AN AUTOMATIC MEETING SCHEDULING FOR

MOBILE USERS

Behrad Assadian, Simon Case, and Fang Wang

British Telecommunications, Intelligent Systems Lab,

Post Point 12, Orion Building, Adastral Park,

Ipswich, England

stract. Emerging mobile computing technologies have increasingly

enhanced the ability of people and organisations to work in collaboration at

different locations and even on the move. By using various handheld devices

such as PDAs, mobile phones, and electronic pens, remote and mobile people

can flexibly exchange text and multimedia information with each other, and

virtually work together as a cohesive group independent of time and location

(constraints). Effective mobile collaboration requires reliable communication

methods to connect people. More importantly, in-time, on-demand meeting

scheduling design is necessary so as to bring fixed or mobile people together to

discuss collective activities, such as planning of group work and allocation of

resources whenever requested. As people usually have different availability and

time arrangement preferences, quickly identifying preferable scheduling

solutions for a group of people with heterogeneous and sometimes conflicting

requirements is a challenge.

1 Introduction

Recent commercial and academic work on automatic meeting scheduling suggests a

promising solution to handle the above challenge, by automating the process of

communication and negotiation between multiple calendars. The existing automatic

meeting scheduling technologies provide automation at diverse levels, ranging from

simple free time display to automatic delivery of negotiated results. These

technologies, however, have several limitations. First the solutions provided by these

technologies are usually unsatisfactory since they lack proper capture and handling of

user personal information such as preferences and requirements during scheduling.

Second these technologies deploy rigid negotiation thus they cannot quickly reach

solutions adaptive to varying scheduling conditions. The computational cost becomes

unfeasibly large with the increased complexity and scale of scheduling. Privacy is an

important issue in group coordination. Some commercial scheduling toolkits allow

mutual access of user availability information. This actually prevents their use in

practice, since many people are reluctant to disclose their personal time arrangement

to others. An ideal meeting scheduler should be able to find the best meeting solutions

fully automatically and flexibly according to different user preferences and

requirements. Moreover, the architecture of the scheduler should be designed to

Assadian B., Case S. and Wang F. (2004).

AN AUTOMATIC MEETING SCHEDULING FOR MOBILE USERS.

In Proceedings of the 3rd International Workshop on Wireless Information Systems, pages 119-126

DOI: 10.5220/0002657701190126

 SciTePress

protect user privacy and with the consideration of the limitations of current mobile

devices so as to work with the low computational power, small memory and low

bandwidth problems encountered by the mobile users.

This paper presents a novel approach, iMeeting, to achieve automatic meeting

scheduling for mobile users. This approach implements intelligent agents and fuzzy

logic functions to achieve the efficient personalised and flexible scheduling. iMeeting

quickly and automatically delivers the best available scheduling solutions for a

proposed meeting, which most satisfy user preferences and requirements. Meanwhile,

user privacy is well protected in iMeeting due to the completely automatic scheduling

on a client-server architecture. iMeeting has a specially designed web portal for the

easy access and use on small handheld devices which have limited functionality.

2 Automatic Meeting Scheduling

Meeting arrangement or scheduling is an ordinary task faced by almost everyone

everyday. Though simple as it looks, meeting scheduling in practice is a time-

consuming and tedious process usually involving multiple iterations of time

proposing, declination, and negotiation. This process becomes even more complex

and difficult, when the number of people invited for a meeting is large and when the

people are distributed at different locations (or time zones) and have different

preferences or bias of time management. To solve this problem, automatic meeting

scheduling that automatically arranges meetings for distributed people based on

various personal requirements becomes a desirable solution.

Today there are a number of commercially available software packages for scheduling

meetings. Well known examples include Microsoft’s Outlook, Lotus’ Notes, and

Meeting Maker. Some e-conferencing software such as Polycom’s Conference Suite

also includes scheduling mechanisms with network awareness. These schedulers

usually provide an interface to view the free and busy time of the attendees of a

proposed meeting. They can also search the next available time that every meeting

attendee is free. The subsequent proposal of available time(s) to other meeting

attendees for agreement and confirmation has to be manually executed by the meeting

proposer himself, usually by sending emails. Because commercial schedulers have no

consideration of user preferences or personal requirements in time search, the

proposed time results are often rejected by other attendees and the proposer has to

find other available times to propose. The above software therefore only provides a

useful tool to assist availability comparing and searching, but they are not automatic

schedulers per se. Frequent communication between users can make scheduling

unnecessarily prolonged and frustrating. This problem would be exaggerated in

mobile collaboration when quick responses cannot be obtained from mobile workers.

