HIIT
Publications
2005-1
Helsinki Institute for Information Technology HIIT, founded in 1999, is a jointresearch institute of the University of Helsinki and Helsinki University of Technology.
HIIT conducts internationally high-level strategic research in information technology and related multi-disciplinary topics, especially in areas where Finnish IT industry has or may reach a significant global role. HIIT works in close co-operation with Finnish universities, research institutes, and industry, aiming to improve the contents, visibility, and impact of Finnish IT research to benefit the competitiveness and progress of the Finnish information society. HIIT also aims at creating a strong network of international partnerships with leading foreign research universities and institutions.
WWW: http://www.hiit.fi
ISBN 952-10-2518-2 (paperback) ISBN 952-10-2519-0 (PDF)
ISSN 1458-9451 (paperback) ISSN 1458-946X (PDF)
P
ROCEEDINGS OF THE
W
ORKSHOP ON
C
ONTEXT
A
WARENESS
FOR
P
ROACTIVE
S
YSTEMS
CAPS
2005
H
ELSINKI
,
F
INLAND
,
16-17
J
UNE
2005
Edited by Patrik Floréen, Greger Lindén,
Tiina Niklander and Kimmo Raatikainen
Flor éen, L ind én, Nik land er, Raatikainen ( eds.): Proc eedings of C AP S 2005 HIIT Publications 2005-1
Proceedings of the Workshop on Context Awareness for Proactive Systems
CAPS 2005
Helsinki, Finland, 16-17 June 2005
Edited by Patrik Floréen, Greger Lindén, Tiina Niklander and Kimmo Raatikainen
HIIT Publications 2005-1
Helsinki Institute for Information Technology HIIT
http://www.hiit.fi
ISBN 952-10-2518-2 (paperback)
ISBN 952-10-2519-0 (PDF)
ISSN 1458-9451 (paperback)
ISSN 1458-946X (PDF)
Helsinki University Press
2005
Preface
The 20 papers in this volume were presented at CAPS 2005, the First Workshop on Context Awareness for Proactive Systems. The objectives of the workshop were to explore what system architectures and infrastructure are needed for man-aging and using context in context-aware systems, how to model and represent context in proactive systems and how to reason with context and predict future context and needs of the user.
Proactive systems make inferences about the user’s future state, activities and goals, and provide services, resources or information that can help the user succeeding in these. Such systems are increasingly deployed in mobile environ-ments like laptops, mobile phones and PDAs. However, this move to a mobile and pervasive environment raises new opportunities and demands on the under-lying systems. In particular, the systems benefit from being personalised and context-aware.
The workshop was held in Helsinki on 16-17 June 2005 and was organised in cooperation with the Department of Computer Science of the University of Helsinki, Helsinki Institute for Information Technology HIIT and the Academy of Finland Research Programme on Proactive Computing (PROACT). The or-ganising committee wishes to thank all referees who evaluated submitted papers. The organising committee also wishes to thank all volunteers who helped to or-ganise this event. We are grateful to the Department of Computer Science, the Helsinki Institute for Information Technology HIIT and the Academy of Fin-land Research Programme on Proactive Computing (PROACT) for sponsoring the workshop.
June 2005 Kimmo Raatikainen
Patrik Flor´een CAPS 2005
Organisation
CAPS 2005 is organised by the Department of Computer Science at the Univer-sity of Helsinki, the Helsinki Institute for Information Technology HIIT and the Academy of Finland Research Programme on Proactive Computing (PROACT). Organising Committee
Patrik Flor´een, chair, Helsinki Institute for Information Technology HIIT Inka Kujala, Helsinki Institute for Information Technology HIIT
Greger Lind´en, Helsinki Institute for Information Technology HIIT Tiina Niklander, University of Helsinki
Kimmo Raatikainen, University of Helsinki Programme Committee
Kimmo Raatikainen, chair, University of Helsinki, Finland Heikki Ailisto, VTT, Finland
Guy Bernard, Institut National des Telecommunications, France David Bonnefoy, Motorola, France
Klaus David, University of Kassel, Germany
Patrik Flor´een, Helsinki Institute for Information Technology HIIT, Finland Stefan Gessler, NEC, Germany
Heikki Helin, TeliaSonera, Finland Theo Kanter, Ericsson, Sweden Mika Klemettinen, Nokia, Finland
Herma van Kranenburg, Telematica, the Netherlands
Martti M¨antyl¨a, Helsinki Institute for Information Technology HIIT, Finland Heikki Saikkonen, Helsinki University of Technology, Finland
Kaisa Sere, ˚Abo Akademi University, Finland Hannu Toivonen, University of Helsinki, Finland Santtu Toivonen, VTT, Finland
Matthias Wagner, DoCoMo Euro-Labs, Germany Proceedings
Greger Lind´en Mika Raento
Sponsoring Institutions
Department of Computer Science, University of Helsinki Helsinki Institute for Information Technology HIIT
Table of Contents
INVITED PAPERS
Building Intelligent Environments Using Smart Daily Objects and
Personal Devices. . . 1
Tatsuo Nakajima, Kaori Fujinami, Eiji Tokunaga (Waseda University)
What Is Wrong with Maria? Privacy Concerns in the ISTAG Scenario . . . 13
Olli Pitk¨anen (HIIT)
CONTRIBUTED PAPERS
Architectures
Designing and Implementing a System for the Provision of Proactive
Context-Aware Services. . . 21
Tapio Pitk¨aranta (VTT), Oriana Riva (HIIT), Santtu Toivonen (VTT)
Architectures: Frameworks
An Architecture Supporting Implementation of Context-Aware Services. . 31
Anders Kofod-Petersen (NTNU), Marius Mikalsen (SINTEF)
Distributed Context Reasoning for Proactive Systems . . . 43
Michael Przybilski (HIIT)
MoBe: A Framework for Context-Aware Mobile Applications. . . 55
Paolo Coppola, Vincenzo Della Mea, Luca Di Gaspero, Stefano Mizzaro, Ivan Scagnetto, Andrea Selva, Luca Vassena, Paolo Zandegiacomo Rizi`o (University of Udine)
Architectures: Ontologies
OWL-SF – A Distributed Semantic Service Framework . . . 67
Bernd Mrohs (FOKUS), Marko Luther (DoCoMo Euro-Labs), Raju Vaidya (FOKUS), Matthias Wagner (DoCoMo Euro-Labs), Stephan Steglich (FOKUS), Wolfgang Kellerer (DoCoMo Euro-Labs), Stefan Arbanowski (FOKUS)
Specification and Design of Framework-Based Context Processing Modules 79
Marco Luca Sbodio (Hewlett-Packard), Wolfgang Thronicke (Siemens)
CAMidO, A Context-Aware Middleware Based on Ontology Meta-Model 93
VI
Privacy
Privacy Management for Social Awareness Applications. . . 105
Mika Raento (HIIT, University of Helsinki), Antti Oulasvirta (HIIT)
Trust and Information
Trust and Privacy in Context-Aware Support for Communication in
Mobile Groups. . . 115
Stefano Campadello (Nokia Research Center), Olivier Coutand (University of Kassel), Christian del Rosso (Nokia Research Center), Silke Holtmanns (Nokia Research Center), Theo Kanter (Ericsson Research), Christian R¨ack (FOKUS), Bernd Mrohs (FOKUS), Stephan Steglich (FOKUS)
Modelling and Simulating Context Data in a Mobile Environment. . . 127
Agathe Battestini, John A. Flanagan (Nokia Research Center)
A Testbed for Proactive Information Retrieval. . . 137
Miikka Miettinen, Ville H. Tuulos, Petri Myllym¨aki (HIIT)
Reasoning
Route Prediction from Cellular Data . . . 147
Kari Laasonen (HIIT, University of Helsinki)
Enabling Proactiveness through Context Prediction . . . 159
Petteri Nurmi (HIIT), Miquel Martin (NEC Europe), John A. Flanagan (Nokia Research Center)
POSTERS AND DEMONSTRATIONS
OpenChannel — A Task-based Framework for Building Multi-Channel
and Context-Aware Enterprise Applications . . . 169
Antti Martikainen, Pekka K¨ahkipuro (SysOpen Digia)
Proactive Context-Aware Service Provisioning for a Marine Community . 175
Oriana Riva (HIIT), Tapio Pitk¨aranta (VTT), Santtu Toivonen (VTT)
Context-Aware Presence Management . . . 177
Bernd Mrohs (FOKUS), Marko Luther (DoCoMo Euro-Labs), Raju Vaidya (FOKUS)
ContextMedia . . . 181
Mika Raento, Renaud Petit (HIIT, University of Helsinki)
Architecture for Higher-Level Context Composition . . . 183
Waltenegus Dargie, Tino L¨offler, Olaf Dr¨ogehorn, Klaus David (University of Kassel)
1
Building Intelligent Environments Using Smart
Daily Objects and Personal Devices
Tatsuo Nakajima, Kaori Fujinami, and Eiji Tokunaga
Department of Computer Science Waseda University
{tatsuo,fujinami,eitoku}@dcl.info.waseda.ac.jp
Abstract. Our daily life will be more attractive when our
surround-ings will be more intelligent. Most of current researches in ubiquitous computing try to build a smart environment by embedding sensors and computers in our living spaces directly. However, the approach is very expensive to make our environment smart. Our approach uses intelligent daily objects and personal devices to build a smart environment. This makes it possible to make our environment smart in an incremental way by replacing existing objects to new intelligent objects whenever a person wants.
