METHODS published: 14 October 2020 doi: 10.3389/fdgth.2020.545949 “Hi This Is NESTORE, Your Personal Assistant”: Design of an Integrated IoT System for a Personalized Coach for Healthy Aging Filippo Palumbo 1*, Antonino Crivello 1 , Francesco Furfari 1 , Michele Girolami 1 , Alfonso Mastropietro 2 , Giorgio Manferdelli 2 , Christina Röcke 3 , Sabrina Guye 3 , Antoni Salvá Casanovas 4 , Maurizio Caon 5 , Francesco Carrino 5 , Omar Abou Khaled 5 , Elena Mugellini 5 , Enrico Denna 6 , Marco Mauri 6 , David Ward 6 , Paula Subías-Beltrán 7 , Silvia Orte 7 , Ciprian Candea 8 , Gabriela Candea 8 and Giovanna Rizzo 2 1 Institute of Information Science and Technologies of the National Research Council of Italy (ISTI-CNR), Pisa, Italy, 2 Institute of Biomedical Technologies of the National Research Council of Italy (ITB-CNR), Segrate, Italy, 3 University Research Priority Program “Dynamics of Healthy Aging” of the University of Zurich, Zurich, Switzerland, 4 Fundació Salut i Envelliment, Universitat Autònoma de Barcelona, Barcelona, Spain, 5 University of Applied Sciences and Arts of Western Switzerland HES-SO, Fribourg, Switzerland, 6 FLEXTRONICS DESIGN SRL, Milan, Italy, 7 eHealth Unit, Eurecat, Center Tecnològic de Edited by: Catalunya, Barcelona, Spain, 8 ROPARDO SRL, Sibiu, Romania Constantinos S. Pattichis, University of Cyprus, Cyprus Reviewed by: In the context of the fourth revolution in healthcare technologies, leveraging monitoring Ming Huang, and personalization across different domains becomes a key factor for providing useful Nara Institute of Science and Technology (NAIST), Japan services to maintain and promote well-being. This is even more crucial for older people, Charis Styliadis, with aging being a complex multi-dimensional and multi-factorial process which can Aristotle University of Thessaloniki, lead to frailty. The NESTORE project was recently funded by the EU Commission with Greece the aim of supporting healthy older people to sustain their well-being and capacity *Correspondence: Filippo Palumbo to live independently. It is based on a multi-dimensional model of the healthy aging filippo.palumbo@isti.cnr.it process that covers physical activity, nutrition, cognition, and social activity. NESTORE is based on the paradigm of the human-in-the-loop cyber-physical system that, exploiting Specialty section: This article was submitted to the availability of Internet of Things technologies combined with analytics in the cloud, Connected Health, provides a virtual coaching system to support healthy aging. This work describes the a section of the journal design of the NESTORE methodology and its IoT architecture. We first model the Frontiers in Digital Health end-user under several domains, then we present the NESTORE system that, analyzing Received: 06 April 2020 Accepted: 28 August 2020 relevant key-markers, provides coaching activities and personalized feedback to the user. Published: 14 October 2020 Finally, we describe the validation strategy to assess the effectiveness of NESTORE as a Citation: coaching platform for healthy aging. Palumbo F, Crivello A, Furfari F, Girolami M, Mastropietro A, Keywords: e-health, virtual coach, IoT, sensor network, digital health Manferdelli G, Röcke C, Guye S, Salvá Casanovas A, Caon M, Carrino F, Abou Khaled O, Mugellini E, Denna E, 1. INTRODUCTION Mauri M, Ward D, Subías-Beltrán P, Orte S, Candea C, Candea G and We live longer and healthier than ever before. This trend is confirmed year after year by the Rizzo G (2020) “Hi This Is NESTORE, Your Personal Assistant”: Design of an proportion of people aged over 65 populating our countries. The prolongation of the life expectancy Integrated IoT System for a is also followed by an increase of the expected quality of the life (1). People aim at remaining active, Personalized Coach for Healthy Aging. healthy and autonomous as much as possible. Such expectations, however, represent challenges Front. Digit. Health 2:545949. hard to meet. In this regard, the e-Health systems have the main responsibility to ease such doi: 10.3389/fdgth.2020.545949 transition, also referred to as healthy aging . Frontiers in Digital Health | www.frontiersin.org 1 October 2020 | Volume 2 | Article 545949 Palumbo et al. The NESTORE IoT Personal Assistant However, healthy aging involves several domains. Physical, The rest of this work is organized as follows. Section 2 shows mental, emotional, and cognitive are all areas that need to be the background of NESTORE with related work and initiatives. considered during the design of ICT-based solutions for older Section 3 explains the core coaching mechanism and its areas adults. Most of the commercial solutions available on the market of intervention. Section 4 details the software and hardware are designed to address only a subset of such domains. This is infrastructure supporting the overall system. The validation the case of apps for tracking the sport activities or for improving of NESTORE is detailed in section 5, while section 6 draws the mental skills. In such cases, users are required and motivated the conclusions. to increase only some skills to our knowledge, only one solution tackles the healthy aging as an holistic process in which all the domains have to be properly trained day by day. This 2. BACKGROUND AND RELATED WORK example is provided by the Matilda system (2, 3) based on a In the last decade, a growing number of research projects and robot for older adult care in residential facilities. It proposes commercial products have dealt with the development of e- different coaching activities for each domain but it doesn’t rely Health solutions to promote and to support well-being with a on a multi-domain coaching model and the proposed activity focus on the so called “silver economy” scenario (5). are not based on international guidelines. This work describes Several EU funded projects have worked on ICT solutions NESTORE, a companion for older adults, whose main goal is to to early detect risks in different aspects of older people’s lives. collect and analyze data from the end-users and to return to them For example, the main ambitions of the EU H2020 My-AHA personalized coaching and feedback. NESTORE is a European project2 are the early risk detection and intervention in order to project, funded under the Horizon 2020 programme, based on support healthy aging both in the physical and cognitive domain. a consortium of 16 partners1 . The project relies on the deployment of Ambient Assisted The innovative idea of NESTORE is to face healthy aging Living (AAL) sensors, wearable devices, and smartphones, and by simultaneously taking into account four key domains of through big data analysis is able to engage users in improving well-being, shaping the human life in terms of both status and their lifestyle both in a short-term and long-term vision. behavior: physiological status and physical activity behavior, In the same context, EU H2020 projects GrowMeUp3 and nutrition, cognitive and mental status, social behavior. To deal Radi4 provide integrated and services through robotic-based with the complexity of aging changes, NESTORE coaching approaches. These two projects mainly monitor the behavioral focuses on one specific aspect for each well-being domain: trend of a healthy older population. The main goal is to support physical activity, nutrition, cognition, and social relations. In and to encourage older persons to stay active longer. PreventIT5 this work, we describe our vision of personalized coaching and REACH6 are focused on monitoring the users’ physical with a set of comprehensive application scenarios as well as activity through the deployment of wearable and/or ambient the design of the NESTORE IoT-based architecture used to sensors. Additional European research projects on the coaching coach the end-users. Our solution firmly relies on unobtrusive of the elderly which feature common elements to NESTORE sensing devices able to collect data from different domains, for are: The CAPTAIN System (6) providing a projected and example, tracking physical activity, detecting nutrition elements, tangible interface; COACH Council of Coaches (7) with multiple assessing sleep quality (4) as part of the physical domain or autonomous, embodied virtual coaches; HOLOBALANCE (8) recognizing social interactions with a fine-grained temporal with their new personalized hologram coach platform. resolution. In turn, such information is analyzed by a Decision The NESTORE project is intended to overcome the limitations Support System (DSS) able to deliver, through the Virtual Coach, of vertical solutions (single-domain oriented works) with personalized feedback to the NESTORE users. The design of a strong holistic approach by addressing several singular NESTORE follows a structured methodology. First, we define domains (cognitive, physiological, social, and nutritional) (9– a physiological, behavioral, and psychological model of the 11) and by considering these domains from a user-centered prospective older adult user through the involvement of domain approach. Furthermore, in order to support healthy aging, experts. Second, we identify those markers to be monitored the NESTORE ecosystem provides a tangible interface to through IoT-based devices. In this vein, it is worth noticing the improve human-machine interactions and the acceptability of design of a custom hardware for detecting some physiological the proposed solution. markers as well as the use of commercial devices. Finally, we From the commercial point of view, several products have evaluate the key health and well-being markers both on the been presented in the last years to hit the market of e- short- and long-term period in order to provide the final users Health coaching systems. In general, these solutions require a with specific coaching plans. The NESTORE approach is under human-machine interaction through tangible objects enriched evaluation in 3 pilot sites across Europe: Spain, The Netherlands, with artificial intelligence capabilities or through smartphone and Italy. Each site involves 30 (20 in the experimental and 10 in a applications. The most important interface that a tangible control condition) individuals recruited on a voluntary basis and followed by experts during the NESTORE coaching intervention. 