Evolving Ambient Intelligence

This workshop is designed to bring together researchers from both industry and academia from the various disciplines to discuss how innovation in the use of technologies to support healthier lifestyles can be moved forward. There has been a growing interest around the world and especially in Europe, on investigating the potential consequences of introducing technology to deliver social and health care to citizens (see for example [1]). This implies an important shift on how social and health care are delivered and it has positive as well as negative consequences which must be investigated carefully. On the other hand there is an urgency provided by the changes in demographics which is putting pressure on governments to provide care to specific sectors of the population, especially older adults, a group which is growing thanks to advances in medicine and greater knowledge on the relationships between lifestyles and health.

Extended periods of time sitting in front of a computer give rise to risks of developing musculoskeletal disorders. In the workplace, computer use contributes considerably to employee injury and results in significant costs to the employer in terms of sick leave and injury claims. Due to these risks there has been significant research into the areas of posture classification and subject intervention to improve posture in an office environment. The Kinect

TM

has been shown to be a suitable hardware platform for posture classification. This paper presents a system for posture classification and novel subject intervention that leverages each of three distinct forms of persuasive computing and explores the success of each type. Our results show significant improvement in posture results from the most effective of our intervention types.

This paper describes a semi-automated web-based system to facilitate digital reminiscence therapy for patients with mild-to-moderate dementia, enacted in a group setting. The system, REMPAD, uses proactive recommendation technology to profile participants and groups, and offers interactive multimedia content from the Internet to match these profiles. In this paper, we focus on the design of the system to deliver an innovative personalized group reminiscence experience. We take a user-centered design approach to discover and address the design challenges and considerations. A combination of methodologies is used throughout this research study, including exploratory interviews, prototype use case walkthroughs, and field evaluations. The results of the field evaluation indicate high user satisfaction when using the system, and strong tendency towards repeated use in future. These studies provide an insight into the current practices and challenges of group reminiscence therapy, and inform the design of a multimedia recommender system to support facilitators and group therapy participants.

Sensor technologies can enable independent living for people with dementia by monitoring their behaviour and identifying points where support may be required. Wearable sensors can provide such support but may constitute a source of stigma for the user if they are perceived as visible and therefore obtrusive. This paper presents an initial empirical investigation exploring the extent to which wearable sensors are perceived as visible. 23 Participants wore eye tracking glasses, which superimposed the location of their gaze onto video data of their panorama. Participants were led to believe that the research entailed a subjective evaluation of the eye tracking glasses. A researcher wore one of two wearable sensors during the evaluation enabling us to measure the extent to which participants fixated on the sensor during a one-on-one meeting. Results are presented on the general visibility and potential fixations on two wearable sensors, a wrist-work actigraph and a lifelogging camera, during normal conversation between two people. Further investigation is merited according to the results of this pilot study.

The mobile ecosystem is rife with applications that aim for individuals to persue a more active and healthier lifestyle. Applications vary from simple diaries that track your weight, calorie intake or blood glucose values towards more advanced ones that offer health recommendations while monitoring your fitness levels during workouts and throughout the day. Leveraging machine learning techniques is a popular approach to recognize non-trivial activities, such as different types of sports. However, such applications face a time consuming training phase before they become practical. In this work, we report on our feasibility analysis of transfer learning as a way to apply learned models from one individual on another, and report on various feature variabilities that may jeopardize the applicability of transfer learning.

Population ageing is an issue that has encouraged the development of Ambient Intelligence systems to support elderly people to live autonomously at home longer. Some key aspects of these systems are the detection of behavior patterns and behavior profiles in their daily life. The information we can infer from these patterns could prove to be very valuable for monitoring the health status of a person, like to control deterioration of diseases or to provide personalized assistive services. In this paper we focus on the unsupervised learning techniques in health monitoring systems for elderly people, which has the advantage of not needing annotations. Collecting these is a tedious job and sometimes difficult to accomplish. We discuss the different existing approaches, identify some limitations and propose possible challenges and directions for future research.