Another important issue in current commercial scheduling software is that they have

no protection of user privacy. They permit the sharing of personal calendar

information by providing free or busy time for others to review. This is unacceptable

for most people, especially if the calendars contain private information reflecting

personal time arrangement and preferences.

120

In recent years there has also been intensive investigation on using agent technologies

to achieve scheduling automation for distributed users [1], [2]. Endowed with certain

knowledge and information about their users and relevant jobs, autonomous or semi-

autonomous agents are employed to process calendars and coordinate with each other

a mutually acceptable time for a meeting on behalf of their users. For coordination

between distributed users, protocols ranging from simple but effective contract net to

recent market mechanisms are extensively employed. In the contract net protocol [3],

an agent for a user (usually the agent for the meeting host) proposes one or more time

proposals according to its user’s availability. The other user agents then bid for the

proposals. User preferences or biases were considered in [4] when proposing and

accepting time slots. User preferences, however, were values defined and input by

users themselves. This obviously increases the workload of users, and more seriously,

decreases the usability of the scheduling software since personal preferences or

requirements are always difficult to define in exact numbers. Because the

coordination and negotiation between user agents always follow a fixed routine (e.g.,

one agent proposes time slots and others bid, a decision on a proposed time slot is

based on bids from all agents, etc.), the scheduling procedure cannot make timely

responses for varied calendar conditions and user requirements. This inevitably causes

ineffective scheduling which has unnecessary information collection and

computation.

2.1 An Effective Solution for Meeting Scheduling

The approach proposed here for automatic meeting scheduling, called iMeeting,

adopts intelligent agents and fuzzy logic to seek the best available solutions based on

user personal information through flexible and efficient coordination. User personal

information such as preferences for a time slot and the importance of a meeting to a

user can be either input by user through natural language, or automatically learned by

agents through the users’ previous time-management behaviour or other exterior

information proactively collected by the agents. Meanwhile, a kind of lightweight

adaptive agent is designed to implement flexible and efficient coordination between

user agents. Based on an easy-to-use web portal, iMeeting provides automatic

meeting scheduling for users at any time and anywhere from any device. Figure 1

illustrates an overview of the iMeeting software. iMeeting develops a client-server

architecture to cope with the limitations of current handheld devices whose

computational power and operating systems (e.g., PalmOS) are generally incapable of

running large data processing and handling tasks such as incoming request in the

background. The iMeeting server stores user calendar information and undertakes

necessary computation tasks for calendar management and meeting scheduling,

whereas client machines provide an intuitive interface for users to request and receive

all kinds of calendaring and scheduling services. All scheduling work is dealt within

the iMeeting server through internal interactions between agents. Even inside the

system, iMeeting agents only request information about other users

from their user

agents when it is really necessary. Users are only informed of automatically searched

solutions (via their user agents) when scheduling is done. iMeeting avoids any

unnecessary disclosure of users’ personal information thus protecting user privacy to

a great extent.

121

Fig. 1. Overview of the iMeeting software. Communication agents(CA)

2.2 Personalised User Assistance

iMeeting provides two kinds of intelligent agents to achieve flexible automatic

meeting scheduling. At the core of iMeeting is a set of personalised agents for users.

Under users’ control, user agents act on behalf of users to process their calendars and

relevant scheduling tasks based on user personal information (e.g., user preferences

and requirements). iMeeting makes use of fuzzy logic to compute with words and

deal with uncertainty. User personal information is represented and processed as

fuzzy words [5]. In consequence users of iMeeting can easily describe their

preferences or requirements by using natural language. For example, a preferred time

for a meeting can be simply specified in natural language expressions such as

morning, afternoon, early morning, around 1:00, etc. The information is then

interpreted by iMeeting using predefined fuzzy logic functions. User personal

information can also be learned by user agents from user actions during time

arrangement or from exterior information such as calendar, user interest profile or

organisation chart. When scheduling a meeting, user agents search and combine

appropriate personal and exterior information to propose suitable time slots for a

meeting. Moreover, the agents employ the information to estimate factors relevant to

a proposed time slot. In particular, the agents can decide on the following features on