2 http://www.activeageing.unito.it/en/home 3 https://cordis.europa.eu/project/rcn/194088/es 4 http://radio-project.eu/ 1 H2020 call: H2020-SC1-2016-2017, topic SC1-PM-15-2017, partners from: 5 http://www.preventit.eu/ Belgium, Italy, Netherlands, Romania, Spain, Switzerland, United Kingdom. 6 http://reach2020.eu/ Frontiers in Digital Health | www.frontiersin.org 2 October 2020 | Volume 2 | Article 545949 Palumbo et al. The NESTORE IoT Personal Assistant object presents is based on voice interaction (e.g., Amazon 3. A PERSONALIZED COACHING Alexa, Google Speech, IBM Watson Conversation, Microsoft EXPERIENCE Bot Framework). The voice interaction is generally well- accepted by older adults and increases the accessibility of the The NESTORE system is mainly aimed at providing multi- interfaces for people who start to have perceptual impairments domain coaching plans specifically tailored to the users’ needs (12). On the other hand, smartphone-based applications offer in order to maximize their engagement and health benefits. advice and coaching activities through messages, notification, NESTORE is focused on physical activity, nutrition, cognition, reminders, and tips. For example, the growing market of and social relations, which are composed by specific subdomains, coaching systems for athletes (e.g., runners, bikers) contains in order to consider, as much as possible, the factors that must several applications able to give personalized suggestions to be monitored and the abilities that must be strengthened during the user in order to improve their performance or to set physiological and psychological aging (Figure 1). The NESTORE up a training schedule (13). The main limitation of these users, after a preliminary phase of assessment which is intended approaches is that they are usually focused on a single to characterize the subject’s status and behaviors using ICT tools, domain. However, addressing human well-being and promoting are invited to select personalized coaching pathways in order to the independence of the older population is inherently a improve or maintain their skills in each specific domain. multi-domain challenge. Once the assessment is performed, different “pathways of well- being” are set up as “tracks” within users make their journey experiencing the NESTORE coaching system (17). A pathway 2.1. Requirements for an E-Coaching of well-being is the process of pursuing a high-level goal to Intervention which the user will commit at the end of the motivational phase. With respect to traditional e-Health interventions, a system Since well-being is intrinsically multi-domain, a pathway spans that monitors user’s behavior and provides personalized across multiple areas (e.g., I want to maintain my everyday suggestions to improve health-related outcomes can be called mental skills while improving my physical activity and nutrition). virtual coach. Such a coach can be simply embedded in Each domain is quite complex and training every aspects of smartphone devices (e.g., app), but can also have a more it could be overwhelming or practically impossible. For this anthropomorphic and physical embodiment, for example in reason, we divided each domain into sub-domains (e.g., the form of an Embodied Conversational Agent (ECA) or a robot. Nutrition domain is divided in “Decrease body weight,” “Achieve As defined in the introduction to a recent IEEE Computer a healthy diet,” and “Maintain muscle mass”). Therefore, a special issue on E-coaching for Health (14), coaching a user pathway could be seen as a selection of sub-domains based means to “frequently, but not continuously, observe, listen on system recommendation (based on the user’s status) and to, question, understand, reason with, teach, and/or advise user preferences. Once the user chooses a pathway, a set of the users in order to change their behavior and to improve coaching activities related to the pathway is proposed. A coaching their health.” To this purpose, they continue, “intelligent activity is a time-bound activity that can be scheduled in the systems are used to encourage progress toward specific health- personal calendar (thus helping the user in planning the activity related goals by providing tailored training and guidance.” and sticking to the plan). The activity can be composed of While in (15) are suggested 9 important features that an a set of specific exercises (e.g., flexibility exercises), or can e-coach system should have (i.e., social ability, credibility, be proposed as an everyday activity (e.g., gardening with a context-awareness, learning abilities, data gathering, proactivity, friend, or buying vegetables in a supermarket). In both cases, reflection, behavior change model integration, planning instructions for carrying out the activities and maximizing the support), authors in (16) reviewed the key components that impact in the involved domains are provided. Coaching activities can significantly affect a variety of health outcomes, the are defined by the experts in each of the NESTORE coaching adherence and the usability of an e-coaching intervention. domains. They can be defined as structured activities (i.e., specific The following behavior change techniques were found to exercises carried out to train the user in a specific domain) positively affect both health outcomes and usability in the or non-structured activities (i.e., everyday activities that can studies reviewed: be conducted as part of the daily routines but that can still contribute to the improvement of well-being in one or more • Setting short-term goals to eventually reach long-term goals; domains). The unstructured activities are intended to avoid • Personalization of goals; forcing the user on performing unwanted activities; thus they • Praise messages; are chosen from a list of actions (dancing, walking with the • Reminders to input self-tracking data into the technology; dog, etc.) that the user likes to perform during his/her everyday • Use of validity-tested devices; life and are suggested by the system based on the specific • Integration of self-tracking and persuasive e-Coaching; user’s choices, also considering the feasibility of performing • Provision of face-to-face instructions during implementation, the activity itself in a specific place and moment of the day. as key components for influencing both health outcomes and However, if the user does not meet the prescribed coaching usability in a positive way; recommendations with the only unstructured activities, the • Provision of personalized content. system proposes also structured activities (for example running Frontiers in Digital Health | www.frontiersin.org 3 October 2020 | Volume 2 | Article 545949 Palumbo et al. The NESTORE IoT Personal Assistant FIGURE 1 | The NESTORE coaching domains and subdomains. or walking in the case of aerobic physical activity) to ensure as walking, running, or cycling to improve cardio-respiratory that the required prescription, as defined by the international fitness; (ii) strength exercises, such as chair squats or knee push- guidelines, is reached. While pathways will be defined in the up to improve upper and lower limbs strength and power; (iii) system as a list of coaching activities, such list can be customized flexibility training, using stretching exercise for upper and lower by the system according to user status, context and preferences. limbs to improve the body range of movement; and (iv) balance For example, regarding the physical activity, the user can select exercises, such as one leg stand or toe-to-line to improve the the weekly frequency of the sessions and the intensity level so overall body stability. The NESTORE coaching system works by that the duration of the daily session is automatically adjusted, in scheduling structured sessions in the user’s agenda; the structured order to meet the needed prescription (18). After the completion activities are individualized according to results obtained by some of the activity, the user can review his/her personal trajectory standard physical performance tests, which are performed in the in the system. Eventually, the system might have recognized a monitoring phase and every 4 weeks. Two main pathways can be part or all the training activities performed. The user can still followed according to users’ needs of retaining or improving their modify what was recognized by the system or add additional physiological status. A score system assigned to each training information (self-reporting). activity will allow keeping track of user progress and personal The NESTORE system includes a plethora of sensors and goal attainment for the selected pathway and in all the involved other data sources (wearable devices, environmental sensors, domains. Such scores can be shown to the user or not, according questionnaires, APIs), which measure data directly from the users to his/her preferences. and process it to extract short-term and long-term indicators. To be compliant to the suggested coaching plans, the user