SEAMLESS COMMUNICATION AND ACCESS TO

INFORMATION FOR MOBILE USERS IN A WIRELESS

ENVIRONMENT

Golha Sharifi, Julita Vassileva and Ralph Deters

University of Saskatchewan, Computer Science Department,

57 Campus Drive, Saskatoon, Saskatchewan S7N 5A9, Canada

Keywords: Mobile Devices, Agents

Abstract: Providing mobile workers with mobile devices such as a Compaq iPaq with a CDPD card can support them

in retrieving information from centralized information systems. More specifically, mobile devices can

enable mobile users to make notifications for schedule changes and add new data into the information

system. In addition these devices can facilitate group communication anytime and anywhere. This paper

presents different ways of providing non-critical information in a timely fashion for nomadic users of

mobile devices using a wireless network. A distributed application prototype to support nomadic users is

proposed, and a simulated environment is used to evaluate the prototype. Since solutions for seamless

access are highly domain specific, the study involves homecare workers at Saskatoon District Health

(SDH). By keeping track of the users’ current context (time, location etc.) and a user task model, it is

possible to predict the information needs of mobile users and to provide context dependent adaptation of

both the content and the functionality. Moreover, to avoid interrupts in the user’s interaction with the main

information sources, methods for mobile transactions management using agent-based smart proxies that

buffer, delay or pre-fetch information/data are introduced.

1 INTRODUCTION

The development of wireless technology and the

commercial success of small screen appliances,

especially cellular phones and Personal Digital

Assistants (PDAs) have advanced the development

of mobile and ubiquitous computing (Weiser, 1991).

Mobile and ubiquitous computing allows nomadic

users to remain “connected” without depending on a

wired infrastructure.

Nomadic workers who use mobile and small

screen devices often need access to existing

information services to enable them to browse

multimedia information anytime and anywhere with

ease. Nomadic workers in specific domains who

might require these services include healthcare,

emergency services, sales, and education. However,

a major problem for the nomadic user using a thin

client is the varying quality of the connection.

Though connections remain fairly stable and good in

areas close to network stations, for instance, Cellular

Digital Packet Data (CDPD) towers, it is not

uncommon to have zones of low connectivity or

even no connection due to physical structures, for

example, high concrete buildings blocking the

signals.

We have investigated different ways of enabling

seamless access of non-critical information for

nomadic users. In this paper we focus on the needs

of homecare workers of the Saskatoon District

Health (SDH). The rest of the paper is organized as

follows: in the next section the domain for this study

and the current problems faced by mobile users are

introduced; next, the general approach to solving

these problems is described, which is explained in

more detail in the third, fourth and fifth sections; the

sixth section outlines an evaluation procedure; a

comparison with other work is presented in the

seventh section; the final section concludes the

paper.

2 RESEARCH DOMAIN

The SDH was chosen as a result of an ongoing

collaboration between Sasktel, TR-Labs and the

University of Saskatchewan. Currently, the focus of

this research is on the use of resource rich devices,

122

Shariﬁ G., Vassileva J. and Deters R. (2004).

SEAMLESS COMMUNICATION AND ACCESS TO INFORMATION FOR MOBILE USERS IN A WIRELESS ENVIRONMENT.

In Proceedings of the Sixth International Conference on Enterprise Information Systems, pages 122-129

DOI: 10.5220/0002624901220129

 SciTePress

like the Compaq iPaq and the CDPD network to

support homecare workers in SDH.

Our research focuses on the problems of

nomadic homecare workers in SDH. In SDH each

patient has a team of homecare workers assigned to

him or her, which consists of nurses,

physiotherapists, home health aides, social workers,

and dieticians responsible for looking after the

patient. Each team meets on a periodic basis to

discuss and coordinate necessary treatment for an

individual patient. Tasks and duties are divided

among the homecare workers based on skills and

available time slots. The relation between patients

and homecare workers can be described as M:N.

Though each of the homecare team members

has different tasks and responsibilities, they all have

one task in common: treatment of a patient.

Homecare workers visit their patients in their homes.