behalf of their users: importance of the meeting to the user, importance of the user to

iMeeting server

User Agent User Agent User Agent

122

the meeting, and overall busyness of each user. Importance of the meeting to the user

is a measure of how useful the meeting is to the end user. Importance of the user to

the meeting on the other hand is a measure of how essential a given user is to the

success of the meeting. A senior manager for example may well be very important for

the organisers of a given meeting (i.e. they have a high Importance of user to meeting

measure) whilst not attaching much importance to the meeting themselves (i.e. they

give a low Importance of meeting to user value to the meeting.) Overall busyness is a

measure of generally how busy the given user is over a time period. If a given user

has a high overall busyness rating then they would probably prefer meetings to be

scheduled at a different time if possible. The agents learn these values from responses

to previous proposals combined with external information such as the position of the

attendees within the organisation. Based on the information of calendar, preferences

and various importances, a value between [0, 1] is automatically generated by the

fuzzy system to imitate the response of a user to a proposed time slot. This value

contributes to the overall fitness of a proposed time slot, which is evaluated in the

agent coordination.

2.2.1 Evaluating User preferences

The evaluation agent uses busyness of user at a given slot (a crisp time and duration),

importance of user to meeting, availability of user, preference of user for the proposed

meeting slot, and the importance of the meeting to the user. The technology requires

that a meeting proposal has the following information: time, duration, attendees. It

may also include some or all of location, agenda, subject and importance of the user

to the meeting. These inputs are combined in the evaluation agent in a fuzzy classifier

to produce an output of the form [0,1] which is passed to the scheduler. Information

exterior to calendar information such as known user interests and organisational

information can be used to generate the inputs to the fuzzy classifier from the

proposal. A fuzzy sub-system in the evaluation agent is used to generate the value of

importance of the meeting to the user by combining and existing information on: the

attendee list, orgenisational chart, acquaintance list the user interest profile and the

agenda and/or meeting subject. A fuzzy sub-system in the evaluation agent combines

information from the users calendar with the location of the meeting to decide

whether a user can attend a meeting or not. A fuzzy sub-system in the evaluation

agent looks at the users calendar about a given slot to allocate a busyness value to the

slot. The evaluation agent adapts the fuzzy rule system when a user reschedules a

meeting made on his/her behalf. The proposer agent accepts a meeting request from

the user (or other agent working on behalf of the user) and generates a meeting

request. The proposer agent maintains a profile of other attendees responses to

meeting request. This profile is used to generate an overall busyness for the user. A

fuzzy system within the proposer agent combines the overall busyness with the

position of the user in the organisation and acquaintance profiles to assign an

importance of user to meeting value to each attendee in the meeting.

123

2.3 Flexible Adaptive Coordination

When arranging a meeting, the meeting proposer sends the iMeeting server a meeting

request including information on the duration, invited attendees, and various other

meeting requirements if known.. Examples of meeting requirements are the minimum

percentage of attendees that should attend this meeting, and the essential attendees

who must attend the meeting. The meeting request is accepted by the user agent of the

proposer and then forwarded to the user agents of the meeting invitees. In contrast to

other academic or commercial schedulers, iMeeting treats each attendee equally so

every user agent involved in a meeting has the ability to propose its own time slots (a

crisp time and duration) which are preferred by its user. To manage the proposed time

slots, a set of lightweight adaptive agents are created to coordinate with each other

and with user agents to find the most successful proposed time slot. These agents are

called coordination agents. Due to the use of coordination agents, user agents are

relieved from the intensive coordination work so that they can be used to provide

other useful personal services for their users, e.g., calendar management or other

meeting scheduling.

Coordination agents are adaptive agents that take flexible actions according to the

current environments and conditions. After creation, they immediately seek other

coordination agents which manage the same time slot, and combine with them into

one agent in order to increase the satisfactoriness of the proposed time slot.

Coordination agents make instant decisions on the feasibility of the proposed time

slots based on the meeting criteria and responses currently collected from some of the

user agents. Thus any inappropriate time slots can be immediately identified, and

accordingly the corresponding coordination agents will be terminated. This avoids

unnecessary scheduling on unsuitable solutions. The resulting scheduling can

therefore be achieved quickly and efficiently with low computational cost. This

property is especially beneficial when there are a large number of people involved in

a meeting or the people have disparate schedules, since the increased heterogeneity

and inconsistency between these people would help to remove unsuitable solutions at

an early stage. Experimental simulations have proved that the computational cost of

iMeeting could be at least one third less than the contract net or heuristic search, when

the number of people is increased to more than 50 (assuming the calendars of these

people are independent).