The raw data, as well as the indicators, are then elaborated by should reach a total weekly score, by performing structured a DSS to profile the users using static and dynamic variables. The and/or non-structured activities. The overall weekly score for the users’ profiling is fundamental to propose personalized coaching domain is 100 points weighting every single sub-part considering plans and performance assessment. Finally, the users interact its importance for the subject quality of life and execution of daily with the NESTORE coaching interfaces (tangible, chatbot, life activity (30 points for aerobic training; 50 points for strength; APPs, serious games, social platform), in order to receive clear 10 points each for both flexibility and balance). Each activity has instructions and feedback to perform the prescribed activities its own score that is calculated considering the impact of that composing the whole coaching plan correctly. A schematic specific activity in the related well-being domain. For example, representation of the NESTORE coaching system is shown 10 min of aerobic activity performed with moderate intensity in Figure 2. corresponds to 2 points for aerobic training. Similarly, dancing for 1 min corresponds to 2 points for the aerobic training, 3.1. Physical Activity sweeping corresponds to 2 points for aerobic training and 0.33 The physical activity coaching is focused on four sub-parts: for strength, whereas walking with the dog corresponds to 1 point (i) aerobic training, consisting of structured activities, such for aerobic training and 0.33 points for strength. The different Frontiers in Digital Health | www.frontiersin.org 4 October 2020 | Volume 2 | Article 545949 Palumbo et al. The NESTORE IoT Personal Assistant FIGURE 2 | The NESTORE coaching system. components (i.e., structured and non-structured activities) add recommended to celiac people) and their nutritional habits. The up to generate the overall score. This implies that the more non- DSS will schedule grocery store outings with the ingredients structured activity a user performs the less structured activity is that will be needed in the upcoming days. Those users that suggested to the user by the system. This choice is intended to track their food intake periodically will receive more customized make the system less intrusive as possible concerning the daily recommendations that those that do not. life usual activities of the users. In order to assess if the user is performing his/her physical 3.3. Cognition activity at best, the wearable device embedded in NESTORE is In the cognitive domain, the NESTORE coaching consists of able to monitor the intensity of the structured activity during three plans, (1) one focusing on memory (involving a structured the whole exercise session and to provide a real time feedback traditional cognitive training task in the domain of working (vibration) to the user if the heart rate is too low or too high. memory), (2) one focusing on broader thinking skills (involving Furthermore, at the end of the training session the system a multi-domain serious game that involves simultaneous tasks in determines the adherence of the user to the prescribed plan. the areas of visuo-motor ability, spatial memory and inhibition), Additionally, to avoid an excessively strenuous training, the and (3) one that targets everyday cognitive functioning (involving system, after each session, asks the user to provide the Borg Scale productive intellectual engagement in the form of unstructured and the Total Quality of Recovery scale as indicators of perceived leisure activities that require learning a novel and complex fatigue and recovery. Based on these user feedback, the system skill, such as digital photography, dancing, etc.). The DSS will can then adapt the intensity and the duration of the next exercise recommend coaching in the cognitive domain based on pre- session in order to maintain the weekly correct prescriptions. defined performance thresholds in the cognitive tasks assessed at baseline. Those users whose cognitive performance is above the 3.2. Nutrition critical threshold will be given the recommendation to keep up The coaching plans for the nutrition domain are structured as their current lifestyle concerning activities as is so as to maintain follows: (1) body weight management, focused to increase or their cognitive functioning. Those individuals below a critical decrease body weight, and consisting of tailored dietary activities, threshold will receive the recommendation to engage in either as well as energy balance monitoring, (2) body composition one of the three cognitive pathways. Performance thresholds are management, focused to increase muscle mass or decrease body defined based on the typical thresholds used in adaptive testing fat, and consisting as well of tailored dietary activities and in cognitive training research, where an accuracy below .6 leads energy balance monitoring, and (3) healthy diet, focused to to a decrease in task difficulty and an accuracy above 0.8 leads to improve their dietary habits by targeting macro-nutrients and an increase in task difficulty. Accordingly, we have chosen to not micro-nutrients intake through both diet monitoring (based on provide a recommendation for additional training if individuals automated food image recognition) and nutrition coaching. The score beyond 0.8, to provide a first degree of recommendation if DSS will suggest one of the aforesaid coaching plans based on the between 0.8 and 0.6 and a strong recommendation for individuals anthropomorphic measures of users (i.e., weight, height, BMI, below 0.6 accuracy. The tests used at baseline are state-of-the- fat free mass, muscle mass, and waist circumference) and their art psychometric cognitive tests intended for the assessment nutrient intake evaluated at baseline. Once the coaching starts, of cognitive capacity in the domain of working memory: N- users will receive recommendations of food to include in their back test and numerical updating) in addition to a validated diet, recipes to cook, and good practices to follow taking into questionnaire on cognitive failures for older individuals with no account the foods they refuse (e.g., none gluten food will be known diagnosis of cognitive impairment. Frontiers in Digital Health | www.frontiersin.org 5 October 2020 | Volume 2 | Article 545949 Palumbo et al. The NESTORE IoT Personal Assistant FIGURE 3 | Overall view of the NESTORE integrated IoT architecture: (A) shared functional modules and (B) used technological components. 3.4. Social Activity Shared modules (components) are referred as: Identity In the social domain, the NESTORE system will target two areas: Manager, Security Manager, User Profile, and Log Service that are (1) The first focuses on (the improvement of) social integration used by the platform and its components. The Identity Manager by suggesting users to join group-based activities in their Module is necessary for end user identities management across community in a domain that matches their personal interests the platform components, APIs, the cloud, mobile, and sensors, (e.g., sports, culture, education, community service). (2) The regardless of the standards on which they are based. On the second pathway targets social skills and their improvement, other hand, the platform exposes a wide range of data, which such as a wide range of communication techniques (face-to- are subject to security and privacy concerns like: user data, face, computer skills in order to better use email or video sensor data, actuator access, service membership. Each of these calls, tolerance of silence in conversations, giving compliments). areas must offer a proper security level which is audited for This pathway requires to join structured classes in the user’s compliance with existing rules. As shown in Figure 3B, the main community and is currently not something that is offered as a components of the NESTORE ecosystem are identified, starting class within the NESTORE system. Given that through a person with: (a) End Users, they will interact with the ecosystem through joining a new group-based activity communication skills are their smartphones and by using web applications; (b) NESTORE also trained, the NESTORE system will primarily focus on the cloud, it hosts all the back-end parts and oversees retrieving, first social pathway. Recommendations for the social domain elaborating and storing the data coming from the platform in general will be given by the DSS based on social integration users and other third-party applications; (c) The End Users are and loneliness screening measures assessed at baseline. Those using the NESTORE applications on their environment/their participants who are well-integrated and not lonely will not house by accessing specific coaching applications (in form of receive a particular recommendation in the social domain Web, smartphone or wearable applications) that are running on other than to keep as socially engaged as they currently are. different devices (tables, phones, wearable) accessing Internet; Two standard questionnaires to assess social integration and (d) The environmental sensors are connected to the NESTORE loneliness with known thresholds are employed at baseline, the platform using WiFi/3G connection, using the IoT/WoT (Web De Jong Gierveld Loneliness Scale (19) and the Lubben Social of Things) approach. Network Scale (20). Both scales have published thresholds for the The underlying integrated software infrastructure provides identification of at risk (for loneliness) individuals and are widely the basis for building the NESTORE system