In our project, we are currently developing sentient artefacts that are intelligent daily objects embedding computers and sensors. Also, we are implementing middleware infrastructures for personal devices that allow us to control embedded services in an easy way.
1
Introduction
The advancement of computer technologies allow everything to be spontaneously embedded into our surroundings. Currently, there are many research projects for making the environment intelligent to exploit the technologies. However, the in-telligent environment that the projects try to achieve has not been realized at the moment. We consider one reason for this is the installation cost. Enabling technologies that are developed in current ubiquitous computing researches re-quire complex infrastructures embedded into our environment, that increase the deployment cost. Also, a new type of devices that requires a user to learn its usage may provide his/her burden.
Our approach uses intelligent daily objects and personal devices. This makes it possible to make our environments smart in an incremental way by replacing existing objects to new intelligent objects whenever a person wants. We are cur-rently developing sentient artefacts that are intelligent daily objects embedding computers and sensors. Also, we are implementing middleware infrastructures for personal devices that allow us to control embedded services in an easy way.
2
Computer mediated daily lives
Our future daily lives will be mediated by many computers containing various in-teraction devices and sensors. Our life style will be changed dramatically because
P. Flor´een, G. Lind´en, T. Niklander, K. Raatikainen (eds.): Proceedings of the Workshop on Context Awareness for Proactive Systems (CAPS 2005), Helsinki, 2005.
2 Tatsuo Nakajima, Kaori Fujinami, and Eiji Tokunaga
our daily lives are enhanced by blending computers in our real world. We call the future life style computer mediated daily lives, where computers will connect a variety of spaces such as home, office, town, car, train, bus, station, and airport in a seamless way. When we develop application software in computer mediated daily lives, it is important to take into account the behavior of human to reduce complexities in our daily lives and to increase pleasurability when interacting information appliances.
The number of information appliances will be increasing in the near future, but a user should not take into account some uninterested appliances by using the user’s preferences. In the past, information technologies were used to automate routine work. Now, too much information are given by the current information technologies. We believe that context-awareness and slow technologies [3] are important to make our daily lives more comfortable. For example, in the future, every object surrounding us, such as chairs and tables will be intelligent [4], but the intelligence may not be useful to make our lives convenient. To make our lives happy or pleasurable, intelligent everyday objects may represent their presence in an artistic way. The approach may be effective to a user for his entertainment. However, the representation should be customized according to a user’s preferences to reduce the amount of information in our daily lives, and the customization needs to be offered by middleware infrastructures to develop applications that support the design of a user’s experience in an easy way.
A key to realizing computer mediated daily lives is to make our environment smart or intelligent in an incremental and spontaneous fashion by coordinating many specialized information appliances and devices. For example, when we wish to add a new sensor device, we do not like to modify or replace existing devices. We wish to use a new information appliance without reconfiguring existing ap-pliances. The characteristic is very important in creating an infrastructure for building practical computer mediated daily lives.
3
Sentient artefacts
We are developing augmenting daily objects with computing capabilities like sen-sors and actuators, that capture a user’s situation,context, and provide him/her its awareness in a natural and implicit way. We call the daily object a “
sen-tient artefact”. We use a sentient artefact as a daily object that has inherent
functionalities as usual. In addition, it detects its state-of-use and utilizes the information as an input to a system. For example, a sentient door and sentient chair are utilized as an ordinary door and chair respectively. However, a sys-tem perceives a user’s presence and state, e.g. inside the room, sitting on the chair, etc., and changes its behavior according to the contextual information. We consider various services that utilize a user’s context will be embedded into our surroundings in the near future, rather than provided through traditional computers. Thus, we believe the sentient artefact approach allows a developer to build context-aware applications easily. Also, from a user’s points of view,
Building Intelligent Environments Using Smart . . . 3
Fig. 1.Sentient Artefacts.
he/she can utilize a context-aware service implicitly and naturally through the interaction with various sentient artefacts.
A sentient artefact has the following advantages.
– A smart space can be incrementally built by increasing numbers of sentient artefacts based on the requirements of an application.
– A user does not need to learn its usage because it keeps its metaphor and does not change its original functionalities.
– It can extract a user’s context implicitly and naturally through the original usage.
We have developed several sentient artefacts that are shown in Figure 1. In the following sections, we describe their brief overviews.
Sentient Chair: We use a chair for doing something with sitting on it. The
activity at the chair is narrowed in conjunction with the type and/or location of the chair. For example, a chair in an office is utilized for supporting a user’s work, while another chair in the kitchen is for eating or chatting. The change of states, i.e. from not sitting to sitting and vice versa, is a trigger for a system to invoke a specific service.
We have developed three types of sentient chairs that can extract different types of contextual information: 1) only an object’s state, i.e. being put by
4 Tatsuo Nakajima, Kaori Fujinami, and Eiji Tokunaga
something or not, 2) a person’s sitting or not sitting with adverbial states, and 3) an approximate direction of the face of a user. The first case was built by simply augmenting a stool with pressure sensors on the seat, where it does not distinguish a person from other objects. In the second case, an office chair with a back seat was adopted, and pressure sensors were attached properly so that the state of legs is extracted. Also, a photo sensor, a pressure sensor and a touch sensor were attached to detect the usage of the back seat. Finally, an office chair was augmented with an electric compass that detects terrestrial magnetism and specifies the direction of the back seat of the chair. These types of chairs show essential design issues on the heterogeneity of types and levels of abstraction of extracted context.