The offices of the homecare workers are usually

scattered throughout different buildings and

maintain separate scheduling and information

systems. To date there has been no centralized

information system for their use, which limits their

ability to share and exchange information regarding

schedules and patients’ medical records.

Homecare workers provide health care services

in different district areas in SDH. The wireless

network’s bandwidth they use is low (< 1 megabits

per second (Mbps)) and varies significantly from

one area to another. Currently, Saskatoon mobile

homecare workers do not have access to patients’

information outside the office, i.e. there is no

available remote access to data files, such as health

records and daily schedules.

A preliminary study was conducted on the

usability of the electronic health records system for

homecare at the SDH (Pinelle and Gutwin, 2001).

The study describes the way that groups of

homecare workers care for a patient and identifies

the existing problems. Results from this study

suggest that there are several obstacles for

communication between the members of a group,

which have led to difficulties in making changes in

the treatment of patients, coordination among

homecare workers, and the scheduling of patient

visits. These difficulties are also observed in

accessing information while the groups of homecare

workers are mobile. In order to address these

problems two approaches are proposed: increasing

awareness and supporting communication.

Following the recommendations given (Pinelle

and Gutwin, 2001), we would like to create a system

that supports the access to information and the

communication between homecare workers. Because

of data-security and safety reasons, we decided not

to focus on core health care data but on ways to

improve the availability of non-critical information,

such as scheduling information and personal patient

data, e.g., address and phone number. Access to the

patient medical record data requires, according to

the general regulation, permission from the patients,

which is hard to obtain.

In SDH, homecare workers work in a variety of

contexts, such as in the office, on the road, and at the

patient’s home. When an office scheduler or a

homecare worker makes a schedule, she must have

knowledge of the other team members’ schedules in

order to avoid conflicts in visiting the same patient.

However, since there is no access to a centralized

scheduling system, it is difficult to coordinate

schedule changes and take into account new data

regarding the patient. For instance, if a homecare

worker visits a patient and notices that the patient’s

condition has worsened, ideally, s/he would

immediately enter this information into the patient’s

record so that the other team members of the group

can access the information and perhaps revise their

schedules. However, this is not possible due to

absence of centralized data storage. We classify the

problems encountered by the homecare workers in

two categories. The first category has to do with lack

of efficient communication between homecare

workers, and the second is related to the inefficient

system of accessing and updating information.

Providing homecare personnel with mobile

devices e.g. iPaq and access to the wireless network

based on CDPD will allow them to retrieve

information, receive and send notification of

schedule changes, update the data/information

system and communicate with other workers.

Mobile devices come in varieties of types and

capabilities. For instance, small screen computing

devices differ in their features and abilities

compared to desktop computers. Their most notable

limitations are their smaller screen size, less

powerful processors, less main memory and limited

bandwidth because of the wireless network. To

ensure scalability, performance, and speed, it is

necessary to develop tools and techniques to

compensate for these limitations.

In this paper we introduce tools and techniques

that enable content and presentation adaptation to

improve the access of workers to information/data.

In particular, we have introduced the use of proxies

that create the illusion of a seamless connection.

Moreover, the adaptation is based on user, task, and

context models. Further, to ensure that the users

have the illusion of seamless access, intelligent

agent-based (Jennings et al., 1998) proxies are used,

which can delay write operations, pre-fetch data and

cache data, as well as use the task, context, and user

models to decide when and how to cache, pre-fetch

or delay.