as a whole. used in gerontological research. From a technological point of view, it can be seen as an ensemble of software and hardware modules that address the functional requirements of the different domains of the coaching 4. THE INTEGRATED INFRASTRUCTURE experience. The details of the particular non-core solutions, such as sleep monitoring (21), social activity (22, 23), indoor behavior The NESTORE project relies on a complex IoT architecture analysis and environmental monitoring (24), can be found in the whose goal is twofold: to retrieve sensing information from referenced works, while the core functionalities of NESTORE will users and to analyze the data collected so that valuable be described in the following subsections. feedback can be given to users. This section describes the Data are initially collected from user’s personal space with NESTORE IoT architecture that integrates three interconnected a sensor kit following the IoT paradigm (section 4.1). Each kit and complementary components essential for the e-Coaching comprises commercial and custom devices able to sense the activity: the Monitoring System, the DSS and the Virtual Coach environment and to send the data to the back-end through a (Figure 3A). These components provide a robust solution for mobile gateway (section 4.2). The raw data collected are then easily adding and integrating 3rd party applications. elaborated on the cloud so that to extract data useful for the DSS Frontiers in Digital Health | www.frontiersin.org 6 October 2020 | Volume 2 | Article 545949 Palumbo et al. The NESTORE IoT Personal Assistant FIGURE 4 | The NESTORE IoT architecture: (A) server-side components and (B) the Web of Things model. (section 4.3). The DSS has the main role to feed the Virtual Coach Connectivity and message routing are implementing the which implements the interface with the end-users (section 4.4). needed protocols for the communication with the IoT devices. AMQP (Advanced Message Queuing) protocol (26) is used, being a protocol widely used for the type of middleware 4.1. The IoT Architecture applications, that implements routing rules, stores messages, An IoT solution is characterized by the connection of several and sets their distribution rules, allowing exchange messaging devices (i.e., things) which can use a gateway to communicate between clients and the server. In the NESTORE context, we on a network to back-end servers running the IoT platform to identified that data must be transferred to the cloud for the integrate the information. The roles of the devices, gateways following scenarios (27): (a) Sensor to Device using different and cloud platforms are well-defined; each provides specific protocols and transmission/transport methods involving the features and functionalities needed for creating a robust IoT Open Systems Interconnection (OSI) level 1 to 4; (b) Device to solution. The NESTORE architecture supports different types Cloud and Sensors to Cloud using HTTP/Web Socket approach of IoT devices that can connect and push data into the cloud, that implement the OSI level 7 Application layer. A special from Sensors to Applications as defined in (25). Security methods device is represented by the gateway that enables machine-to- are implemented from the devices to the cloud. Features such machine communication. All data moving to the cloud, or vice- as authentication, encryption, and authorization are part of the versa, go through the gateway, which can be either a dedicated solution stack. hardware appliance or software program. The gateway provides A high level overview of the proposed architecture is depicted a place to pre-process data locally at the edge before sending in Figure 4. The device registry component ensures the definition it on to the cloud. When data is aggregated, summarized, and of the device categories and the data type that are supported tactically analyzed at the edge, it minimizes the volume of by each one. In order to assure data analytics and data data that needs to be forwarded on to the cloud, which can interoperability for the device data, the format and description have a big impact on response times and network transmission of data must be taken into account from the beginning. Thus, costs. REST web services are used to access the IoT system. a definition of the ontologies and metadata across heterogeneous The proposed architecture defines and implements a common domains is a key requirement. Device management is responsible strategy Application Programming Interface (API) for managing for the enrollment of each “thing” with the system and managing the IoT platform and for acquiring data from various devices. each device instance. After a successful enrollment a “thing” can Through APIs, it is possible to send data from devices (e.g., communicate (send/receive) data with the IoT system. The data sensors, mobile apps, etc.) to the cloud by Web of Things devices management module handles data about how the user interacts or gateways. Reusing and adapting patterns commonly used for with the “thing” and it traces its usage among different users. It is the Web, we enabled the IoT architecture as Web of Things, assuring data security by separation of data generated by the same using Web servers (28) on smart things and applying the REST “thing” for different users on different time intervals. To take architectural style (29, 30) to the physical world. The essence of full advantage of the IoT system, the event and analytics module REST is to focus on creating loosely coupled services on the Web ensures the possibility to define, manage, and use events and basic so that they can be easily reused (31). analytics by integration of the well-known Grafana engine7 . A device to be connected to the IoT platform needs to implement a Web of Things Agent (wotAgent) that transforms a 7 https://grafana.com/ sensor or a device into a compatible WoT element. The wotAgent Frontiers in Digital Health | www.frontiersin.org 7 October 2020 | Volume 2 | Article 545949 Palumbo et al. The NESTORE IoT Personal Assistant can be configured at the Cloud level when the wotAgent support battery level status is signaled with a specific alert notification OSI level 3 and 4 and it translate the data to the cloud from and a buzz. Charging station takes about 5 h to completely the sensors and devices. When it is configured to run on a recharge the wristband battery and during the recharging process mobile device on top of an operation system (i.e., Android) or an animation is showed on the wristband display to inform the transforming a device/sensor into a smart device then wotAgent user about the charging progress. implements locally the OSI level 1 to 4 and it communicate with BLE beacons are deployed in the user’s environment to detect the IoT cloud using REST. Figure 4B shows the wotAgent model social interactions among NESTORE users and their relatives adopted in NESTORE. (bringing with them keyfobs equipped with mobile BLE tags) with their duration, function, location, and number. We exploit 4.2. The Monitoring System the capability of calculating the proximity between BLE devices The kernel of the NESTORE system is developed around a from Received Signal Strenght Indicators (RSSIs) (32–34) also wearable and environmental devices. The wearable device and for detecting the interaction of the user with pieces of furniture its charging station are part of the multi-sensor monitoring in the house. We deploy fixed beacons that give us insights system of NESTORE representing a multi-parameter observatory on the users level of sedentariness (35, 36). We customized and data recorder in the form of a wristband together with the NESTORE BLE beacons deployed in the house to advertise Bluetooth Low Energy (BLE) beacons. The devices communicate additional information like motion (to increase the level of via a BLE chip with both wearable and environmental beacons accuracy in detecting interactions with point of interests in the as well as smartphones and tablets. The wristband can work house), temperature and humidity [to calculate the indoor air for about 18 consecutive hours considering a standard use case quality indicator (37)]. In this context, the wristband also acts expected by Nestore pilot. Battery level status is always visualized as a hub for the beacons (W for wearable beacon and E for qualitatively with an icon on the wristband screen and the low environmental beacon). Function and features for the wristband and beacons vary and are summarized in Table 1. The wearable device is recharged with a dedicated wired battery charger while the beacons have replaceable coin-cell type batteries as shown TABLE 1 | Functions and features of wristband/beacons. Figure 5. Function and feature Wristband Beacon As with all system engineering driven projects, the starting point is the definition of the user, user journey, usage User steps counter X environment and stakeholder needs and relevant requirements Distance estimation X (38). Since NESTORE is a human-engineered system, the first Activity type recognition (i.e., “no activity,” X batch of requirements concerns the overall business mission of walking or running) the system described in the business requirements specification Sedentariness monitoring X (BRS). This is followed by stakeholder requirements (StRS) Stairs (Up and Down) X and concludes with systems (SRS) and relevant sub-system Energy expenditure X requirements. The BRS represent the business intention of RSSI W and E NESTORE (denoted “enterprise”) that connects the external Battery level monitoring W and E environment or market needs or the Concept of Operations Motion