Functional status is a major determinant of clinical outcomes. The Barthel Scale or Barthel Index (BI) is an ordinal scale used to measure performance in activities of daily living. A higher BI is associated with reduced length of stay in hospital and a greater likelihood of being able to live at home with a degree of independence following discharge from hospital. Currently on admission to hospital the BI is assessed subjectively by nursing staff. This work explores the possibility of using wearable wireless inertial measurement as a means of automating and detecting changes in BI. Preliminary findings for a study comprising of 16 patients suggest a correlation (0.7613) between average cumulative movement over 24 hours and variance in BI over the same period.

Different degrees and types of visual impairment have become a common condition among the elderly, as aging inevitably affects the health and lifestyle of individuals. Partial or complete lack of sight is often accompanied by other ailments and conditions which further hinder the individual’s activities. In this work, a novel Ambient Assisted Living (AAL) platform is proposed, aiming to support the functional capabilities of the elderly and visually impaired, thus ameliorating their lifestyle. This platform is based on indoor tracking of commonly used objects, such as medication packages. Accompanied by a proposed behavioural modeling methodology, the application also offers valuable observations that may indicate developing ill-health conditions in an early stage. The proposed platform was tested and evaluated by end-users in Spain, Greece and Finland.

To deal with the large amount of data produced by telemonitoring of patients with chronic diseases, a decision support system (DSS) was developed. The DSS uses sensor data and the data from a patient’s electronic health record as the input. It assesses the risk to the patient’s health by exploiting the existing medical knowledge. The risk assessment can show the contribution of the individual monitored parameters to the risk, and can be tailored by the doctor to each patient.

Robotic ecologies often comprise a large number of environmental sensors and actuators, that, on the other hand, operate by means of their proper standards, among which ZigBee is one of the most widely used. Being designed to build autonomous sensor networks, the use of sensors based on ZigBee requires a deep knowledge of its logic and protocols. In order to facilitate interoperability between ZigBee sensors and external applications we designed ZB4O, an application-level gateway that exports ZigBee services in external networks. This paper describes the experience of integration of ZB4O and ZigBee networks within the robotic ecology GiraffPlus which is being developed within the EU project GiraffPlus.

As part of the ACCOMPANY project we are researching the use of a companion robot for elderly people within a sensorised home. One of our goals is to give end users, such as care workers, relatives and the elderly persons themselves, the ability to create robot behaviours based on events within the home. We employ a knowledge-based approach in dealing with the creation of complex robot behaviours and need to have flexibility in dealing with events occuring at different times and in different timeframes and periods. In this paper we describe our overall approach to creating an environment where robot behaviours can be taught and focus especially on how we deal with the temporal aspects associated with this issue.

As the field of configuration planning grows, so does the need for objective comparisons of algorithms and results. As the community stands today, different approaches to formalise and solve the problem at hand exist, and little or no importance has been given to compare results of different research groups. In this paper we summarize the definitions used by a few different research groups, and we explain two empiric method for comparing planning algorithms, based on statistics. While the methods themselves do not solve all the problems of comparative studies, it is a first step towards numerically comparing performances of the different configuration planning methods proposed by the community.

The paper presents a simple people-centric adaptive ecology between a humanoid robot and a virtual human (social agents) to perform a real-world common complex social task. The social task was to assist the social agents each other in searching for a hidden object in a homely environment. In order to develop the ecology between the agents, we developed the agents with various similar functionalities, interaction modalities, sensing abilities, intelligence, autonomy etc., and integrated them through a common communication platform based on a novel control algorithm. In order to assess people’s acceptance of the ecology between the social agents and to benchmark the ecology, we studied human’s interactions with those agents and with some other allied agents for that task.We evaluated the attributes and performances of the social agents in their cooperations for the task, analyzed the attributes and performances and benchmarked them with the standards. The results showed that both of the social agents within the ecology could perform satisfactorily to accomplish the common social task though the performances varied slightly between the agents.We also found a trade-off between the attributes and the performances of the social agents. We then proposed to use the results to develop adaptive social ecologies with intelligent social agents of different realities to assist the humans in various real-world complex social tasks in smart spaces, or to get the real-world social tasks done in cooperation between the social agents.

