Abdul Samad Shibghatullah
43
30052008
negotiation, the participants are the active components that determine the direction of the search. The minimum requirement of a negotiating agent is the ability to make and
respond to proposals. To improve the efficiency of the negotiation process, the recipient needs to be able to provide more useful feedback on the proposals it receives. Various
negotiation methods have been defined in literature, and most of them are inspired by human negotiations that are market-based, plan-based, game theory-based, and artificial
intelligence based. Kraus, 1997; Faratin et al., 1998; Jennings et al., 2001. Market-based negotiation is the simplest and the most renowned negotiation protocol,
and the most widely used in agent-based systems is the contract net protocol involving offers, bids, and contracts Nwana et al., 1996; Beer et al., 1999; Shen and Norrie,
1999; Jennings et al., 2001; Shen et al., 2001. It is a high-level negotiation protocol
that provides many advantages and most important is its flexibility and dynamic nature, which suits industrial agent-based applications.
Plan-based negotiation is based on cooperation strategies for resolving conflicts among plans of a group of agents. Laasri and Lesser 1990 described a three-phase cycle
negotiation plan. This model of negotiation could be centralised or distributed. Game theory-based negotiation employs techniques based on game theory to structure
and organise negotiation between the agents Nwana et al., 1996; Jennings et al., 2001. The key concepts in the game theory approaches are utility functions, a space of deals,
strategies, and negotiation protocols. Artificial Intelligence based negotiation considers negotiation as an iterative activity
and Sycara 1991 exploited case-based reasoning in this iterative process. Sycara 1991 argued that human negotiators draw from past negotiation experiences to guide
present and future decisions.
2.5.4 MAS Communication
In MAS, communication is the basis for interactions and social organisations. Communication enables the agents to cooperate, coordinate their actions, and carry out
tasks jointly resulting in systems that are more coherent. A number of communication
Abdul Samad Shibghatullah
44
30052008
languages have been developed for inter-agent communication, and the most widely used ones are KIF Knowledge Interchange Format Genesereth and Fikes, 1992,
KQML Knowledge Query and Manipulation Language Finin et al., 1994, and ACL
Agent Communication LanguageLabrou et al., 1999. Currently, XML Extensible
Markup Language has started to show its performance as a language to encode messages exchanged between agents, in particular in agent-based e-commerce to
support the next generation of Internet commerce Glushko et al., 1999; Korzyk, 2000; Turowski, 2002.
Several modes of communication have been defined that are shared data and message passing Ferber, 1999; Shen et al., 2001; Weiss, 1999; Wooldridge, 2002. Repository
is a common shared data repository, i.e. a blackboard. A blackboard is used by agents to write messages, post partial results, and obtain information. This mode of
communication is used in blackboard architectures. Message-passing communication is a widely used approach. In the message-passing
approach, agents communicate with each other by sending asynchronous messages. Asynchronous communication is the primary mode of interaction in most agent-based
applications. There are two basic message types: assertions and queries Weiss, 1999. Every agent, whether active or passive, must have the ability to accept information. In
its simplest form, this information is communicated to the agent by means of an assertion. In order to assume a passive role in a dialogue, an agent must additionally be
able to answer questions, i.e. it must be able to accept a query from another agent and send a reply to the agent by making an assertion. In order to assume an active role in a
dialogue, an agent must be able to issue queries and make assertions. With these capabilities, the agent then can potentially control another agent by causing it to respond
to the query or to accept the information asserted. There are several methods of communication in message-passing mode. There are
point-to-point, broadcast, and multi-cast Weiss, 1999; Wooldridge, 2002. In point to point communication an agent sends a message to another specific agent. In broadcast,
an agent sends out a message to all other agents in the system, and in multi-cast, an agent sends out a message to a selected group of agents.
Abdul Samad Shibghatullah
45
30052008
Formalisms for representing communication in agent theory tend to be based on speech act theory Wooldridge and Jennings, 1995; Ferber, 1999; Weiss; 1999; Wooldridge,
2002, as originated by Austin in 1962, and further developed by Searle in 1969 Wooldridge and Jennings, 1995. The key principle of speech act theory is that
communicative utterances are actions, in the same sense that physical actions are. They noticed that a certain class of natural language utterances or speech acts had the
characteristics of actions, in the sense that they change the state of the world in a way analogous to physical actions. They observed that most things people say are not simply
propositions that are true or false, but are performatives that succeed or fail. Since the early 1990s, speech act theories have directly informed and influenced a number of
languages that have been developed for agent communication, such as KQML and ACL. In KQML and ACL, each message has a performative a class of the message
and a number of parameters to describe the format of the message sender, receiver, content, etc.. The most important differences between these two languages are in the
collection of performatives they provide.
2.5.5 Application of MAS in Scheduling