detection E only (ConOps) while the subsequent StRS and SRS represent Temperature monitoring E only the operations concept (OpsCon) (39). The opening high- Humidity monitoring E only level project business, stakeholder and system requirements FIGURE 5 | The NESTORE monitoring artifacts: (A) the wearable device; (B) the charging station; (C) the social and environmental beacons. Frontiers in Digital Health | www.frontiersin.org 8 October 2020 | Volume 2 | Article 545949 Palumbo et al. The NESTORE IoT Personal Assistant TABLE 2 | Business (BRS), Stakeholder (StRS), and System Requirements data sources encompassing the four target domains. Once this Specifications (SRS). data is processed, it is inputted into the DSS so a customized BRS StRS SRS coaching intervention can be designed for each user. The DSS is comprised by multiple modules: Business model of Wearable bracelet Wristband status-diagnostics: NESTORE • Profiling individuals: A complete profile of users is built based 1. Sensor status on the extracted indicators, which include demographic and 2. Charging status environmental data, information about their preferences and 3. Battery level status habits, and data describing their nutritional, social, physical, Development Easy-to-use and small Wristband alerts: and cognitive status and behavior. A user’s profile is translated challenges wearable device 4. Low battery into tags that capture the information that will be used to 5. User sedentariness personalize the given coaching. 6. Wristband “not worn” • Pathway suggestion: Rule-based Reasoning (RBR) is used 7. Connection needed to assess users’ status in relation to all pathways. The Development time and Non-invasive Wristband data availability: standardization of the thresholds between pathways comes cost 8. Non-structured activities with a three-level scale to measure the users’ status per 9. Sedentariness periods pathway. 10. Social interactions • Coaching and the tagging system: After users choose the 11. Environmental interactions pathways to focus on, the tagging system comes into play by 12. Structured activities proposing suitable Coaching Events (CEs) to users. 13. Evaluation tests 14. Device error logs These three modules use AI techniques covering a wide Type and number of Does not complicate User characteristics: range of methodologies: regression analysis to infer the daily users user everyday life 15. Age routines of users; statistical modeling to measure the similarity 16. Gender among users; RBR to select the most adequate pathway to recommend (44). 17. Weight Personalization in NESTORE has the following dimensions: 18. Height tags to describe users and CEs, ratings, and history analysis. 19. Resting heart rate Tagging is the process of assigning meta-data to content in 20. Maximum heart rate the form of keywords. Users’ profiles are tagged automatically 21. Language preference, etc. thanks to a given ontology and expert driven-criteria. Tags Innovation brief Mobile app provision 22. Battery duration together with ratings are inputted into the tagging system, which User experience-user Ease of 23. Battery recharging time journey maintenance/installation employs recommendation techniques to tailor coaching events to users. History analysis is meant to assure diversity in the User needs and user Data security 24. Display size and type definition recommendations given. The analysis provides a further level of understanding of the status of the final users, which helps in personalizing all the interaction with the coach. As depicted in Figure 8, the DSS is fed by different Public Cloud Services as specifications are simplified and summarized in Table 2. In the LogMeal API and the information coming from the wearable deriving the user perspectives concerning: devices and sensing units through the IOT SubSystem. They can • Identification of needs, values and suggestions for co-design; be grouped as shown next: • Transferability of participants’ perspectives to technologists; • Sensing system: It is composed of a wearable device and • Co-design user-end participation for prototype improvement; an ensemble of environmental wireless devices that detect these were provided by the relevant co-design specialists involved the status of the users’ living space and their behavior. in the project and then translated into an overall Industrial Those include a smart scale, a BCG for sleep quality, and Design (ID) framework. This led to the generation of three BLE beacons. wristband concepts and a charging station as shown in Figure 6. • Food recognition: The LogMeal API is used to automatically To complete the NESTORE monitoring system there is build a food diary based on images captured by the users. also a smart scale to collect anthropometric, musculoskeletal • Questionnaires and tests: A psychological model is built based characteristics, and balance (40) and a ballistocardiographic on the social and cognitive profile of users. These forms are (BCG) system in order to perform sleep monitoring (41–43). The also used to describe users’ preferences and likings. NESTORE system and the outline of the device interactions is • Serious games: Cognitive and physical exercises are suggested illustrated in Figure 7. through game-based initiatives to foster well-being in an entertaining way. 4.3. The Decision Support System • Context: Personalization is improved by adapting the coaching The DSS can be considered the intelligence of NESTORE (18). Its events (CEs) to users’ daily life contexts, e.g., by using aim is to provide tailored coaching to users based on integrated meteorological or location-related data. Frontiers in Digital Health | www.frontiersin.org 9 October 2020 | Volume 2 | Article 545949 Palumbo et al. The NESTORE IoT Personal Assistant FIGURE 6 | Graphical ID framework translation of user input from co-design specialists report: (A) wearable device and (B) charging station. FIGURE 7 | Overview of the monitoring system architecture. 4.4. The Virtual Coach produce an intervention aimed at improving the user’s behavior The NESTORE project aims at presenting a new concept of considering all the peculiar characteristics of that specific an e-coach, which consists of an artificial intelligence that is individual. The NESTORE e-coach is able to communicate able to monitor the user’s behavior through an extensive and with the user via multiple multimodal interfaces, dynamically heterogeneous network of sensors and devices, and that can adapting the interaction to the type of content, the contextual Frontiers in Digital Health | www.frontiersin.org 10 October 2020 | Volume 2 | Article 545949 Palumbo et al. The NESTORE IoT Personal Assistant FIGURE 8 | Overview of The DSS architecture. situation and the user’s current activity. The e-coach interacts smart bracelet) or from self-reporting. Finally, the mobile app with the user using different modalities in a context-aware provides the possibility to accept and schedule in the calendar manner. For example, Jana, who is 68 years old and retired wants the activities of the pathways proposed by the DSS. The mobile to increase her social integration. The e-coach then proposes a application is a convenient way to invite people to join and share list of fitting activities via the chatbot, where she can ask about an activity. If the invited person is not yet part of the NESTORE all the details of the proposed activity list. Moreover, the mobile ecosystem, she/he will receive an e-mail notification and the app allows to create an event choosing the preferred parameters, possibility to register to NESTORE through the Social Platform. directly sharing with her network and creating an appointment Concerning the Social Platform, its main role is to connect in her calendar. Once the invitations to her social network are NESTORE’s users with other users and stay in touch with friends sent, Jana can go back to her daily routine and for this reason she and family members. This is done through classic tools such as activates the tangible interface with a gesture. While cooking, she a forum and blog accessible via a simple web interface. Posts can casually chat with the tangible interface. The vocal interaction are regularly added to provide additional tips and tricks related is quite convenient while the user is occupied in other tasks to the four domains and use of NESTORE itself. In addition, that require her attention. Jana can ask the e-coach through the through the “Events” page, the social platform shows geolocalized tangible interface if any of her friends have already accepted her information about nearby events and activities that may interest invitation. Once she will be done eating, she can prepare to go to the users. These events may be added by partner associations or bed and with a simple gesture she can put the tangible interface by the users themselves. in sleep mode. Strictly interconnected with the mobile app, the chatbot Within NESTORE’s mobile app, the interaction with the provides an easy and intuitive interface to obtain personalized user is mainly managed by five interconnected interfaces: the support (e.g., reminders, encouragement messages), receive application, the social platform, the chatbot, the serious game, coaching advice (e.g., instructions and tips), and enter personal and the tangible coach. The mobile application represents the information and preferences (e.g., during the registration main entry point for the user to interact with the NESTORE process). The chatbot is also used to get a feedback from system. Its main role is to provide feedback on the different the user after an activity or to help him to comply self- domains