SEAMLESS COMMUNICATION AND ACCESS TO INFORMATION FOR MOBILE USERS IN A WIRELESS

ENVIRONMENT

123

3 GENERAL APPROACH

In order to address these problems, we propose

combining techniques from agent technologies,

distributed database transaction management, and

user and task modeling. One of the goals is to ensure

seamless access to the information despite the

frequent interruptions in the connection caused by

the mobility of the homecare workers and the

unequal coverage of the area. The interface has to

be designed for small wireless devices allowing

homecare workers to conveniently perform typical

operations, including the following:

- Retrieving information from the system

(downloading files)

- Entering new data into the health records

(uploading/updating files)

- Accessing the schedule (reading schedules

and making appointments)

- Communicating with other members of the

home care division (sending/receiving messages)

It has been noted (Vassileva, 1996) that

healthcare and homecare workers have typical tasks,

which have standard information needs, e.g., the task

of visiting a patient requires information regarding

the patient’s address. Therefore, it is possible to (1)

predict what kind of information will be needed by

the homecare worker using her schedule and (2) pre-

fetch and/or adapt information appropriate for the

task and (3) present it according to the user’s

preferences and the limitations of the device on

which the information is going to be viewed. More

specifically, the techniques proposed to achieve the

goals stated in the previous section are listed below:

The use of agents: we use agents to give the

users the impression of seamless connection. Agents

hide the changes in the network bandwidth from the

user by providing the needed information for the

users before their information needs arise by the

means of task, user, and context modeling and

adaptation techniques.

The identification of read/write operations:

To ensure data consistency, we identified two

different kinds of read and write operations in the

SDH domain. The first kind is reading or writing to

data types needing a lock which prevents users from

accessing and/or modifying the data simultaneously

(e.g. a schedule); the second kind is information that

will not be updated simultaneously and therefore

does not require a lock (e.g. a patient’s record).

Providing only necessary functionalities:

When the context (time, location) and the current

task of the user is known, the interface can provide

only the functionalities needed by the user to

perform the task in the specific context. In this way,

the interface can be adapted to the constraint of the

small device.

4 USER AND TASK MODELING

The combination of techniques described in the

previous section is integrated in an application

architecture called “Mobi-Timar” (Farsi for “mobile

caregiver”), which has an interface for homecare

workers, an agent middleware and a server side. The

following assumptions are made to help us

undertake the design:

- There is a centralized schedule for all

homecare workers.

- The mobile devices are resource-rich, e.g.,

have at least 64 megabytes (MB) RAM, a wireless

connection with a bandwidth of at least 19.2 kilobits

per second (kbps), and the ability to run third party

applications.

- Patient and schedule-related data is stored in a

centralized repository.

- The wireless network coverage in the area has

varying but known bandwidth and latency.

The next sections describe the main components

of the Mobi-Timar architecture.

4.1 User Model

In the SDH, the homecare workers can be classified

based on their profession as nurses, physiotherapists,

social workers, home aide workers and dieticians.

Therefore, the user models are based on the

stereotype approach, as there are stereotypes

corresponding to the professions of the homecare

workers. There is a general stereotype for “homecare

workers” which has subclass stereotypes

corresponding to the professions. The stereotype

user model is retrieved and initialized when the user

logs in for the first time (she has to enter her

profession). The stereotype user model does not

change over time i.e., it is a long-term model, since

the profession of a homecare worker remains the

same. The individual user models are instances of

the stereotype user classes containing specific

individual preferences that may change over time,

for example, “personal info”, “Chief” or “expert” in

Figure 2.

Figure 1 shows an example of the stereotype

hierarchy. There are two different relations in this

hierarchy. One relation is the inheritance between a

subclass stereotype and the general stereotype, e.g.,

all nurses are homecare workers (see Figure 1). The

other relationship that is shown in the stereotype

hierarchy is the relation between a type of profession

referred to as “user class” and its task hierarchies,

e.g. the user class “Nurse” has a nurse task

hierarchy.

ICEIS 2004 - SOFTWARE AGENTS AND INTERNET COMPUTING

124

Figure 1: A Stereotype Hierarchy

The individual user model extends the

stereotype of the particular user class with the

personal information, preferences, rank and the

experience of the user. Figure 2 shows an example

of an individual user model and illustrates the

relationship between this model and the stereotype

user model.

Figure 2: An Example of Individual User Model

4.2 Task Model

In SDH each class of users has to perform

standardized tasks. Each task in this domain needs

specific types of information to be successfully

performed, and the information needs of tasks

typically do not vary over time. The task hierarchies

for each profession contain all the tasks that can be

performed by that class of users. Figure 3 shows the

task hierarchy for the user class “Nurse.”