Robotic ecologies are networks of heterogeneous devices (sensors, actuators, automated appliances and mobile robots) pervasively embedded in everyday environments, where they cooperate to the achievement of complex tasks. Their successful application opens important research questions for both their engineering and their interaction with human users. In this paper we illustrate a testbed built to support interaction studies between human users and robotic ecologies. The testbed consists of an interactive and autonomous robotic ecology that is able to engage with human users, react to their activities, and even introduce them to the main concepts behind robotic ecologies. We describe how such a testbed is built using a middleware purposefully designed for robotic ecologies, and we report our experiences in its application to a number of project demonstrations and human-robot interaction studies.

To provide a better understanding of the relative strengths of Machine Learning based Activity Recognition methods, in this paper we present a comparative analysis of the robustness of three popular methods with respect to sensor noise. Specifically we evaluate the robustness of Naive Bayes classifier, Support Vector Machine, and Random Forest based activity recognition models in three cases which span sensor errors from dead to poorly calibrated sensors. Test data is partially synthesized from a recently annotated activity recognition corpus which includes both interleaved activities and a range of both temporally long and short activities. Results demonstrate that the relative performance of Support Vector Machine classifiers over Naive Bayes classifiers reduces in noisy sensor conditions, but that overall the Random Forest classifier provides best activity recognition accuracy across all noise conditions synthesized in the corpus. Moreover, we find that activity recognition is equally robust across classification techniques with the relative performance of all models holding up under almost all sensor noise conditions considered.

Dempster-Shafer (DS) theory, and its associated Dempster rule of combination, has been widely used to determine belief based on uncertain evidence sources. Variations to the original Dempster rule of combination have appeared in the literature to support particular scenarios where unreliable results may result from the use of original DS theory. While theoretical explanations of the rule variations are explained, there is a lack of empirical comparisons of the DS theory and its variations against real data sets. In this work, we examine several variations to DS theory. Using two real-world sensor data sets, we compare the performance of DS theory and several of its variations in recognising situations. The empirical results shed insight on how to select these fusion rules based on the nature of sensor data, the relationship of this data over time to the higher level hypotheses and the choice of frame of discernment.

Wireless sensor networks are attracting increasing interest but suffer from severe challenges such as low data reliability. To improve the data reliability, many sensor fault detection techniques have been proposed. Behind these methods, mathematical models are usually employed to serve as comparing metric to find faulty data in the absence of ground truth. In this paper, we firstly discuss sensor data features and their relevance to fault detection. Criteria that should be met to become a competent data model for the purpose of fault detection is summarised. Some existing sensor data modelling methods for fault detection are presented and qualitatively compared.

The aim of reducing greenhouse gases and increasing energy efficiency faces a number of challenges to date. A significant portion of overall energy expenditure in residential and commercial sectors is considered as wastage. Finding technological methods in order to reduce wastage has been the main focus of researchers in recent years. Non-Intrusive Load Monitoring (NILM) is perceived as a cost-effective approach to monitor appliance level energy consumption in a building. However, this approach still faces a number of problems that need to be addressed. In this study, we propose an approach by which uncertainty of appliance’s identification that have similar signatures, is addressed. Unlike other approaches, our approach uses occupant’s behavioural information to aid appliance disaggregation algorithms. We also demonstrate our technique through experimentation in a household.

Ambient intelligence is increasingly used for monitoring of health data, which contributes to personalized and preventive healthcare. While efforts regarding well-being in technology-enhanced spaces are currently very much limited to high-income countries, this article explores the cost effectiveness, emerging necessities and existing opportunities to invest to the transfer of this technology to low and middle-income countries as soon as possible. While this transfer is appropriate for the prevention as well as treatment of many diseases, this article focuses on the particularly relevant example of diabetes, which has also become one of the major health challenges in low and middle-income countries.