of the current pathway and to help schedule the reporting tasks. A notable example is the possibility to take activities. Once logged in, the app presents a quick overview on pictures and log the current meal directly during a chat (see the progressions made in the current pathway. The user can then Figure 10). The questionnaires used to evaluate the user’s navigate to get a more detailed view of the achievement and status in the different domains are also administered via stats in all four domains. Each of these views has a specialized the chatbot. In this case, the interface changes according interface developed with the support of the domain experts. to the needs of the questionnaire (Likert-like scale, multiple These views show several dimensions of the user’s temporal choice answers, open questions, etc.) to provide the optimal evolution (daily, weekly, and/or monthly). For instance, the response format. social view (Figure 9) shows stats about perceived loneliness, the In parallel with the app, a serious game has been developed type of social interaction having taken place (by telephone, in to foster users’ motivation (45). The serious game tackles person, etc.) and the time spent with people of the user’s “local specific subdomains such as strength, balance, and flexibility social circle” (i.e., family and friends taking part in NESTORE). training for the physical domain or spatial navigation, inhibition The interface clearly states whether these values come from a and visuo-motor functioning for the cognitive domain (46). monitoring device (e.g., from the WOT agent connected to the This is done thanks to an in-game virtual character (called Frontiers in Digital Health | www.frontiersin.org 11 October 2020 | Volume 2 | Article 545949 Palumbo et al. The NESTORE IoT Personal Assistant FIGURE 9 | Feedback on the Social domain via the app interface. Nestor) leading the user through a series of mini-games and and to display different types of information. Therefore, the guided exercises. Serious game sessions are suggested by the horizontal position has been designed considering that the user DSS and added to the user’s calendar similarly to any other can see the LEDs on the base in order to receive complementary coaching activity. information. The behavior of the tangible interface intends to The tangible interface is connected to the same modules that be life-like in order to establish an affective relationship with manage the chatbot. This allows implementing a conversational the user for a long-lasting bonding. The enclosure is covered agent with the same characteristics and behavior of the with a soft textile to provide multisensorial experience while chatbot but based on vocal interaction. This solution allows interacting with it, and the textile provides a sensation of warmth to have a coherence in terms of a user’s perception of how that the user can associate to the embodiment of the e-coach the e-coach behaves and interacts with his/her despite being in a domestic environment, which should be intimate, cozy, embodied in different devices and using different modalities. The and relaxing. The interaction with the tangible interface wants tangible interface is a physical object designed for domestic use to replicate these sensations of intimate, cozy, and relaxing (Figure 11); it has a particular shape that allows for placement experience. The e-coach is wireless and has a base to charge in two different positions: one vertical with the wide base its battery. In order to be positioned on the charging station, down, and one horizontal with the wide base toward the user. the tangible interface has to be in vertical position, which The first position, the vertical one, corresponds to the sleep implies that it is in sleep mode. The vocal interface is not only position, which implies that the e-coach will not disturb the convenient for the interaction while performing other tasks, but user. The second position (horizontal) means that the user is it also provides the opportunity for sentiment analysis, which open to interact with the e-coach. The wide base integrates LEDs is used to understand affective processes in order to adapt the that are controlled by the system in order to provide different conversation flow and the behavior of the tangible interface (e.g., patterns that accompany different behaviors of the e-coach acknowledgment + encouragement). Frontiers in Digital Health | www.frontiersin.org 12 October 2020 | Volume 2 | Article 545949 Palumbo et al. The NESTORE IoT Personal Assistant FIGURE 10 | Meal logging via the chatbot interface. 5. VALIDATING THE NESTORE SYSTEM All objectives will be addressed by the design outlined in Figure 12 that describes the data collection flow carried out The validation of the NESTORE system has to address the in the three NESTORE pilot sites (Spain, The Netherlands, question of whether the proposed and developed system will and Italy). We will use a pre-/post-test study comparing the indeed serve the intended purpose of having a positive impact on group of individuals that receives the technology-based and well-being and health, and show that the different components personalized health coaching intervention to a control group that and tools are safe and practical in their use. Likewise, validation receives a standard, non-personalized intervention. The pilot- must offer information on the feasibility of carrying out a study will consist of the following phases (1) pre-test assessment, clinical trial to demonstrate the effectiveness of the system in (2) setup of & training with system, (3) motivational and pre- achieving its objective of promoting healthy aging, quality of life intervention tracking phase in daily life, (4) coaching phase, (5) and well-being. post-intervention tracking, and (6) post-test assessment. The study design was selected considering that the user The overall sample consists of a convenience sample of needs to adopt the NESTORE system for at least 3 months to 60 persons (20 per site) using the NESTORE system and 30 allow a sufficiently long time period for the interventions to elicit changes in the respective coaching domains. Moreover, controls (10 per site), gender balanced (15 men and 15 women in order to assure that any observed effects can be attributed at each site). NESTORE users need to be cognitively, physically to the NESTORE intervention specifically, and not just to and mentally healthy enough to be able to use the system. the fact that individuals have engaged in anything new, a At the same time, the system is developed for healthy older well-defined active control group is necessary that engages in adults who do not have certain conditions that could possibly activities comparable with the NESTORE system, but without the make the interventions harmful to themselves. Thus, all chronic “active ingredients” of the NESTORE coach (i.e., personalization diseases (cardiovascular, metabolic, neurological, immunologic) and technology-based features). One feasible solution that will are an exclusion criterion. Only persons with certain controlled be followed in the initial pilot study is to deliver printed chronic conditions can be included as they could benefit from general recommendations about healthy habits in all NESTORE the NESTORE system or the NESTORE system will suggest domains (nutrition/physical activity/cognition/social) to the prospective users to turn to their primary physician to support control group, which would allow a comparison between the in managing their conditions. Participants will be recruited with personalized and technology-based NESTORE interventions age between 65 and 75 years, living alone or community dwelling with more traditional general (i.e., non-personalized) and paper- in autonomy, with ability to use standard ICT devices and based coaching material. availability of internet connection at home. Frontiers in Digital Health | www.frontiersin.org 13 October 2020 | Volume 2 | Article 545949 Palumbo et al. The NESTORE IoT Personal Assistant 5.1. Preliminary Findings affects the overall perception of the system, despite its The fieldwork started on February 2020 with the baseline good functioning. assessments and the first installations of the intervention group. Regarding the first finding, through the definition of a A prototype of the Personalized Coach kit has been deployed procedure were established the main needs to be covered in in two Pilot sites of the NESTORE project in Barcelona (Spain) each of the 3 stages: pre-installation (where it is structured and Monza (Italy). Ten older people (4 Women, 6 Men) have and planned all the necessary aspects related to deployment been equipped with the kit before COVID-19 lock-down. After and installation like the definition of workflows, training completing a maximum of 4 weeks of system use, 2 or 3 weeks of materials and resources), installation (when all the steps are daily tracking and one or two of intervention, some lessons learnt performed, but also it is reinforced the user training in the emerged from this short period. system use) and post-installation (when the system is completely The preliminary results revealed two main aspects: the deployed and prepared for the participant autonomous use). significance of a detailed and well-specified procedure is In the pre-installation it was considered crucial to do a essential for the correct progress of the installation process user-centered installation that means respectful with the daily and secondly, the significant failure of a part of the system rhythms, reducing the presence in their homes and based in personalize attention of