Figure 3: The Task Hierarchy of Nurses Stereotype Class

A detailed view of one task from this task model

is shown in Figure 4. The task model is designed

hierarchically, and each task is decomposed to

subtasks (see Figures 3 and 4). Some of the subtasks

are restricted to certain user classes. For the task

“Visiting patient” in Figure 4, if the user is a nurse,

he needs to perform the following three tasks

sequentially:

- Retrieve the patient’s medical record

- Retrieve the patient’s personal information

- Record the report.

However, if the user is a home-health aide, she

will perform only the latter two subtasks. The reason

for having task hierarchies defined separately from

user task hierarchies is that there can be more than

one class of users performing this task. Depending

on the user class, different subtasks and different

information may be relevant. For example, when a

nurse and physiotherapist perform the same task, i.e.

retrieving a patient’s record, the information that is

provided to them is different. Again, depending on

the users’ profession, access rights to different

information are given in each task. For example, in

Figure 4, if the user is in the nurse class, he has the

right to access the patient’s medical record while

other classes of users (e.g. social aide workers) do

not have access to that information. The lower levels

of the task model contain the detailed

decompositions for each task and the information

needed to perform each subtask. The description of

each subtask includes information, such as the users’

class and context in which the task can be

performed.

As can be seen from Figure 4, each successive

level in the task model differentiates subtasks based

on either user type or context parameters. At the

second level in Figure 4, the task hierarchy is user

profession-specific because each task can be carried

out differently and can provide different information

based on the user’s class. In the second level task

decomposition takes place. At the third level in

Figure 4, the subtasks are organized based on

network bandwidth because different available

bandwidth allows the task to be carried out

differently. Depending on the current bandwidth

SEAMLESS COMMUNICATION AND ACCESS TO INFORMATION FOR MOBILE USERS IN A WIRELESS

ENVIRONMENT

125

Figure 4: Task Hierarchy for Task “Visiting Patient”

availability, only one of the subtasks will be

performed. The changes of task structures and the

required information for tasks are regulated and can

be changed only by the system administrator.

4.3 Schedule

The schedule is used to predict the task that a user

will perform and the context (time, location), so the

needed information can be pre-fetched according to

the task model. Therefore, based on the schedule,

which gives the time and location in which a task

will be performed, the system uses the context

model to predict the network bandwidth at a specific

location.

4.4 Connectivity Model

The system’s connectivity model contains a map of

the network connectivity in the various known

locations shown as an interval of available

bandwidth for each location.

4.5 Library of Cases

Case-based reasoning (CBR) binds user model, task

model and connectivity model together. It is possible

to create a library of cases based on the task and the

type of location of homecare workers. Figure 5

shows the way a case is constructed. In Figure 5, the

User description represents the stereotype user

model, Task represents the task model and

Connection interval represents the connection

model. The cases are indexed by task-ID (t4 in

Figure 5) and

the bandwidth necessary to achieve

the task.

Figure 5: An Example of Case

The process of pre-fetching consists of

predicting network connectivity using the

connectivity model and the user's schedule (time,

location and task) and pre-fetching/hoarding the

needed information

using the user's task model.

Figure 6 shows this

process of pre-fetching

information. The current

location of the user and the

current time is used to find the next appointment

from the user's schedule (step 1). Knowing the

current location and the next location, the bandwidth

is retrieved from the connectivity model (step2).

When the task is known from the schedule, the

necessary information for the completion of the task

is retrieved from the task model (step 3). Having

knowledge of the bandwidth and the task, the user

model is checked for any specific preferences (step

4). These four steps represent the process of

retrieving an appropriate case from the case library.

The fifth step is to pre-fetch/hoard the necessary

information. In Mobi-Timar if a user performs

different tasks which are not included in her/his

schedule, the system will not guarantee the

availability of information in disconnected

condition.

