Call for Participation CAIA-94 Workshop on Coordinated Design and Planning Marriott Riverwalk - San Antonio, Texas March 1, 1994
The goal of this workshop is to find common ground among various approaches to the coordination of constructive problem solving--particularly design and planning--among multiple agents. Such approaches include, but are not limited to, DAI, case-based reasoning, constructive problem solving, and multiagent systems. The intent is to discover where these approaches complement and compete. We also wish to discover what parts of the problem space may not yet be covered. Because the terminology among the approaches has not yet been unified, and because many of the reported problems have different characteristics, this workshop uses the tactic of grounding the discussion on a common application example.
A meeting scheduling example has been chosen, because it is simple and easily understood without technical jargon. It is also of a ubiquitous problem type: the assignment of resources to tasks within discrete timeframes under constraints. Examples are assigning gates to airplanes, rooms to hotel guests, teachers to courses, and employees to jobs.
The meeting scheduling problem also has features common to many cooperative probelems. For instance, each meeting attendee may have goals that conflict with that of the others. While they have the common goal of finding days on which to meet, attendees may have independent reasons to prefer different days. Also, it is assumed that each attendee may have some local calendar or meeting scheduler. Some combination of the attendees and these systems communicate over a wide-area network (WAN), such as the Internet, to solve the meeting scheduling problem. Any number of intermediate software systems, interacting via the WAN, can be posited to facilitate communications and problem solving.
Workshop participants are expected to relate their approach to the example by answering in their paper at least the six questions that follow the example. It is not required that the approaches or systems described actually be in use, since it is well-known that scheduling systems are rarely successful in implementation. (Speculation about this fact is invited.) The example was chosen to simply explicate some common features of distributed design and planning problems.
The workshop will consist of at most thirty-two (32) participants, each of whom should submit a paper for the working notes. No more than eight (8) papers will be selected for presentation and discussion. Authors will not present their own papers. Instead, the papers will be assigned to eight groups of three nonauthors for presentation and evaluation. Authors will make a short response.
A comparison and evaluation of the eight papers will be the subject of a panel discussion. The panel will be also composed of nonauthors, and will attempt to identify where the eight approaches presented compete with or complement each other. ________________________________________________________________________
Problem Description: The Secretaries' Nightmare.
Axel, Brigitt, Carl, and Dirk wish to plan two all-day meetings. The scenario is as follows:
The meetings must be scheduled on weekdays during April 1994. The first meeting concerns preliminary business agreements among three companies. Axel, Brigitt, and Dirk can each represent their respective three companies' business interests and marketing goals. Carl's company will be a subcontractor to Dirk's company and he need not attend this first meeting.
The second meeting is a technical discussion of proposed architectures for the joint project. Carl's attendance is necessary, although Dirk's is not. Axel and Brigitt must attend.
Axel is based in New York City, Brigitt in Palo Alto, Carl in Los Angeles, and Dirk in Austin. Travel should be minimized by scheduling two days in a row if possible. The faster the meeting can take place, the better. The meetings can be in either Austin or Los Angeles, but it is preferable for the second meeting to be in Los Angeles because of the availability of a demonstration. If the meetings are back-to-back in the same city, Los Angeles is preferable.
Axel is available in April the week of the 4th, the 18th and 19th, and the 25th and 26th. Brigitt is available the 7th, 8th, 19th, and the week of the 25th. Carl is available on the 7th, 19th and 26th. Dirk is available on the 7th, 8th, 18th, and 25th.
The best option would be to hold consecutive meetings on the 7th and 8th, but Carl cannot meet on the 8th. The meeting cannot occur on the 18th and 19th because Brigitt is not available on the 18th. One available option is for Axel, Brigitt, and Dirk to meet on the 7th. Then Axel, Brigitt, and Carl can meet on the 19th. Another is for Axel, Brigitt, and Dirk to meet on the 25th and for Axel, Brigitt, and Carl to meet on the 26th. The first option is chosen because it is important to meet as soon as possible and because Dirk and Carl would like to meet between the two meetings if possible.
Under this first option, the two meetings can be in different cities. The first meeting will be held in Austin, because it is more central, especially for Axel. The second meeting should be held in Los Angeles because of the demonstration facilities. Under this option, the need for another meeting is generated. Since Carl was not present at the first meeting, Dirk wants to brief Carl to make sure the business decisions are reflected in the technical design.
But it turns out that Dirk and Carl cannot get together between the 8th and the 19th. They object to the current plan. But now they also object more strongly to the meetings being held on the 25th and 26th. They push for another option. The current options will be held as backup plans; possibly Dirk and Carl can get by on a phone call. But all participants agree to negotiate to see if anyone will relax their constraints. Each must persuade the others of the importance of their current commitments, but avoid revealing possibly confidential information.
Axel declines to make any more days available because he has already made all of the week of the 4th open. Carl and Dirk offer to be available on the 5th and 6th, but Brigitt can make neither meeting because of nonrefundable travel plans. Finally Brigitt offers to cancel another meeting and be available on the 18th. This means that the two meetings can be held back-to-back on the 18th and 19th. Now both meetings will be held in Los Angeles.
Then Carl's schedule changes. It turns out he can meet in Austin on the 8th. This means that the separate meeting option is now available - Axel, Brigitt, and Dirk can meet in Austin on the 7th; Dirk and Carl can meet in Austin on the 8th; and Axel, Brigitt, and Carl can meet in Los Angeles on the 19th. This also means that everyone could meet on the 7th and 8th, an even better option. But Axel objects saying that he has already planned another meeting in Los Angeles on the 19th to take advantage of his presence there. He would rather not cancel that meeting.
They agree to the separate meeting option so that Brigitt can reschedule the meeting she canceled. Everything works as planned. _____________________________________________________________________
Questions:
1. What parts of the problem does your approach help with and how?
2. Which functions are distributed and which are centralized?
3. What other tools and methodologies do you presume?
4. Do you allow heterogeneous agents, such as different calendar systems, to participate? What changes are required to the systems?
5. What messages, or kinds of messages, are exchanged between your agents?
6. Why is your approach better than email between people? Between systems? _____________________________________________________________________
Workshop Committee
Charles Petrie <petrie@sunrise.stanford.edu> (chair) Center for Design Research Stanford University Bldg. 560, Panama Street Stanford, CA 94305-2232
Michael Huhns <huhns@mcc.com> J. Marty Tenenbaum <marty@eitech.com> MCC EITech 3500 West Balcones Center Dr. 459 Hamilton Avenue Austin, Tx 78759-6509 Palo Alto, CA 94301
Please send papers to Charles Petrie to arrive no later than January 14th, 1994. Notifications will be sent by February 15th. Papers should be no more than ten (10) pages in length. Clarity and specific relevance to the scenario described will be of the utmost importance in paper selection.
Electronic submission in PostScript format is encouraged. However, Apple users must beware that their files may be too large to email and even if received may not print. LaTeX submissions are also welcome. But please check with the Chair about style files and included PostScript figures.
If you plan to participate in this workshop without registering for the main conference, there will be a $75 fee.
This workshop will be repeated at CoopIS-94 in May in Toronto. A book including the best papers from both workshops is planned.