We assume that smart environments in combination with mobile devices will increasingly appear in contexts where interactions become transactions. Since most deal-making activities involve at least two parties and a process of negotiation, we propose a generic interface focusing on pre-negotiation, negotiation and contracting within

one

system. We present two prototypes based on the visual metaphor of a marketplace allowing for simple drag and drop actions. This combination of functionalities aims to bridge the gap between deal-making and its legal representation. Visualized interactive contracts introduced as a general convention may improve transparency of legal texts increasing the overall understanding and legal literacy of consumers and professionals alike. An intuitive visualization can differentiate between repetitive elements of standardized contracts, variables and amendments. The proposed platform is applicable to legal contractual contexts including b2b portals, ecommerce, online licensing agreements, financial instruments etc., and may help to transform social networks into transactional ones.

One of the open questions in the concept of ambient intelligence regards user interfaces to these invisible computers. If at all, how do they show up – and how does ambient intelligence in general and the user interfaces in particular change architectural space. As computers become ubiquitous or ambient, they create spatial relations towards other devices and to the place that they are located in. This paper formulates chances and challenges for both architecture and HCI.

Representation and reasoning about context information is a main area of research in Ambient Intelligence (AmI). Given the openness and decentralization of many AmI applications, we argue that usage of semantic web technologies for context modeling brings advantages in terms of standards, uniform representation and expressive reasoning. We present an approach for modeling of context information which builds and improves upon related lines of work (SOUPA, CML, annotated RDF). We provide a formalization of the model and an innovative realization using the latest proposals for semantic web standards like RDF and SPARQL. A commonly encountered ambient intelligence scenario showcases the approach.

Several studies have shown the value of using proximity data to understand the social context of users. To simplify the use of social context in application development we have developed Proximates, a social context engine for mobile phones. It scans nearby Bluetooth peers to determine what devices are in proximity. We map Bluetooth MAC ids to user identities on existing social networks which then allows Proximates to infer the social context of the user. The main contribution of Proximates is its use of link attributes retrieved from Facebook for granular relationship classification. We also show that Proximates can bridge the gap between physical and digital social interactions, by showing that it can be used to measure how much time a user spends in physical proximity with his Facebook friends. In this paper we present the architecture and initial experimental results on deployment usability aspects of users of an example application. We also discuss using location for proximity detection versus direct sensing using Bluetooth.

In this paper we discuss the possibilities of context-aware systems in providing more secure user authentication. We describe some approaches in using context information in adaptive security systems, especially in adaptive user authentication. In addition, we discuss some recent results in applying the context itself as an authentication factor. Recent advances in cryptographic protocol design and adaptive, context-aware systems enable the linking of the context information to the cryptographic keys and authentication. Furthermore, new protocols make adaptive user authentication easier as it is possible to combine several different factors in a single protocol. We give some examples of this and discuss the further potential of these methods.

Ontology-based model in context-aware systems offers more expressiveness, semantically sharing and interoperability. It supports reasoning tasks in a better way than other approaches. However, the main concern with ontology implementation is the expensive computational request for the reasoning process which makes this model unsuitable for critical time applications. This applies to tourism recommender systems when user is moving. Late response based on current tourist location might recommend him points of his interest that are already behind him. In this paper, we propose an ontology-based tourism mobile system that uses current user location and his speed to provide him with a recommendation about points of his interest before reaching them.

Urban environments are intelligent spaces where a wide set of heterogeneous variables that directly influence the behavior of the individual converge. In this paper we present UrbanContext, a new model for urban platforms that follows an individual centered approach and validates the use of the Theory of Roles to understand the behavior of the individual within a social environment. The roles defined in UrbanContext allow the interpretation of the states of the individual, facilitating his interaction with the environment and offering services without damaging his privacy. We describe the UrbanContext model and the fundamental principles that have been identified in this design; likewise we present a first validation scenario for UrbanContext.

The rapid increase in mobile data activity has raised the stakes on developing innovative new technologies and cellular topologies that can meet these demands in an efficient manner. Femtocell seems to be a good candidate that can potentially play a much broader role by enabling a new class of multimedia and family communications services. This concept may have great consumer appeal because it has the potential to transform the way people stay connected with the two things that matter most: their family (or friends) and their media. This new paradigm is called: the femto-group or Femtozone.

In this work, we present an MDA-based tool (Model driven architecture) for ubiquitous service discovery between heterogenous user’s devices and services in a Femtozone.