Figure 6: The Process of Pre-Fetching Information

Case1:User description: nurse, physiotherapist

Task: (t4) retrieve the patient’s personal information

Connection interval: [100,375]

Result: load the address and the name of the patient

ICEIS 2004 - SOFTWARE AGENTS AND INTERNET COMPUTING

126

5 ADAPTATION

The proposed Mobi-Timar is based on client-server

architecture in which the agents are the middleware.

The client is a Mobile Device (MD) and the server

runs on a desktop computer. The server side contains

the adaptation mechanism, the library of cases, the

schedule, the user model, the task model, the context

model, the user agents and proxy agents. The server

side is encoded in Java and uses the JADE [4]

platform. A proxy agent and a user agent on the

server side represent each client. The client side

contains a proxy agent, user agent, an application

agent and an individual user model. The client side

is encoded in C#. The agents in C# and Java

communicate through ACL messages. Figure 7

illustrates the communication architecture.

Figure 7: The Mobi-Timar Architecture

5.1 The Use of Agents

To give the users the impression of a seamless

connection, agents are used. Agents hide changes in

the network bandwidth from the user by

automatically pre-fetching/hoarding the needed

information from/to the server ahead of time (as

shown in Figure 6).

5.2 The Identification of Read/Write

Operations

To ensure that the read and write operations are

transparent, we use two sets of agents (user agent

and proxy agent) and two copies of data, one on the

MD and the other on the server (data on the MD is a

copy of the data on server). The user agent on the

server side looks at any changes and updates on the

server side. In the case of disconnection the user

agent on the client and the server side will priorize

the queue of tasks and messages that are waiting to

be executed based on the user’s preferences from the

user model. The proxy agent of a client on the server

side plays the client role during disconnection mode,

while the proxy agent on the MD plays the server

role during disconnection mode. The proxy agent

receives the messages from server/client, waits for

the connection, and then forwards the messages to

the client/server. The user agent performs automatic

hoarding, and the proxy agent performs automatic

reintegration. The needed data for the user is cached

on the MD, and the primary copy of the data is

stored on the server. If the client side is disconnected

and the client is executing a write operation, the

client will continue with the task on MD. When the

client is finished with the task on MD, the agents

will wait for the connection and then execute the

task on the server side and let the user know the

result of the task. The status of the network

bandwidth is also available for the user, so the user

knows the accuracy of the data that she is using.

There are two different kinds of read and write

operations, one that needs a lock and one that does

not need a lock. When a user reads/writes to the

schedule or the agent is pre-fetching/updating the

schedule, the server side will lock the schedule to

avoid any inconsistencies, such as the loss of

information updates. If the client gets disconnected

while reading the schedule, the schedule lock will be

broken after a certain amount of time. If the client

stays connected or only has a brief disconnection,

the schedule stays locked until the client finishes the

transaction. For agents/clients that are writing to the

schedule, the schedule is locked and will not be

updated until the transaction is committed. In case of

an abort, the client on the MD will be notified.

Access to patient records does not require

locking since information cannot be deleted or

modified by homecare workers, who can only add

more data. But adding the new information to the

existing information cannot change the old content

due to SDH policy, which states that all transactions

related to homecare workers and patients should be

recorded and archived. This is illustrated in Figure 8.

After visiting a Patient at time t1, the Physiotherapist

adds information to the Patient’s record using her

mobile device in an area of poor connection. Some

time later at time t2 in an area of good connection, a

Nurse adds information to the Patient’s record. Her

transaction is committed on the server side fairly

quickly at time t3. Therefore, at a later time, t4,

when the Physiotherapist’s transaction is finally

committed on the server side, the additions to the

Patient’s record will be ordered based on the time

when each transaction was originally performed on

the client’s mobile device. For this reason the clocks

of all mobile devices are synchronized with server’s

clock.