Before a coordination agent is terminated, the agent will inform its agreed user agents

of the failure. The user agents can then consider initiating new time intervals by

relaxing the constraints of their users. Due to the parallel scheduling of multiple

coordination agents, iMeeting can not only promptly deliver the best available

solution that most satisfies meeting requirements and user preferences, but also

suggest several (2-3) alternative high quality solutions for users to choose. In

particular, when a solution satisfying all user requirements is absent, iMeeting will

provide several results that are likely to be accepted by the users. The users

(especially the meeting host) then have the option of accepting one of the results.

124

2.4 Easy to use and access

iMeeting is easy to access and use. It provides a general web portal that can be

accessed by a wide-range of devices ranging from lightweight devices (e.g., smart

'phones, PDAs and laptops) to desktops, workstations and enterprise servers. In

particular, a simplified and succinct web interface is specifically designed to

accelerate and faciliate the use of iMeeting on mobile devices. Recently a team of

researchers at the University of Maryland and Microsoft created a “calendar-on-

steroids” for keeping track of schedules on Pocket PCs [6]. Though this calendar

provides a functional interface to plan and analyse schedules, a question is whether

mobile users would accept such complex tasks on handheld devices which have a

limited functionality particularly with respect to the input side of the interface. In

iMeeting, we keep the interface compact and easy to understand. The relevant

management and computation functions are automatically implemented by iMeeting

with least human involvement.

iMeeting offers flexible scheduling that can work with varying constraints for varied

user requirements. The input for a meeting proposal can simply be the basic meeting

information, i.e., time, duration, and attendees. It can also include some or all of the

other meeting concerns such as location, agenda, and subject. iMeeting always seeks

the best available solution to meet all kinds of user requirements when scheduling a

meeting.

3 Conclusion

The demand for effective mobile collaboration from an increasingly mobile work

force has motivated extensive commercial and academic work on distributed

automatic meeting scheduling. The iMeeting software introduced in this paper

provides a new and efficient way to automatically schedule meetings for mobile users.

A Survey carried out confirmed the usefulness of iMeeting as an automatic meeting

scheduler. This survey investigated the attitudes and opinions of the people toward

automatic meeting scheduling using iMeeting as ordinary users. Around 70% of the

people deemed automatic meeting scheduling as an important and useful tool, and

more than 80% people commended the easy input of the natural language and the

personal consideration in scheduling as advantages of iMeeting. iMeeting is

highlighted by its privacy protection which is a serious concern of many users but

neglected in most commercial products. The survey also revealed worries from some

involved people, who suspected automatic schedulers won’t deliver solutions which

are the most suitable for their own preferences.

Like many other tools, the value of iMeeting is determined by how and how often

people use it. Regular update of personal calendars and frequent use of iMeeting for

meeting arrangement will increase the opportunities for iMeeting to capture and learn

the exact information from user behaviour.

125

References

[1] L. Garrido-Luna and K. Sycara, “Towards a Totally Distributed Meeting Scheduling

System”, Lecture Notes in Computer Science, vol. 1137, 1996, pp.85-97.

[2] S. Sen, “An Automated Distributed Meeting Scheduler”, IEEE Expert, vol. 12, no. 4, 1997,

pp.41-45.

[3] R.G. Smith, “The contract net protocol: High-level communication and control in a

distributed problem solver”, IEEE Transactions on Computers, vol, C-29, no. 12, 1980,

pp.1104-1113.

[4] S. Sen, T. Haynes and N. Arora, “Satisfying User Preferences While Negotiating

Meetings”, International Journal of Human-Computer Studies, vol. 47, 1997, pp.407-427.

[5] W.R. Wobcke and A.G. Sichanie, “Diary Management with Fuzzy Preferences”,

Proceedings of the Fifth International Conference on the Practical Application of Intelligent

Agents and Multi-Agent Technology, 2000, pp.105-120.

[6] B.B. Bederson, M.P. Czerwinski and G.G. Robertson, “A Fisheye Calendar Interface for

PDAs: Providing Overviews for Small Displays”, HCIL Tech Report #HCIL-2002-09,

Computer Science Department, University of Maryland, 2002.

126