AwareMirror: AwareMirror is an augmented mirror that displays information
relevant to a person in front of it on the periphery of his/her sights [1]. The augmentation is natural because we usually “use” a mirror to know our internal state, e.g. health, as well as external state, e.g. looks, through the reflection on the surface of it, that is easily extended to virtual features, e.g the schedule of the day, the weather forecasting to the destination. Such virtual features is considered to be contextual information. AwareMirror suggests an ideal interaction between a person and a computer embedded into our daily lives. The state-of-use of a mirror is defined as the detection of something in front of it, that has been realized by two infra-red range finders in consideration of a feeling of privacy violation. It is not sufficient to identify a user by the mirror itself. The detection of utilization of a co-located sentient toothbrush is utilized for this purpose. This is an example of the artefact integration to extract more meaningful information.
Sentient Personal Device: Recently, a personal device like a cellular phone
and personal digital assistant (PDA) becomes popular in our daily lives. So, we have included it into a sentient artefact. In contrast to other artefacts described above, a personal device is used for general purpose that means the usage mostly depends on an application running on the device. Therefore, different types of sensors should be included for various requirements on extracting a user’s con-text.
We have developed an experimental personal device called Muffin in coopera-tion with Nokia Research Center Tokyo. Muffin has five categories of 15 sensors, including 1) a temperature, relative humidity, barometer , and alcohol sensor for retrieving environmental context, 2) a heart rate (pulse), galvanic skin resistance (GSR), and skin temperature sensor for extracting physiological context, 3) an electronic compass, global positioning systems (GPS), and ultrasonic range finder for retrieving location context, 4) a three-axis accelerometer and grip sensor for reasoning activity context, and finally 5) a radio frequency identification (RFID) tag reader, two video cameras, microphone for general recognition. Also, Muf-fin has actuation functionalities including an audio speaker, vibrator, and liquid crystal display (LCD).
These sensors are utilized for two reasons: 1) implicit and 2) explicit context extraction. For example, if a user is holding Muffin and facing to the display, then
Building Intelligent Environments Using Smart . . . 5
he/she would be happy if information is presented using visual cues, e.g. blinking LED, displaying a pop-up, etc. On the other hand, the fact of just holding and walking should let a user be notified by the vibration of the terminal. These sorts of state-of-use are obtained implicitly through its usage for other purposes. In the latter case, the terminal is utilized as a tool for processing gesture inputs, that means that a mobile terminal is used to control some applications explicitly by using gesture inputs such as shaking or rotating.
4
Personal coordination server
In the future, there are many applications in our surrounding environments. It is important to control these appliances in an easy way. The middleware infrastructure called a personal coordination server [7] allows us to aggregate them by using a personal device. While designing the middleware, we take into account the following issues.
– We like to control appliances from various presentation documents such as HTML and Flash.
– A way to control appliances is changed according to a user’s preference. Our middleware offers high level abstraction to specify appliances that a user likes to control, and each user’s personal device contains rules for personalizing the control of appliances. Since a personal coordination server can be carried with a user, he can aggregate appliances by using the same preferences in a seamless way anytime anywhere.
4.1 Basic Architecture
A personal coordination server that is carried by a user is implemented in a per-sonal device like a cellular phone, a wrist watch, or a jacket. Thus, the server can be carried by a user anytime anywhere. The personal coordination server collects information about appliances near a user, and creates a database storing infor-mation about these appliances. Then, it creates an HTML-based presentation document containing the attributes of appliances and the commands to con-trol them. A display near the user also detects the personal coordination server, and retrieves the presentation document containing the automatically generated user interface. The display shows the presentation document on the display. The document contains URLs (Uniform Resource Locator) embedding the attributes of appliances and their commands specified by using our URL-based naming scheme. Also, the presentation document is customized according to the user’s preference. When the user touches the display, a URL containing the attributes of an appliance and its command is transmitted to his/her personal coordina-tion server via the HTTP protocol. The server translates the URL to a SOAP command by accessing a database containing information about the appliance that s/he likes to control. Finally, the SOAP (Simple Object Access Protocol) command is forwarded to the target appliance.
6 Tatsuo Nakajima, Kaori Fujinami, and Eiji Tokunaga
4.2 Spontaneous Appliance Detection
URL-based Naming Scheme: Our framework allows a user to access one or
more appliances through a personal coordination server. We introduce a URL-based naming convention for specifying and controlling appliances. In our ap-proach, by embedding the attributes of appliances and commands in URLs, an HTML-based presentation document can be used to control appliances. The con-vention is defined within the standard URL but the path elements of the URL form can contain some additional information.
The URL definition is very flexible because we can specify various attributes to identify a target appliance. We can also use attributes that represent con-text information such as location. A personal coordination server can select an appliance in a context-aware way.
Service Management: In our system, the service management module in
a personal coordination server knows respective appliances via SSDP, and re-trieves service specification documents represented as RDF (Resource Descrip-tion Framework).
The service database in a personal coordination server contains all service specification documents detected currently. It contains a link to a WSDL (Web Service Description Language) document identifying commands that can be ac-cepted. If an appliance contains several functionalities, its specification document may contain several links to WSDL documents. Also, attributes of the document are used to identify a target appliance.
Personalization Management: A personal coordination server customizes a
presentation document according to a user’s preference encoded in a preference rule. Now, a personal coordination server detects several types of light appliances. We assume that a rule to filter light alliances whose type is not a ceiling light is stored in the personal coordination server. The presentation document contains only ceiling lights that reside in a room where a user is. A personal coordination server omits information about other types of light appliances. The preference rule is encoded in an everyday object, and it can be registered in a personal coordination server by closing the object to a user’s personal device.
4.3 An Application Scenario
In this scenario, we assume that surrounding various objects will embed RFID tags. Since these tags contain different preference rules, the customization is changed according to objects near a user. For example, if a child holds a stuffed animal that contains tags, the rules encoded in the tags are registered in the child’s personal coordination server. S/he can customize how to control informa-tion appliances by changing stuffed animals that s/he holds currently.
For example, we assume that a child is holding a stuffed toy dog. The dog contains a tag including a rule for selecting televisions because the child believes that the dog likes to watch televisions. Thus, a display shows a user interface to
Building Intelligent Environments Using Smart . . . 7
control televisions. On the other hand, when the child holds a stuffed toy rabbit, the display shows a user interface for music players because she believes that the rabbit likes to listen to music.
The tags can also be embedded in our daily goods like clothes, accessories, and shoes. Especially, a young person usually wants to put on these goods according to his/her feeling or emotion everyday. The goods reflect his/her preferences and current mental condition either consciously or unconsciously. Customizing services depending on what a user puts on today makes his/her daily lives more pleasurable.
5
Vidgets interaction framework
TheVidgets (Virtual Tangible Widgets) interaction framework enables a user to
control everyday embedded services in our surroundings using his/her personal mobile device equipped with several sensors and a camera. A user searches and selects the services through the camera, and actions and feedback through the sensors offer virtually implemented tangible user interfaces.
The vidgets interaction framework combines the following two features. – Tangible intense feelings enabling a user to intuitively interact with everyday
services.