Context-awareness is one of the most relevant research areas in Pervasive Computing scope. However, in more complex environments, loosely coupled data and dynamic behavior of the environment can become a source of serious problems. These problems can spread throughout whole proactive behavior and in this way make them difficult to handle. In this paper we argue that there is a lack of research done about the data quality and quality of awareness particularly in this kind of composite systems. We investigate on how to build the semantic layer of the DOHA middleware due to reach the goal of context-awareness. To achieve that we are using a data quality approach and data profiles. In addition, we use ontologies as information representation method and to carry over the information about the profiles to the pervasive environment. By doing so, we assure proper representation of the real world and we are improving semantic data and context-awareness in such environments.

OUI (Organic User Interface) together with NUI (Natural User Interface) and TUI (Tangible User Interface) falls under the field of Ubiquitous Computing, aiming to give the digital life a deeper human sense. Accordingly, we presented an OUI framework for designing OUIs to formalize the input and output interaction techniques for OUI users, and discuss it based on novel perspective of the organic interaction possible styles. We enhance the SMaG (Speech/ Manipulation/ air-Gesture) interaction model for organic interaction styles, in addition to our OUI design principles and design-specific guidelines in a perspective based on the usability aspects of the look, feel and design. Moreover, we propose a new usability model in this paper to enable measuring and testing the usability standards of organic systems, in which we called the 3Es usability model, referring to Efficiency, Enjoyment and Easiness.

Finally, we developed the ’OUI Sketcher’ application as a case study for our proposed OUI framework and thus applying and testing the SMaG model. OUI Sketcher enables the user to create sketch figures and geometric drawings with all three OUI natural interactions (voice commands, touch and air-gestures) as interaction techniques, to apply the SMaG model perspectives and measure the usability standards against our OUI framework, using the 3Es usability model.

This paper describes a system for ubiquitous monitoring of health and physical parameters, suitable to run on home-based infrastructures and personal mobile settings. The system is built on a micro Web of Things Open Platform (

μ

WoTOP), enabling easy integration of sensing and processing modules. It also facilitates building ubiquitous services by accessing biometric sensors through uniform interfaces based on REST. Additionally, the system is capable of processing user context and generate alerts based on the Common Alerting Protocol, making its output compatible with already existing medical solutions. The system has been customized to deploy a fall/faint detection solution, which enables multiple supervisors with the capability of receiving real time alerts on monitored subjects.

This paper illustrates the interest of the technological platform developed by the European project FI-WARE for applications that are dedicated to smart home usage. Indeed, this project delivers a novel service infrastructure. This infrastructure is offering a set of Generic Enablers (GE) as generic as possible to be reused by the FI-PPP use case projects as well as by different partners of FI-WARE. In this paper, we selected two generic enablers to be tested and evaluated. This in progress study aims to better assess these generic enablers and use them to elaborate realistic innovative services.

Use of mobile devices for the personal and corporate purposes is growing rapidly. Context-awareness is an essential feature of the mobile apps. In this paper, we present an approach to predict the next place for a mobile phone by using an online learning method. We represent the model in the form of state-action representation. Each state is a distinct context and behavior of the app is represented in the form of actions applicable at that state. The results show that online learning based approach performs better than two state-of-the-art mobility prediction approaches. Performance is measured in term of accuracy to predict the next location of a mobile host.

With the ever increasing device count and the introduction of wireless pervasive computing in building automation, in order 1) to enhance people life comfort and safety and 2) to optimize the consumption cost of public utility services like energy and water; it is becoming more than crucial to automatically manage these commonly known smart buildings. Indeed, in this research work, we propose a platform that addresses the problem of energy management of different electric devices in a smart Information and Communication Technology (ICT) building and particularly high performance computing (HPC) data centers. This platform integrates the mobile technology as a means to facilitate the maintenance operations usually executed by the mobile workforce agents on smart building electric devices. Another important issue integrated in this proposed platform deals with the monitoring of a life safety plan in case of an emergency scenario due, for instance, to fire detection. The emergency rescue monitoring is eased by the use of mobile devices such as the tablets, usually handled by people in a smart ICT building.