaka, R. and Bateman, J. 2020)). This concept is generally used to contextualize robot’s capabil- ities with respect to the properties and features of elements (e.g., objects) composing an environment and dynamically identifying opportunities of actions. In SAR, affordances may allow (autonomous) assistive robots to adapt or take advantage of action possibilities that can facilitate assis- tance. In this work we propose an interpretation of af- fordances as situations characterizing opportunities of as- sistance that link health needs of a patient to the capabil- ities of a robot. To support such reasoning features, the capabilities of a robot are here described with respect to general health needs of a person. And we consider also the concept of Function introduced by (Borgo et al. 2014; Borgo et al. 2009) to define a taxonomy characterizing the capabilities of agents in manufacturing domains. Functions are classified according to their effects on the qualities of domain entities (e.g., the color of physical objects). This in- terpretation supports flexible reasoning and pursues a clear separation between the capabilities of an “acting entity” and the concrete implementation (instance) of such an entity. We thus refine this concept to define the capabilities of an assis- tive robot and characterize them in terms of the effects they have on the health state of an assisted person. The above concepts are deployed within a cognitive con- Proceedings of the 17th International Conference on Principles of Knowledge Representation and Reasoning (KR 2020) Special Session on KR and Robotics 917