– Universal remote controller capability to find and use these services by his/her personal device.
In the Vidgets interaction framework, we classify the use of the personal device into the following three stages, (1) searching services, (2) grasping one of the services and (3) using the service. Theseamless transition of these three stages offered by our framework is very important to improve the overall usability of the interaction with embedded services.
5.1 Seamless Transition
In the vidgets interaction framework, we assume that a user has his/her personal device equipped with a camera, several physical sensors such as an accelerometer and a force sensor, a handy size display and a speaker. We call such a personal device Vidgets Universal Controller(VUC). We also assume that everyday ser-vices deployed in ubiquitous computing environments equip with visual tags. We call the visual tags indicating respective services vidgets, and the services are calledVidgets Ubiquitous Services (VUSs).
Searching Vidgets: The first stage is the Searching Vidgets stage. In this
stage, a user looks for vidgets represented as recognizable visual tags through a camera installed on his/her VUC. We choose a visual tagging technique to develop our framework not other tagging technologies such as RFID, because they are probably the least expensive and most flexible. It is possible to create
8 Tatsuo Nakajima, Kaori Fujinami, and Eiji Tokunaga
them by usual printers, attach them to almost anywhere, display them on almost any digital screens, and recognize them with a small and mobile camera.
When a user’s VUC’s camera recognizes a vidgets visual tag, a display shows a superimposed image on the captured image. The superimposed image indicates “what it controls” and “how a user interacts with the service indicated by the
image”.
Grasping Vidgets: The second one is the Grasping Vidgets stage. When a
user finds out a vidgets visual tag, he/she can grasp it with his/her VUC. When the VUC recognizes a user’s grasping action, it initializes to download a mobile code to control a service indicated by the visual tag, that should be downloaded asynchronously to the VUC. The downloaded mobile code provides interactive visual and sound feedback that indicates grasping the vidget, ready to use and how to control the service.
The grasping action should be handled by physical sensors or buttons at-tached to the VUC. Although it can be handled by traditional GUI systems, such as pen-based GUI, we should not use such traditional user interaction tech-niques. The interaction modality in the former searching stage and pen-based GUI is quite different. Thus, the stages transit seamlessly. The grasping action should be more physical or tangible in which a user can change the stages from the search stage to the grasp stage seamlessly.
Using Vidgets: The third one is the Using Vidgets stage. While a user is
holding on the vidget grasped in the previous stage, he/she can interact with it by using physical sensors attached to his/her VUC. For example, an accelerometer sensor can handle not only traditional widget-like interaction such as buttons and scrollbars but also virtual physical actions such as shaking, rolling and tilting. Reacting to these actions, VUC provides real-time visual and sound feedback that indicates the current state of the virtual physical controller and improves a user’s affordance.
When a user stops to grasp the vidget by releasing the sensor used for grasp-ing the vidget, the use stage is terminated and transits to the first stage. In order to make the overall interaction between VUC and services comfortable and in-tuitive for a user, the interaction cycle should seamlessly transit through these stages.
5.2 Prototype Implementation
We have implemented the prototype system realizing the vidgets interaction framework. VUC, as described in the previous section, is a personal device equipped with a camera and tangible interactive capabilities. VUC finds and interacts with VUSs that are services embedded in the real world. Each VUS is deployed with its own Vidget ID that is a recognizable tag representing the VUS’s information. It registers the mappings of its ID andVidgets Mobile Code
Building Intelligent Environments Using Smart . . . 9
Fig. 2.Vidgets Interaction Framework.
vidgets mobile code from each service that handles tangible actions, sends con-trol events, and returns feedback actions to a user. When a user finds and grasps a certain Vidget ID, the user’s VUC dynamically loads the corresponding vid-gets mobile code and initializes it. Each component hasVidgets Communication
Infrastructure and the downloaded code submits control events to the service
through it.
Figure 2 shows our prototype implementation of Vidgets Universal Con-troller. We adopt Sony VAIO type U (VGN-U50) that is a keyboard-less micro PC as the base of the VUC. It provides real-time visual feedback on its display and sound feedback using the build-in speaker. The visual marker detection and mobile augmented reality are realized by using a digital camera. We currently adopt ARToolKit that is a software library for detecting a visual marker and superimposing a 3D image on the detected marker. The superimposed image is described by VRML, and registered by VUS. A force sensor transmitting a grasp-ing action and an accelerometer producgrasp-ing several tangible actions are composed in one sensor box.
One of vidgets ubiquitous services that we have implemented is Media Dialer that enables a user to play (or pause), fast-forward (or rewind), and select movies by interacting with three types of vidgets shown in Figure 2. The first vidget, the left visual marker, is used to play or pause a movie on the screen. If the movie
10 Tatsuo Nakajima, Kaori Fujinami, and Eiji Tokunaga
is not playing, the screen shows the play visual marker superimposed by the 3D play button image. On the other hand, if the movie is playing, it shows the pause visual marker, and it is superimposed by the pause button image. When a user grasps each vidget, he/she can play or stop the movie.
The second vidget, the middle visual marker, is used for fast-forwarding or rewinding a movie. While grasping the vidget, the dial shaped image is shown on VUC. When the user tilts his/her VUC, the dial is revolving according to its gradient. The more the user tilts it to the right, the faster the movie is forwarding. On the other hand, the more the user tilts it to the left, the faster the movie is rewinding.
The third vidget, the right visual marker, allows a user to select movies on its screen. While grasping the vidget, the service shows the menu for selecting movies, and the dial controller is shown on his/her VUC. If a user tilts the VUC to the right, the next movie is focused on. Similarly, if the user tilts it to the right, the previous movie is focused on. If the user keeps tilting his/her VUC, the focused movie is switched continuously.
6
Conclusion and future direction
In the future, a lot of services will be embedded in our surroundings. We believe that our vidgets framework is very effective to detect and choose the services. Also, sentient artefacts are very useful to retrieve context information about us in a natural way.
We are also working three topics currently. The first topic is to develop a middleware infrastructure to integrate sentient artefacts [2]. It enables us to develop services using several sentient artefacts in an easy way. The second topic is to develop the world model for developing context-aware services easily. We believe that the appropriate world model is very important to develop context-aware services that are natural to users. The final topic is to collect experience with building middleware infrastructures for ubiquitous computing [5,6]. We especially believe that it is important to consider the effect of human factors in the design of middleware infrastructures.
We found that the large gaps between the assumptions of software infras-tructures and the requirements of applications cause serious problems to support multiple applications. Especially, the gap between the description of the context and the corresponding real objects is a serious source of the difficulties to ac-cept applications developed by ubiquitous computing research communities. We need to investigate how to tame the gap in the future to build useful ubiquitous computing environments.
References
1. K.Fujinami, F.Kawsar, T.Nakajima, “Aware Mirror: A Personalized Display using a Mirror”, In Proceedings of Pervasive 2005.
Building Intelligent Environments Using Smart . . . 11 2. K.Fujinami, T.Yamabe, T.Nakajima, “Bazaar: A World Model with Self-Descriptive Objects for Context-Aware Applications”, In Proceedings of 2nd International Sym-posium on Ubiquitous Computing Systems, 2004.
3. L.Hallnas, J.Redstrom, “Slow Technology - Designing for Reflection”, Personal and Ubiquitous Computing, Vol. 5, No. 3, 2001.
