AA110 CF Levchenkov Software Agent Negotiation by dandanhuanghuang


									                           MCDM 2004, Whistler, B. C., Canada, August 6-11, 2004


                                          Anatoly Levchenkov
                                     Transport and Machinery Faculty
                                          Riga Technical University
                                      1, Kalku St., Riga, LV-1658, Latvia
                                   mailto:levas@latnet.lv fax: +371 7324362

Keywords: software agent, schedule, cooperative games, negotiation, group decision making and logistics


The purpose is to work out an algorithm of negotiation for software agents for the group decision support
system for scheduling, intended for solution of logistics tasks of minimization of the usage of resources and
coordination of actions of the logistics process participants.

In logistic s process of deliveries there can be two kinds of objects: the processors are immovable objects -
warehouses, shops etc., and the jobs - objects, which move between processors and are served by them - for
example, trucks. Each object in a logistics chain has own software agent. Also there is a super agent, which
makes the schedule by using group decision support system.

For solution of the task a set of processors and a set of jobs are given. Each job consists of a number of
operations and each operation is carried out by one processor. Also duration of each operation and directive
terms are given. General algorithm is proposed to solve the task of making a schedule to co-operate the work
of software agents.

The scheduling task is represented as a cooperative, non-coalition game, where all players are interested in
receiving maximal profit with minimal costs, not disturbing each other. All the data for solution of the task
have to be incorporated in a database, and appropriate programming languages and serv ers will be used for
realization of algorithms using group decision support system

In scheduling game the players are the software agents, where each is responsible only for its own object.
The software agent has the following information about the object: the directive term - time of a desirable
beginning and ending of work, the penalties in case of infringements, serving time and time spent on route.

Each player-job has playing strategies - orders of visiting of necessary processors. The best strategy is that,
where the transport charges are minimal. The software agent defines the best strategy and offers the
schedule, convenient for the object. This schedule is the base of negotiation for the group decision support
system. The super agent realizes negotiation of these software agents, defining a negotiation set. It plays a
role of the arbitration judge, which nominates what players have to do proceeding from results of negotiation.

The offered algorithm gives possibility to realize negotiation between software agents and super agent for
making schedule. Charges for schedule without negotiation in some cases are twice larger than when the
proposed algorithm is used.

To top