their needs. During the installation, an ordered written procedure set the steps to avoid mistakes. Furthermore, it was video recorded the first installation as a model for the rest of installations. And finally, the post-installation was encompassed a non-intrusive method to monitor the data flow to assure that the system was working correctly. Some qualitative results were obtained before the intermediate assessment thanks to the voluntarily self-report of users using the provided channels by researchers. From a performance point of view, the first results showed that it is critical the information that users received before installation. When users started to use the system, they modified their perception related to the system usefulness, because their expectancy did not match with the real life first impressions of use. However, it is important to note that the project was stopped just in the moment when participants were familiarizing and learning how to use FIGURE 11 | The tangible interface. the system. FIGURE 12 | NESTORE Study Design: Experimental and Control Group. Frontiers in Digital Health | www.frontiersin.org 14 October 2020 | Volume 2 | Article 545949 Palumbo et al. The NESTORE IoT Personal Assistant In reference to the sensing layer the first findings tasks and the benefit directly obtained from the system demonstrated that the wristband is one of the system devices used, affected they usefulness perception after the daily that centered more interest in users during the recruitment stage. tracking period. The first results suggest a corpus of perceptions which created a framework providing significant opportunities to improve certain aspects in functional and non-functional areas described 6. DISCUSSION AND CONCLUSIONS in the following: The NESTORE virtual coach personalized strategies includes • Power consumption: Although, the battery life is 18 h, the self-monitoring, context-awareness, and different interface comparison with commercial solutions affect the participant’ modalities to better address when, how, and what coaching perception about the device reliability. messages to deliver to the user in an automated and intelligent • Battery charge base: In the post-installation phase, some way. However, besides the significant challenges related to participants reported problems to the correctly place the optimization and orchestration of the different hardware of the wristband in the charging base, even though (i.e., set of devices) and software (i.e., back-end modules) it was provided a manual and they were instructed to artifacts involved in the system, it is important to provide do it. to the user a valid tool, as a virtual companion, in terms of • Monitoring the battery status: Uncertainty generated usability and effectiveness. For these reasons, we presented in some participants when the battery life is less the overall architecture of the NESTORE system, the different than expected. hardware and software modules enabling the virtual coaching, • Switch system: reported difficulties to turn on and off the and a validation study about the feasibility of carrying out a wristband after the training session. clinical trial to demonstrate the effectiveness of the system in achieving its objective in promoting healthy aging, quality of life, With respect to the software layer, some barriers were reported and well-being. by users after the daily tracking described in the following: Since the design phases of the system, we put the muldi- • Exceeding the push notifications number over user’s domain approach at the core of NESTORE both in the modeling expectations: The theoretical number of push notifications of the end-user and in the development of the different is not in accordance with the real users’ tolerance frame. modules in the back-end. Each service, thanks to the modularity But also, the findings showed that a group of participants offered by the IoT architecture, exchanges information about were unconscious of receiving them, and a reduce part of related domain which is then orchestrated by the DSS to participants informed that they deliberately ignored them provide to the user personalized pathways of well-being. This after few days. On the other hand, users did not appreciate approach has its roots in the participatory design techniques receiving duplicate notifications and affected their satisfaction adopted, in terms of co-design and co-creation of the offered perception. services. In this vision, we also embraced the technological • Connection between NESTORE coach app and NESTORE outcomes from the community created around EU programme connect app: When the participant is prepared to on “Active aging and self-management of health,” offering an perform some formal tests using the NESTORE app, the IoT architecture supporting the integration with other strategic communication between it and the NESTORE connect create projects addressing similar objectives. a loop that generate the perception that something is not working well. It happened very often in the beginning of the pilot study because some Android functionalities in the DATA AVAILABILITY STATEMENT firmware. The original contributions presented in the study are included • Task repetition and workload during intervention: Users in the article/supplementary material, further inquiries can be reported the feeling that they always have done the same type directed to the corresponding author/s. of task related to the intervention like fulfill questionnaires, but also some tasks were reported as heavy. • Motivation triggers: The system did not integrate enough AUTHOR CONTRIBUTIONS mechanisms a part of notifications that could sustain over the intervention the users’ interest to complete new activities. FP developed the idea and led this writing of the manuscript. AC, During the co-design participants asked for having a model of FF, and MG contributed to the section related to the integrated coaching friendly, but the prototype is not perceived in this infrastructure. AM, GM, CR, SG, AS, and GR described the direction by the moment. personalized coaching experience as domain experts. MC, FC, • Usefulness perception of NESTORE Coach app: Some OA, and EM contributed to the description of the virtual coach. functionalities were not used during the first weeks of the EM, ED, MM, and DW contributed to the description of the intervention because there were some technical incidences. monitoring system. PS-B and SO described the decision support On the other hand, participants reported problems using system. CC and GC contributed to the description of the IoT the nutrition app or the cognitive tests. These problems architecture. All authors edited the drafts of the manuscript and but also the unbalance between the time invested to do provided critical revisions. Frontiers in Digital Health | www.frontiersin.org 15 October 2020 | Volume 2 | Article 545949 Palumbo et al. The NESTORE IoT Personal Assistant FUNDING ACKNOWLEDGMENTS This work has been funded by the EU H2020 project The authors wish to thank Marta Civiello, Paola Nonini, and Novel Empowering Solutions and Technologies Andrea Mangone from Flex Ltd for their contribution to the for Older people to Retain Everyday life activities design of the wearable device and all the partners involved in (NESTORE), GA769643. the project. REFERENCES 17. Angelini L, Mugellini E, Khaled OA, Röcke C, Guye S, Porcelli S, et al. The NESTORE e-coach: accompanying older adults through a personalized 1. Preedy VR, Watson RR. Handbook of Disease Burdens and Quality of Life pathway to wellbeing. In: Proceedings of the 12th ACM International Measures. New York, NY: Springer (2010). doi: 10.1007/978-0-387-78665-0 Conference on PErvasive Technologies Related to Assistive Environments. 2. Khosla R, Chu MT, Nguyen K. Enhancing emotional well being of Rhodes: ACM (2019). p. 620–8. doi: 10.1145/3316782.3322763 elderly using assistive social robots in Australia. In: 2013 International 18. Orte S, Subias P, Maldonado LF, Mastropietro A, Porcelli S, Rizzo G, et al. Conference on Biometrics and Kansei Engineering. Tokyo (2013). Dynamic decision support system for personalised coaching to support active doi: 10.1109/ICBAKE.2013.9 aging. In: Bandini S, Cortellessa G, Gorrini A, Palumbo F, editors. 4th Italian 3. Khosla R, Chu MT. Embodying care in Matilda: an affective communication Workshop on Artificial Intelligence for Ambient Assisted Living (AI*AAL.it). robot for emotional wellbeing of older people in Australian residential care No. 2333 in CEUR Workshop Proceedings - AI*IA Series. Aachen (2018). p. facilities. ACM Trans Manage Inf Syst. (2013) 4:2544104. doi: 10.1145/2544104 16–36. Available online: http://ceur-ws.org/Vol-2333/paper2.pdf 4. Crivello A, Barsocchi P, Girolami M, Palumbo F. The meaning of sleep 19. Gierveld JDJ, Tilburg TV. A 6-item scale for overall, emotional, and social quality: a survey of available technologies. IEEE Access. (2019) 7:167374–90. loneliness: confirmatory tests on survey data. Res Aging. (2006) 28:582–98. doi: 10.1109/ACCESS.2019.2953835 doi: 10.1177/0164027506289723 5. Oxford Economics. The Silver Economy, Final Report - Study. Technopolis 20. Lubben J, Blozik E, Gillmann G, Iliffe S, von Renteln Kruse W, Beck JC, (2018). et al. Performance of an abbreviated version of the Lubben Social Network 6. Konstantinidis EI, Petsani D, Conti G, Billis A, Conotter V, Chican G, et al. A Scale among three European community-dwelling older adult populations. new approach for aging at home: the CAPTAIN system. In: Ohno-Machado Gerontologist. (2006) 46:503–13. doi: 10.1093/geront/46.4.503 L, Séroussi B, editors. MedInfo. (2019). p. 1704–5. 