4. L.E.Holmquist, H.W. Gellersen, G.K.Kortuem, A.Schmidt, M.S.Stohbach, S.A. An-lifakos, F.M. Michahelles, B. Schiele, M. Beigl, R. Maze, “Building Intelligent Envi-ronments with Smart-Its”, IEEE Computer Graphics & Applications Jan/Feb 2004. 5. T.Nakajima, D.Ueno, “Experiences with Building Middleware Infrastructures for Home Computing on Commodity Software”, In Proceedings of 10th International Conference on Real-Time and Embedded Computing, Systems, and Applications, 2004.
6. T.Nakajima, K.Fujinami, E.Tokunaga, H.Ishikawa, “Middleware Design Issues for Ubiquitous Computing”, In the 3rd International Conference on Mobile and Ubiq-uitous Multimedia, 2004.
7. T.Nakajima, “Personal Coordination Server: A System Infrastructure for Designing Peasurable Experience”, In Proceedings of International Conference on Pervasive Services, 2005.
8. E.Tokunaga, T.Nakajima, “Virtual Tangible Widgets: A Seamless Universal Inter-action with a Mobile Sensing Device”, Technical Report, Waseda University, 2005.
13
What Is Wrong with Maria?
Privacy Concerns in the ISTAG Scenario
Olli Pitk¨anenHelsinki Institute for Information Technology HIIT
1
Introduction
Scenarios are important tools to systematically analyze the circumstances related to future information and communication technologies. They are often helpful in requirement elicitation, but they can also assist in analyzing possible business models and legal challenges. [8,9,10]
One of the best known examples of ICT scenarios is the work that the IST Advisory Group (ISTAG) has conducted. ISTAG has tried to get a higher level of focus and a higher pace of development in Europe on Information and Com-munication Technologies. As a part of this work, ISTAG launched a scenario planning exercise in 2000. The scenarios were developed by the IPTS (part of the European Commission’s Joint Research Centre) in collaboration with DG Information Society and with the active involvement of 35 experts from across Europe. The aim was to describe what living with ‘Ambient Intelligence’ might be like for ordinary people in 2010. [4]
ISTAG scenarios can be considered quite optimistic — even unrealistic what comes to the pace of technological development. Yet, they present the European vision of high-tech development in the field of ambient intelligence. As they have been produced by the advisory group of European Commission, they also portray the somewhat “official” image of the future. Therefore, it is interesting to analyze the scenarios more in detail and especially compare them to the requirements that European legislation states. In this paper, I am focusing on one of the scenarios and — although many legal areas are related to the scenario — I am analyzing it particularly from the privacy and data protection viewpoint.
ISTAG scenario that I am concentrating on in this paper isMaria (see Ap-pendix 1). It is a scenario about a busy business person traveling abroad and using highly automated communication systems. From the legal point of view it is notable that most transactions — both private and public — are automated; hardly any human interaction is required. The machines make very significant decisions, like they seem to decide who is allowed to enter a country, and they make binding contracts on behalf of someone else. The context sensitive services utilize a lot of personal information. Especially location information is needed for many services.
P. Flor´een, G. Lind´en, T. Niklander, K. Raatikainen (eds.): Proceedings of the Workshop on Context Awareness for Proactive Systems (CAPS 2005), Helsinki, 2005.
14 Olli Pitk¨anen
2
Privacy and Data Protection Law
By its very nature, a significant proportion of information managed by ubiquitous systems is private. People object to information on their location, behavior, habits, interests, transactions, finance, social situations, and health among other, be communicated without their consent. Therefore, a priority for ubiquitous systems is privacy and confidentiality protection.
On the other hand, many companies and public agencies are keen to ac-cess such data. For example, a commercial company would be able to direct marketing to the right individuals much more accurately if they held detailed information about their habits and circumstances. Some customers might be willing to benefit from the situation while others are so concerned about their privacy that they would not even consider letting this information be utilized under any circumstances.
The European Union has set rather strict rules regulating the use of personal data via the data protection directives. However, in other regions, for example in the USA, the discussion about privacy protection has not led to comparable statutes thus far often opting for self regulation of markets. It remains unclear which approach ultimately will prove to be more attractive: although privacy is perceived as an extremely important aspect of legal systems and must be protected appropriately, too strict a privacy protection legislation may lead to unintended results with useful services for consumers being simply unsustainable in the marketplace.
The ambient intelligence technologies described in Maria scenario represent huge challenges to privacy. The interconnected computing devices must have access to a large amount of private information to be able to provide the services. This might poses severe risks to privacy. The scenario does not refer to any such problems: the system is working perfectly and it honors the users’ privacy. Nothing however ensures that. If the system has so much private information about people, it is very easy to — intentionally or by mistake — use it wrongfully or distribute it too widely. Actually, often the best solutions from the purely technical point of view are unacceptable from privacy perspective. For example, access control mechanisms that prohibit unauthorized use of information are complex to implement and decrease the overall performance and usability of a system. Therefore it is often tempting to leave such mechanisms away or at least make them as light as possible. Unless a paying customer insists or a law requires, a system provider easily ignores privacy protection.
In the European Union, several directives and other statutes — notably Data Protection Directive (95/46/EC ) [2] and Directive on Privacy and Electronic Communications (2002/58/EC) [3] — have been introduced to protect privacy and personal information. However, the autonomic nature of ubiquitous comput-ing has implications that cannot be adequately addressed by existcomput-ing legislation. An example of this situation relates to the use of location data to provide es-sential context for many ambient intelligence services. Processing such location information falls under the provisions of the directive which requires explicit consent by the user. In an ubiquitous commerce environment where a number
What Is Wrong with Maria? 15
of services by different service providers are used in tandem it is difficult to notify and receive the consent of users to process location data every time this is necessary. In fact, the directive requires that the user accepts separately the use by each service of their data and even more, services must provide continu-ally the “possibility, [of] using a simple means and free of charge, of temporarily refusing the processing of such data for each connection to the network or for each transmission of a communication.” Indeed, it appears that this aspect of the directive completely excludes the possibility of federated service provision. Moreover, if users are required to accept separately the use of their private data by each service then in practice, it is most likely that users simply would not use the services rather than accept this management overhead. Surely, usability improvements and automatic mechanisms can make the situation much easier, but ultimately the user must have control and the ability to refuse the process-ing of location data in order to fulfill the requirements of the directive. Finally, while the directive aims to harmonize legal systems and guarantee certain level of protection within the EU, obviously it does not apply in countries outside its boundaries.
In Maria scenario for instance, it seems that most services would benefit
from her location data. However, the situation becomes very complex if Maria needs to accept separately each service to use the data, and each service must provide her with the continuing “possibility, using a simple means and free of charge, of temporarily refusing the processing of such data for each connection to the network or for each transmission of a communication”. In practice, it would probably be easier for Maria simply not to use the services. Surely, usability studies and automatic mechanisms can make the situation much easier, but ultimately the user must have control and the ability to refuse the processing of location data in order to fulfill the requirements of the directive.
The directives aim at harmonizing legal systems and guarantee certain level of protection within the EU, but obviously they do not apply in countries outside the Union. Therefore, exchanging information within the Union has been tried to make flexible. On the other hand, it is highly restricted to transfer personal data from the member countries to “unsafe” countries outside the Union.