SEAMLESS COMMUNICATION AND ACCESS TO INFORMATION FOR MOBILE USERS IN A WIRELESS

ENVIRONMENT

127

Transaction committed

on client side

(On MD)

Transaction committed

on server side

(On MSS)

Time

T1 t2 t3 t4

Nurse

Write to

Patient A

Write to

Patient A

Physio-

therapist

Write to

Patient A

Write to

Patient A

Figure 8: An Example for Second Type of Read/Write

5.3 Providing Only Necessary

Functionalities

The user agent can predict the next type of

information access to be performed by the homecare

worker using the task model (known from the user

class), the connection model, the schedule and a

library of cases. Based on the screen size limitation

of mobile devices, the user agent provides only the

suitable functionalities for the homecare worker’s

current access to information and not the full

functionality of Mobi-Timar since it knows which

operations are going to be performed. This allows a

consistent way to design the user interface around

specific user tasks.

6 EVALUATION

The evaluation of the Mobi-Timar architecture is

done using a simulation environment for the location

and available bandwidth. This simulation depicts a

virtual world that provides the location and the

network bandwidth. The movement of each user

with her MD is represented with a moving avatar in

the virtual world. Each user has a pre-assigned

schedule, which instructs her to move to certain

locations (visit patients) and perform certain tasks.

The user moves her avatar in the simulated location

and performs the tasks requiring access to the patient

data and schedule on her MD.

Figure 9 illustrates one of the network coverage

models for this simulation. Since it is important to

test the ability of Mobi-Timar to adapt quickly to

drastic changes in network coverage, different

models are considered, which allows for a more

realistic discontinuous pattern of coverage.

The simulation shows the movement of the user

through different network areas, which allows

testing the transparent read and write operations and

the availability of data access regardless of the

location and the network bandwidth. By testing

these two aspects, we will be able to evaluate the

functionality of Mobi-Timar.

Figure 9: Different Network Coverage in the Simulation

GUI

The evaluation of Mobi-Timar will be

conducted by running experiments involving a

centralized database containing information for four

patients. Two graduate students will take the roles of

a nurse and a physiotherapist. Each homecare

worker will perform either a full-day or a half-day

schedule, which includes all six types of operations:

downloading files, updating and uploading files,

sending messages, receiving messages, reading

schedules, making or changing appointments. All

these operations will be performed in two different

conditions: connected and disconnected mode. This

will allow us to evaluate the process of pre-fetching

information based on the available bandwidth,

accessing information during disconnection,

reintegrating information and maintaining the

consistency of data and user preferences on both

client and server side.

The evaluation procedure involves the

following two steps:

• Analyzing log data drawn from running the

Mobi-Timar in several different bandwidth

simulation models. The log file contains the

homecare workers and agents transactions

plus all the information transferred between

the clients’ devices and the server with time

stamp allowing the computation of the

speed of task execution.

• Using interviews to perform a usability

study based on the following criteria:

- Ease of use

- User satisfaction

- Interface design and the availability of

information and tasks

The evaluation process will enable us to

determine the number of failures and successful

read/write operations. It will also be possible to

deduce how well the system performs when the

design assumptions made earlier are valid. We will

also be able to ascertain how transparent the

read/write operations were by looking at the

consistency of data on both client and server side

and the numbers of failed and successful operations.

ICEIS 2004 - SOFTWARE AGENTS AND INTERNET COMPUTING

128

7 RELATED WORK

There is a great deal of research on adaptation of the

application and web infrastructure for mobile

devices. Most of this research, for example

CRUMPET (Poslad et al., 2001a; Poslad et al.,

2001b), MyCampus (Sadeh et al., 2002) and Hippie

(Oppermann et al., 1999; Specht and Oppermann,

1999), focuses on content adaptation, information

retrieval, interface adaptation and representation

adaptation. Furthermore, the focus of most of this

research is on adaptation based on the user type and

the level of experience of the user. In contrast,

Mobi-Timar adapts also based on context (e.g.

current time, user task, and connection availability).

Projects, such as Broad-Car (Console et al., 2002),

propose adaptation based on the context of the user

and use multimedia output, e.g. voice, video or text,

for presenting the data to the user. In these studies,

parameters used for adaptation are bandwidth, user

type, task, screen size, output channels of the device,

and the location. However, there are no write

transactions performed in any of these systems, and

synchronization and transaction management are not

needed. Unlike previous approaches, as in

CRUMPET, Hippie and MyCampus, our Mobi-

Timar addresses the problem of disconnection.