21. Crivello A, Palumbo F, Barsocchi P, La Rosa D, Scarselli F, Bianchini 7. op den Akker H, op den Akker R, Beinema T, Banos O, Heylen D, M. Understanding human sleep behaviour by machine learning. In: Bedsted B, et al. Council of coaches a novel holistic behavior change Klempous R, Nikodem J, Baranyi PZ, editors. Cognitive Infocommunications, coaching approach. In: 4th International Conference on Information and Theory and Applications. Cham: Springer (2019). p. 227–52. Communication Technologies for Ageing Well and e-Health, ICT4AWE. doi: 10.1007/978-3-319-95996-2_11 Funchal: SciTePress (2018). p. 219–26. doi: 10.5220/0006787702190226 22. Baronti P, Barsocchi P, Chessa S, Crivello A, Girolami M, Mavilia F, et al. 8. Kouris I, Sarafidis M, Androutsou T, Koutsouris D. HOLOBALANCE: Remote detection of social interactions in indoor environments through an augmented Reality virtual trainer solution forbalance training and fall bluetooth low energy beacons. J Ambient Intell Smart Environ. (2020) prevention. In: 2018 40th Annual International Conference of the IEEE 12:203–17. doi: 10.3233/AIS-200560 Engineering in Medicine and Biology Society (EMBC). Honolulu, HI: IEEE 23. Mavilia F, Palumbo F, Barsocchi P, Chessa S, Girolami M. Remote detection of (2018). p. 4233–6. doi: 10.1109/EMBC.2018.8513357 indoor human proximity using bluetooth low energy beacons. In: 2019 15th 9. Mastropietro A, Röcke C, Porcelli S, del Bas J, Boqué N, Fernández Maldonado International Conference on Intelligent Environments (IE). Rabat: IEEE (2019). L, et al. Multi-domain model of healthy aging : the experience of the p. 16–21. doi: 10.1109/IE.2019.000-1 H2020 NESTORE project. In: Leone A, Caroppo A, Rescio G, Diraco G, 24. Palumbo F, Baronti P, Crivello A, Furfari F, Girolami M, Mavilia F, et al. Siciliano P, editors. Lecture Notes in Electrical Engineering. (2018). p. 13–21. Exploiting BLE beacons capabilities in the NESTORE monitoring system. In: doi: 10.1007/978-3-030-05921-7_2 Bandini S, Cortellessa G, Palumbo F, editors. 5th Italian Workshop on Artificial 10. Girolami M, Mavilia F, Delmastro F, Distefano E. Detecting Social Intelligence for Ambient Assisted Living (AI*AAL.it). No. 2559 in CEUR Interactions through Commercial Mobile Devices. In: 2018 IEEE Workshop Proceedings - AI*IA Series. Aachen (2019). p. 66–84. Available International Conference on Pervasive Computing and Communications online at: http://ceur-ws.org/Vol-2559/paper5.pdf Workshops (PerCom Workshops). Athens. (2018). p. 125–30. 25. Guinard D, Trifa V. Building the Web of Things: With Examples in Node.js and doi: 10.1109/PERCOMW.2018.8480397 Raspberry Pi. Shelter Island, NY: Manning Publications Co. (2016). 11. Girolami M, Barsocchi P, Chessa S, Furfari F. A social-based service 26. Fernandes JL, Lopes IC, Rodrigues JJ, Ullah S. Performance evaluation of discovery protocol for mobile ad hoc networks. In: 2013 12th Annual RESTful web services and AMQP protocol. In: 2013 Fifth International Mediterranean Ad Hoc Networking Workshop (MED-HOC-NET). (2013). p. Conference on Ubiquitous and Future Networks (ICUFN). IEEE (2013). p. 103–10. doi: 10.1109/MedHocNet.2013.6767417 810–5. doi: 10.1109/ICUFN.2013.6614932 12. Jacko JA, Leonard VK, Scott IU. Perceptual impairments: new advancements 27. Rayes A, Salam S. (editors) The internet in IoT-OSI, TCP/IP, IPv4, IPv6 and promoting technological access. In: Sears A, Jacko JA, editors. Human- internet routing. In: Internet of Things From Hype to Reality. Cham: Springer Computer Interaction: Designing for Diverse Users and Domains. (2009). p. (2017). p. 35–56. doi: 10.1007/978-3-319-44860-2_2 93–110. doi: 10.1201/9781420088885.ch6 28. Guinard D, Trifa V, Pham T, Liechti O. Towards physical mashups in the 13. Higgins JP. Smartphone applications for patients’ health and fitness. Am J web of things. In: Proceedings of INSS. Vol. 9. Pittsburgh, PA (2009). p. 17–9. Med. (2016) 129:11–9. doi: 10.1016/j.amjmed.2015.05.038 doi: 10.1109/INSS.2009.5409925 14. Banos O, Nugent C. E-coaching for health. Computer. (2018) 51:12–5. 29. Fielding RT, Taylor RN. Principled design of the modern Web architecture. doi: 10.1109/MC.2018.1731070 ACM Trans Internet Technol. (2002) 2:115–50. doi: 10.1145/514183. 15. Kamphorst BA. E-coaching systems. Pers Ubiquit Comput. (2017) 21:625–32. 514185 doi: 10.1007/s00779-017-1020-6 30. Ruby S. RESTful Web Services. Sebastopol, CA: O’Reilly (2007). 16. Lentferink AJ, Oldenhuis HK, de Groot M, Polstra L, Velthuijsen H, van 31. Pautasso C, Wilde E. Why is the web loosely coupled?: A multi- Gemert-Pijnen JE. Key components in eHealth interventions combining self- faceted metric for service design. In: Proceedings of the 18th International tracking and persuasive eCoaching to promote a healthier lifestyle: a scoping Conference on World Wide Web. Madrid: ACM (2009). p. 911–20. review. J Med Internet Res. (2017) 19:e277. doi: 10.2196/jmir.7288 doi: 10.1145/1526709.1526832 Frontiers in Digital Health | www.frontiersin.org 16 October 2020 | Volume 2 | Article 545949 Palumbo et al. The NESTORE IoT Personal Assistant 32. Baronti P, Barsocchi P, Chessa S, Mavilia F, Palumbo F. Indoor bluetooth low fields. Pers Ubiquit Comput. (2018) 22:227–43. doi: 10.1007/s00779-017- energy dataset for localization, tracking, occupancy, and social interaction. 1038-9 Sensors. (2018) 18:4462. doi: 10.3390/s18124462 43. Barsocchi P, Bianchini M, Crivello A, La Rosa D, Palumbo F, Scarselli F. 33. Palumbo F, Barsocchi P, Chessa S, Augusto JC. A stigmergic approach to An unobtrusive sleep monitoring system for the human sleep behaviour indoor localization using bluetooth low energy beacons. In: 2015 12th IEEE understanding. In: 2016 7th IEEE International Conference on Cognitive International Conference on Advanced Video and Signal Based Surveillance Infocommunications (CogInfoCom). Wroclaw: IEEE (2016). p. 91–6. (AVSS). Karlsruhe: IEEE (2015). p. 1–6. doi: 10.1109/AVSS.2015.7301734 doi: 10.1109/CogInfoCom.2016.7804531 34. Barsocchi P, Crivello A, Girolami M, Mavilia F, Palumbo F. Occupancy 44. Subías-Beltrán P, Orte S, Vargiu E, Palumbo F, Angelini L, Abou Khaled O, detection by multi-power bluetooth low energy beaconing. In: 2017 et al. A decision support system to propose coaching plans for seniors. In: International Conference on Indoor Positioning and Indoor Navigation (IPIN). 2019 IEEE 32nd International Symposium on Computer-Based Medical Systems Sapporo: IEEE (2017). p. 1–6. doi: 10.1109/IPIN.2017.8115946 (CBMS). Cordoba: IEEE (2019). p. 592–95. doi: 10.1109/CBMS.2019.00123 35. Barsocchi P, Cimino MG, Ferro E, Lazzeri A, Palumbo F, Vaglini G. 45. Kniestedt I, Lukosch S, Brazier F. User-centered design of an online Monitoring elderly behavior via indoor position-based stigmergy. mobile game suite to affect well-being of older adults. In: International Pervas Mobile Comput. (2015) 23:26–42. doi: 10.1016/j.pmcj.2015. Conference on Entertainment Computing. Poznan: Springer (2018). p. 355–61. 04.003 doi: 10.1007/978-3-319-99426-0_44 36. Palumbo F, La Rosa D, Ferro E. Stigmergy-based Long-Term Monitoring 46. Binder JC, Zöllig J, Eschen A, Mérillat S, Röcke C, Schoch S, et al. Multi- of Indoor Users Mobility in Ambient Assisted Living Environments: the domain training in healthy old age: Hotel Plastisse as an iPad-based serious DOREMI Project Approach. In: Bandini S, Cortellessa G, Palumbo F, editors. game to systematically compare multi-domain and single-domain training. AI*AAL2016 Artificial Intelligence for Ambient Assisted Living. Aachen (2013). Front Aging Neurosci. (2015) 7:137. doi: 10.3389/fnagi.2015.00137 p. 18–32. 37. Occupational Safety and Health Administration. Indoor Air Quality in Conflict of Interest: ED, MM, and DW were employed by company Commercial and Institutional Buildings. Washington, DC: Maroon Ebooks FLEXTRONICS DESIGN SRL, while CC and GC were employed by company (2015). ROPARDO SRL. 38. ISO/IEC/IEEE 15288:2015. Systems and Software Engineering: System Life Cycle Processes. (2015). The remaining authors declare that the research was conducted in the absence of 39. Shortell TM. INCOSE Systems Engineering Handbook: A Guide for System Life any commercial or financial relationships that could be construed as a potential Cycle Processes and Activities. Hoboken, NJ: John Wiley & Sons (2015). conflict of interest. 40. Bacciu D, Chessa S, Gallicchio C, Micheli A, Pedrelli L, Ferro E, et al. A learning system for automatic Berg Balance Scale score estimation. Copyright © 2020 Palumbo, Crivello, Furfari, Girolami, Mastropietro, Manferdelli, Eng Appl Artif Intell. (2017) 66:60–74. doi: 10.1016/j.engappai.2017. Röcke, Guye, Salvá Casanovas, Caon, Carrino, Abou Khaled, Mugellini, Denna, 08.018 Mauri, Ward, Subías-Beltrán, Orte, Candea, Candea and Rizzo. This is an open- 41. Inan OT, Migeotte PF, Park KS, Etemadi M, Tavakolian K, Casanella access article distributed under the terms of the Creative Commons Attribution R, et al. Ballistocardiography and seismocardiography: a review of License (CC BY). The use, distribution or reproduction in other forums is permitted, recent advances. IEEE J Biomed Health Inform. (2014) 19:1414–27. provided the original author(s) and the copyright owner(s) are credited and that the doi: 10.1109/JBHI.2014.2361732 original publication in this journal is cited, in accordance with accepted academic 42. Alfeo AL, Barsocchi P, Cimino MG, La Rosa D, Palumbo F, Vaglini practice. No use, distribution or reproduction is permitted which does not comply G. Sleep behavior assessment via smartwatch and stigmergic receptive with these terms. Frontiers in Digital Health | www.frontiersin.org 17 October 2020 | Volume 2 | Article 545949