InMariascenario, European citizen is traveling outside Europe. Her personal data mainly originates from the Union but is needed in Asia. Presumably Maria is willing to use those personalized services and therefore accepts the transfer of her personal data between at least her home-country and the Asian country. Yet, in accordance with the directives and European national laws, she has to explicitly accept the transfer of data from Europe to the Asian country. This effectively protects her privacy, but introduces severe challenges to the designers of the services. Also, it decreases the efficiency of the concept that was emphasized by ISTAG. According to ISTAG, “Ambient Intelligence works in a seamless, unobtrusive and often invisible way.” The need to get consent from the user makes this goal hard to achieve.
When developing scenarios, it is often difficult to estimate how much change can occur in a given time period. According to WIIO’s law, people tend to
16 Olli Pitk¨anen
overestimate the near future and to underestimate the far future. [7] With respect
toMaria scenario, it seems unlikely that technology could develop that far in a
few years, but it is even more improbable that the legal systems could change by 2010 that much. Especially regarding privacy and data protection, the legislation in Europe has gone in a more restrictive direction in the recent years. While protecting most important values it has not made this kind of a scenario easier to achieve. At least, all this will not happen by 2010 as the scenario suggests.
3
Balancing Interests
Mobility, context-awareness, and ubiquity will bring computer networks even into the most intimate places and walks of life. As the scenario illustrates, future computing and communication devices are not only capable of accessing people’s private information but many useful services are highly dependent on it. There will an increasing dilemma: people are requesting and can benefit from services that jeopardize their privacy.
Probably, the opposite interests of getting useful services and protecting pri-vacy tend to seek balance. People are willing to disclose reasonable amount of private information to get the services they need, but not more than that. Cer-tainly some people are more cautious of their privacy while some others do not care even if quite a lot of information on them is available for others. It certainly depends also a lot on the situation, social context, the services, and other factors, how much somebody is willing to disclose. Privacy is most relative. In a busi-ness meeting everybody is expected to introduce oneself while in an anonymous discussion group it is acceptable to use a pseudonym. Usually, it is at one’s own discretion how much personal information she or he is willing to reveal. [1]
On the other hand, for certain service providers there may be incentives to collect as much private information from people as they can, because that information can be worth a lot of money. Also, as discussed above, it is often more difficult and expensive to build technical systems that secure private information than ignore privacy needs. Therefore service providers easily disregard privacy unless customers insist it or a legal system forces them to honor people’s privacy. The recent changes in legal systems, like European directives on data protec-tion, have substantially improved privacy protection. Some of the chosen actions, however, make it very difficult to develop services that users would like to have. For example, ISTAG Maria may turn out to be unfeasible, because it is very difficult to implement the services in an efficient but legal way.
In summary, the opposite interests in privacy and useful services need to be balanced. The legal system needs to take into consideration both sides and also understand what is technically feasible. Challenges to privacy are much greater than ever before.
What Is Wrong with Maria? 17
4
Other legal areas
Maria scenario also has difficulties in other legal areas besides privacy. In this paper, I am not going into details in those other areas, but I am simply pointing out that according to the scenario all the immigration and border control proce-dures as well as negotiation processes have been automated. In most cases, legal systems assume that human beings make the final decisions: there is a boarder guard to decide who is allowed to enter the country; human representatives of legal entities make agreements on behalf of the organizations, and so on. Those human beings are able to make the decisions even using incomplete information and taking reasonable risks, they are authorized to use due liberation, and even-tually they are also responsible for their decisions. It will take a long time before — if ever — computer systems are able to deliberate complex problems, make answers with incomplete input information, and take intentional risks. There are fundamental difficulties to make a machine liable for its decisions.
Additionally, extensive processing of information is always related to intel-lectual property rights (IPR). In this case, especially copyright and database protection questions are important. (e.g. [6])
5
Conclusion
Several factors affect the importance of privacy in relation to ambient intelli-gence. First, new technologies make privacy increasingly vulnerable. It is a fun-damental property of many information and communication technologies that they need to manage information that can be related to individuals.
Second, private information can be used to improve and add value to in-formation products and services. Maria scenario describes situations in which the end-users benefit from the fact that a product or a service can be adapted in accordance their personal circumstances. A context-aware service or a prod-uct that is adapted according to end-user profile can — at its best — provide end-users with exactly the service or information that is needed in a certain situation. While communication networks rapidly grow, the amount of available information or the information flood explodes. End-users are progressively more in need of technologies that filter away unnecessary noise and give them only the information they want. Therefore, the avail of personal information is not only sometimes acceptable, but increasingly necessary to let the end-users ben-efit from the possibilities that the information and communication technologies offer them.
Third, taking care of end-users privacy presents additional costs. Unless cus-tomers insist or laws require, it does not usually make sense for commercial companies to take measures to secure private information. Sometimes a com-pany may get competitive advantage by announcing that it follows a privacy policy, which protects end-users’ information. Usually, though, individuals are not that concerned about their privacy that they would require companies to take the extra steps to protect it, nor get companies much advantage from voluntary
18 Olli Pitk¨anen
privacy policies. Therefore, the laws of economics make the rational companies to ignore privacy protection as much as they legally can — or even as long as they don’t get caught.
In conclusion, it is mostly up to lawmakers to find the right balance between privacy needs and the useful avail of personal information. As of now, they have not quite succeeded. In the European Union, the laws are too numerous and com-plex, they cover unnecessary wide spectrum of information and circumstances, they include unnecessary provisions, and sometimes they may even harm use-ful businesses as shown above. If ISTAG Maria scenario is to represent a set of eligible future products and services, the laws should not prevent it as they seem to do today. On the other hand, in most countries data protection laws are too weak or even non-existing. To support the development of information and network society, a decent minimum level is required.
References
1. A. Acquisti and J. Grossklags. Privacy Attitudes and Privacy Behavior: Losses, Gains, and Hyperbolic Discounting. In: J. Camp and S. Lewis (Eds) The Economics of Information Security. Kluwer Academic Publishers. 2004.
2. Directive 95/46/EC of the European Parliament and of the Council of 24 October 1995 on the protection of individuals with regard to the processing of personal data and on the free movement of such data, 1995.
3. Directive 2002/58/EC of the European Parliament and of the Council of 12 July 2002 concerning the processing of personal data and the protection of privacy in the electronic communications sector (Directive on privacy and electronic communica-tions), 2002.
4. K. Ducatel, M. Bogdanowicz, F. Scapolo, J. Leijten, J-C. Burgelman (eds.). ISTAG Scenarios for Ambient Intelligence in 2010, Final Report, IPTS-Seville, 2001. 5. J. Hong and J.A. Landay. A Context/Communication Information Agent. Personal
Technologies. Special Issue on Situated Interaction and Context-Aware Computing. 2001.
6. R. Merges, P. Menell and M. Lemley. Intellectual Property in the New Technological Age. 2nd edition. Aspen Law & Business. 2000.
7. G. M˚artenson, M. Otala, O. A. Wiio. Tietotekniikka 1990-luvulla (in Finnish). Sitra, B-78, 1985.
8. O. Pitk¨anen. Legal Challenges to Ubiquitous Commerce. In: G. Roussos (ed.), Ubiq-uitous and Pervasive Commerce: New Frontiers for Electronic Business, Springer SMB, 2005.
9. O. Pitk¨anen, M. M¨antyl¨a, M. V¨alim¨aki and J. Kemppinen. Assessing Legal Chal-lenges on the Mobile Internet. International Journal of Electronic Commerce 8(1):101-120. 2003.