Agents are used to pre-fetch information relevant to

the user task and context. This research doesn’t

focus on just-in-time retrieval; instead, it aims to

support seamless access to information and

modification in a timely fashion.

8 CONCLUSIONS

In this paper we described Mobi-Timar, an

architecture for a mobile communication and

information system for homecare workers, which

uses a variety of techniques including user

modeling, task modeling, case-based adaptation,

multi-agent systems and mobile transaction

management.

The result of this research is an application and a

middleware for a mobile environment that can be

used in other domains with similar characteristics,

for example, supporting mobile technicians,

salespeople, real estate agents, etc. The application

ensures seamless access to data by using mobile

transaction management performed by the user agent

and the proxy agent. The user model, task model and

context model allow the agents to select and pre-

fetch relevant information needed for the user’s next

task and to create the illusion of seamless connection

for the users even when the bandwidth is low or the

user is disconnected.

ACKNOWLEDGEMENT

This research has been supported by TR-Labs.

REFERENCES

CDPD (Cellular Digital Packet Data),

http://www2.picante.com:81/~gtaylor/cdpd.html

Compaq iPaq,

http://athome.compaq.com/showroom/static/ipaq/hand

held_jumppage.asp

Console, L., Gioria, S., Lombardi, I., Surano, V., Torre, I.,

2002. Adaptation and Personalization on Board Cars:

A Framework and Its Application to Tourist Services.

In 2nd International Conference on Adaptive

Hypermedia and Adaptive Web Based Systems, AH

2002, 112-121.

JADE: http://sharon.cselt.it/projects/jade/

Jennings, N., Sycara, K. and Wooldridge, M., 1998. A

RoadMap of Agent Research and Development.

Journal on Autonomous Agents and Multi-Agent

Systems, 1, 275-306.

Oppermann, R., Specht, M., Jaceniak, I., 1999. Hippie: A

Nomadic Information System. Gellersen. W. (Ed.):

Proceedings of the First International Symposium

Handheld and Ubiquitous Computing (HUC'99),

Karlsruhe, September 27 - 29, 330 - 333.

Pinelle, D. and Gutwin, C., 2001. Collaboration

Requirements for Home Care. Research Report,

University of Saskatchewan,

http://hci.usask.ca/publications/index.xml

Poslad S., Laamanen H., Malaka R., Nick A., Buckle P.

and Zipf, A., 2001a. CRUMPET: Creation of User-

friendly Mobile Services Personalised for Tourism.

Proceedings of: 3G 2001 - Second International

Conference on 3G Mobile Communication

Technologies. London, UK, 26-29.

Poslad S., Charlton P. In Marik V., Stepankova O., 2001b.

Standardizing agent interoperability: the FIPA

approach. Proceedings of the Multi-Agent Systems &

Applications Conference, ACAI-01 (invited paper).

Sadeh, N., Chan, E., Shimazaki, Y., Van, 2002.

MyCampus: An Agent-Based Environment for Context-

Aware Mobile Services. In AAMAS02 Workshop on

Ubiquitous Agents on Embedded, Wearable, and

Mobile Devices, Bologna, Italy.

Specht, M., and Oppermann, R., 1999. User Modeling and

Adaptivity in Nomadic Information Systems.

Proceedings of the 7. GI-Workshop Adaptivität und

Benutzermodellierung in Interaktiven

Softwaresystemen (ABIS99), Universität Magdeburg,

September 27 - October 1, 325 -328.

Vassileva, J., 1996. A Task-Centered Approach for User

Modeling in a Hypermedia Office Documentation

System. User Modeling and User Adapted Interaction,

6, 2-3, 185-223.

Weiser, M., 1991. The computer for the twenty-first

century. Scientific American, September, 94-100.

SEAMLESS COMMUNICATION AND ACCESS TO INFORMATION FOR MOBILE USERS IN A WIRELESS

ENVIRONMENT

129