10. B. Wendell. Foundations of Futures Studies. Volumes I-II. Transaction Publishers. 1997.
What Is Wrong with Maria? 19
Appendix 1: ‘Maria’ — Road Warrior Scenario
After a tiring long haul flight Maria passes through the arrivals hall of an airport in a Far Eastern country. She is travelling light, hand baggage only. When she comes to this particular country she knows that she can travel much lighter than less than a decade ago, when she had to carry a collection of different so-called personal computing devices (laptop PC, mobile phone, electronic organisers and sometimes beamers and printers). Her computing system for this trip is reduced to one highly personalised communications device, her ‘P—Com’ that she wears on her wrist. A particular feature of this trip is that the country that Maria is visiting has since the previous year embarked on an ambitious ambient intelli-gence infrastructure programme. Thus her visa for the trip was self-arranged and she is able to stroll through immigration without stopping because her P-Comm is dealing with the ID checks as she walks.
A rented car has been reserved for her and is waiting in an earmarked bay. The car opens as she approaches. It starts at the press of a button: she doesn’t need a key. She still has to drive the car but she is supported in her journey downtown to the conference centre-hotel by the traffic guidance system that had been launched by the city government as part of the ‘AmI-Nation’ initiative two years earlier. Downtown traffic has been a legendary nightmare in this city for many years, and draconian steps were taken to limit access to the city centre. But Maria has priority access rights into the central cordon because she has a reservation in the car park of the hotel. Central access however comes at a premium price, in Maria’s case it is embedded in a deal negotiated between her personal agent and the transaction agents of the car-rental and hotel chains. Her firm operates centralised billing for these expenses and uses its purchasing power to gain access at attractive rates. Such preferential treatment for affluent foreigners was highly contentious at the time of the introduction of the route pricing system and the government was forced to hypothecate funds from the tolling system to the public transport infrastructure in return. In the car Maria’s teenage daughter comes through on the audio system. Amanda has detected from ‘En Casa’ system at home that her mother is in a place that supports direct voice contact. However, even with all the route guidance support Maria wants to concentrate on her driving and says that she will call back from the hotel.
Maria is directed to a parking slot in the underground garage of the newly constructed building of the Smar-tel Chain. She is met in the garage by the porter — the first contact with a real human in our story so far! He helps her with her luggage to her room. Her room adopts her ‘personality’ as she enters. The room temperature, default lighting and a range of video and music choices are displayed on the video wall. She needs to make some changes to her presentation — a sales pitch that will be used as the basis for a negotiation later in the day. Using voice commands she adjusts the light levels and commands a bath. Then she calls up her daughter on the video wall, while talking she uses a traditional remote control system to browse through a set of webcast local news bulletins from back home that her daughter tells her about. They watch them together.
20 Olli Pitk¨anen
Later on she ‘localises’ her presentation with the help of an agent that is spe-cialised in advising on local preferences (colour schemes, the use of language). She stores the presentation on the secure server at headquarters back in Eu-rope. In the hotel’s seminar room where the sales pitch is take place, she will be able to call down an encrypted version of the presentation and give it a post presentation decrypt life of 1.5 minutes. She goes downstairs to make her pre-sentation. . . this for her is a high stress event. Not only is she performing alone for the first time, the clients concerned are well known to be tough players. Still, she doesn’t actually have to close the deal this time. As she enters the meeting she raises communications access thresholds to block out anything but red-level ‘emergency’ messages. The meeting is rough, but she feels it was a success. Com-ing out of the meetCom-ing she lowers the communication barriers again and picks up a number of amber level communications including one from her cardio-monitor warning her to take some rest now. The day has been long and stressing. She needs to chill out with a little meditation and medication. For Maria the medi-tation is a concert on the video wall and the medication. . . a large gin and tonic from her room’s minibar.
21
Designing and Implementing a System for the
Provision of Proactive Context-Aware Services
Tapio Pitk¨aranta1, Oriana Riva2, and Santtu Toivonen1
1 VTT Technical Research Centre of Finland
P.O. Box 1203, FIN-02044 VTT, Finland
{tapio.pitkaranta, santtu.toivonen}@vtt.fi
2 Helsinki Institute for Information Technology HIIT
P.O. Box 9800, FIN-02015 HUT, Finland
Abstract. In this paper we discuss our experiences in designing and
implementing an event-based system aimed to support the provision of proactive context-aware services. We present our findings about the crit-ical design decisions of such system and give a preliminary performance evaluation of the system prototype. Such analysis permits us to identify crucial performance checkpoints of our system to be taken into consid-eration in the future work.
1
Introduction
The ways of finding and utilizing content from the Web are constantly evolving. Using a desktop computer and a Web browser to access content in HTML-format has until now been the dominant mode of operating, and will likely continue to dominate in the foreseeable future. However, other ways to use the content are emerging and in some application areas fast replacing the browsing in its traditional sense. In DYNAMOS3 project, carried out jointly by VTT and HIIT, these emerging modes of operating with Web content are in a central role. The crux of the DYNAMOS project lies in the notion of context. Mobile users equipped with wireless devices go through several contextual changes as they move around in physical and social surroundings. These contextual changes are used as filters for providing the users with appropriate content. The content the users are provided with can be created by third party service providers, or by other users. Particularly, we investigate content that is originally generated by a service provider, but later on annotated by one or more users. These service an-notations can be for example numerical ratings or natural language descriptions, which transmit the users’ attitudes towards a certain service.
We call this approach ahybrid service provision model consisting of service provider to consumer interaction on one hand, and consumer to consumer on the other hand. Figure 1 depicts the approach on a general level. A more detailed
3Dynamic Composition and Sharing of Context-Aware Mobile Services URL:http://www.vtt.fi/tte/proj/dynamos/
P. Flor´een, G. Lind´en, T. Niklander, K. Raatikainen (eds.): Proceedings of the Workshop on Context Awareness for Proactive Systems (CAPS 2005), Helsinki, 2005.
22 Tapio Pitk¨aranta, Oriana Riva, and Santtu Toivonen
Fig. 1.Hybrid approach for service provision [1].
description can be found in [1]. During the project this model will be tested and evaluated from various angles. Among these, a user trial with boaters in Turku archipelago will be carried out in the summer 2006. This paper evaluates the system on a more technical level, concentrating on the performance of various software components constituting the system.
The rest of the paper is organized as follows. Section 2 presents related work for the research. Section 3 discusses the core concepts of the DYNAMOS sys-tem and gives some implementation insights. In Section 4, some performance measurements of the DYNAMOS system prototype are presented and discussed. Finally, Section 5 draws our conclusions and outlines the future work.
2
Background and Related Work
A lot of work has been done in the area of context-aware service provisioning in the past few years. The wide deployment of easily accessible mobile wireless communication networks and the rapid diffusion of small and powerful portable devices have promoted the concrete deployment of numerous location-based sys-tems. By simply integrating some positioning techniques, applications can em-ploy location information to provide variegated services. Examples of location-based applications are navigator assistants, location-aware information services such as weather services or tourist guides, location-sensitive games, and so forth. In particular tourist guide and navigation systems are rather relevant to our work given the maritime scenario we address. Some recent examples are GUIDE [2], which focuses on implementing a context-sensitive tourist guide for visitors in the city of Lancaster, and WebPark [3], which provides location-based services in protected and recreation areas, such as coastal, rural, and mountainous areas. However, most of existing location-based systems lack in addressing two im-portant features. First, as pointed out by a study on the potential benefits of
ad-Provision of Proactive Context-Aware Services 23
vances in mobile technologies for tourism [4], tourists typically benefit from shar-ing visitshar-ing experiences, and they usually do ”pre-visitshar-ing” and ”post-visitshar-ing”. Pre-visiting is about planning the visit; post-visiting is about reminiscing and sharing the experience. The support for sharing user experience along with the possibility of storing such content for future usage by the user herself, by her friends or anyone are central functionalities in the DYNAMOS System. The con-cept of attaching signs or leaving marks to visited environments reminds previous studies such as the tour of Disneyland application of Pascoe [5], GeoNotes [6], the K-Trek architecture [7], virtual graffiti [8], and InfoRadar [9].
Second, even though location is currently the most dominant context param-eter, advances in sensor technologies will soon allow the inclusion of more con-textual information in service adaptation, such as environmental conditions, user activity, social surroundings, and so forth. Sensors will be integrated in mobile de-vices or in the environment, and dede-vices will communicate with each other to ac-quire further accurate information. Many research activities have been dedicated to the integration of sensor technology, processing of sensor data, and final rep-resentation into context information. Specifically, context-infrastructures can be effectively employed to reduce the burden due to supporting context-awareness and to provide useful abstractions for application developers [10]. In addition they potentially provide typical middleware properties, such as interoperabil-ity, transparency, and modularity. Examples of existing context-infrastructure include the TEA framework [11], the Wireless Toolkit of Dey et al. [12], the Sen-tient Model [13], Gaia [14] and RCSM [15]. In addition to context parameters retrieved automatically with the aid of different sensors, some parameters can be manually or semi-automatically entered by the user. In DYNAMOS an example of such is the user activity, which plays a central role in the service-matching.
In DYNAMOS project we specifically target three usages of context informa-tion. First, context is employed in order to provide services in a proactive and context-aware fashion. Secondly, context is utilized when creating reminders for past or future events in the form of service annotations and user-generated notes. Third, information-sharing is context-based and is supported by the possibility of exchanging such user-generated content.
3
The DYNAMOS System
3.1 Core Concepts
The major functionality of the DYNAMOS System is notifying mobile users about services in a context-sensitive fashion. This is carried out by providing users withservice descriptionsgenerated by service providers andservice annota-tionscreated by other users. In addition to annotating services and sharing such annotations with each other, the users can create and share content that does not refer to any specific service description. This kind of content is calleduser notes. For instance, users can create notes in the form of warning messages (e.g.,
”Dangerous underwater rocks nearby!”), to notify other users of accidental situ-ations that occur in a certain location and time. Alternatively, these notes may
24 Tapio Pitk¨aranta, Oriana Riva, and Santtu Toivonen
Fig. 2.Overall System Architecture.
be plain notifications of some state of affairs, such as bird sightings. In a ma-rine setting, such messages can be sorted as emergency/urgency/safety/routine events.
Service descriptions capture the most essential semantics of the services. In other words, the services are categorized according to their business branches. Additionally, a natural language description about the services can be given. These descriptions are matched against user profiles when providing them with services. The same applies for service annotations; if a service annotation at-tached to a service potentially of interest to a user matches the user’s current context, it is provided to her. Among our future work is considering the user pref-erences (found in user profiles) and their impact on the provision of annotations as opposed to service descriptions, and vice versa.
3.2 Architectural Components
The overall architecture of the DYNAMOS System is depicted in Figure 2. The three main components of the system are the following:
– The Client:the client can be a regular Web browser (thin client) or a
stand-alone application (thick client) that uses event-based communication. Using events rather than for example RMI (as in [8]) has advantages such more fault tolerant communication and wider penetration of currently available compliant client devices.
– Content Servers:the actual content is stored in the DYNAMOS content
servers. The User Content Server stores user related information (i.e., pro-file and context data). The Service Content Server stores content provided by service providers (i.e., service descriptions) and user-generated content (i.e., user notes and service annotations). This provides us with flexibility in distributing the content through multiple channels.
– Distribution Servers:since we want to support different ways in accessing
such content, there are multiple types of access server, namely Web Servers and Event Servers.
The access to the DYNAMOS System is currently supported by means of two different interfaces. The Web interface can be used with any device that supports
Provision of Proactive Context-Aware Services 25
a web browser. However, the current Web interface is designed for devices with large screens. For mobile users often equipped with limited capabilities devices we implemented a stand-alone client application. Such application is context-aware, meaning that it constantly monitors context parameters such as location and activity of the user and notifies the Event Server of relevant context changes. The major benefit in doing this is that the mobile user usually equipped with small screen devices are proactively provided with a compact summary of available services, without the need of browsing the whole content stored in the system.
The DYNAMOS Content Servers contain the whole content of the system. The Distribution Servers use a middleware solution to access the Content Servers. In the current implementation we use Java RMI. We chose Java RMI since it outperforms most of the other middleware solutions, such as Web services [16] and it is also easy to implement and to maintain. The Content Server uses a combination of objects and relational databases to store and to offer the content to the Distribution Servers. Due to performance reasons most of the content is kept in memory and only storing content includes time consuming MySQL operations. To test our application with the selected sailing scenario we have currently provided the server with a database of tourist services in Turku area4. The size of the database is around one thousand service descriptions entries.
The Event Distribution Server we have integrated in our system is the Fuego Core event server [17]. It provides a standard event-based platform and allows us to accessing and updating the content stored in the Content Servers in an event-based manner. The reason for adopting an event-based infrastructure is twofold. First, a main property of context-aware applications is to react to con-text changes occurring in their execution environment. Such applications must be able to subscribe to context changes events and be notified when such events occur. Second, context changes and user generated content must be added and shared among users in an event-based manner in order to support an efficient and real time provisioning and sharing of such content.
3.3 DYNAMOS Event Distribution
Figure 3 shows the Fuego Core architecture and the DYNAMOS components so far implemented. The figure shows on the right side the DYNAMOS Event Server and on the left side the DYNAMOS client application. The Fuego Core stack provides us with an event-based distribution channel and several enhancements for wireless internet. More details can be found at the project web page [17]. In the following, we briefly describe the main DYNAMOS software components.
– TheUser Note and Annotation Listeners(both on the client and the server
side) are in charge of detecting the presence of new user-generated notes and service annotations and of notifying the upper layer. User-generated notes and service annotations are made available into the system respectively by
theUser Note Producerand theAnnotation Producer. Therefore, clients are
capable of adding and sharing such notes in a real-time manner.
26 Tapio Pitk¨aranta, Oriana Riva, and Santtu Toivonen
Fig. 3.DYNAMOS Event Distribution Server.
– The Context Listener (on the server side) listens to context changes (i.e.,
user profile and context changes) occurring on the client side and constantly update the server with the current user context and profile. Context and profile change notifications are produced on the client’s side by theContext
Manager.
– TheMatcher (on the server side) plays a key role in the DYNAMOS Server.
It performs the matching between active users and service descriptions thus to provide users with a compact summary of relevant services. The match-ing occurs between context/profile information of the user (e.g., interest, location) and service description (e.g., service category, location). The same matching applies to annotat
