Clearing outdated information in stabilizing systems of cooperative communicating information agents | |
| Author | Do Duc Hanh |
| Call Number | AIT Diss. no.CS-10-03 |
| Subject(s) | Intelligent agents (Computer software) Information retrieval Information storage and retrieval systems |
| Note | A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Engineering in Computer Science, School of Engineering and Technology |
| Publisher | Asian Institute of Technology |
| Series Statement | Dissertation ; no. CS-10-03 |
| Abstract | We consider systems of cooperative communicating information agents situated in unpredictable environments. Agents needs to cooperate to provide information to users as they individually may not have the knowledge or the capability to sense all relevant changes in their environment. Due to the unpredictability of their surroundings, agents may possess temporarily wrong information, propagate outdated information to other agents. Stabilization refers to a capability of the agents to eventually get correct information despite unpredictable environment changes. In other words, agents in “stabilizing” systems have capability to clear outdated information from circulation. It could be viewed as a correctness criterion of systems of cooperative communicating information agents. We argue that the stabilization of a system of cooperative information agents could be understood as the con-vergence of the behavior of the whole system toward the behavior of a “superagent”, who has the sensing and computing capabilities of all agents combined.There are two modes of communication between agents: pull–based and push–based. Inpush-based communication, agents periodically send information to specific recipients with-out being requested. On the other hand, in pull-based communication, agents have to request for information and wait for replies.Until now pull–based communication is prevalent in multiagent research. For efficiency,a piece of information obtained from other agents may be used to answer many queries.We formulate general requirements for information sharing in agent systems. We presenta formal framework for studying stabilization with/without information sharing, in which agents are represented by abductive logic programs with abducibles being literals the agents could sense or receive from other agents. We point out that surprisingly, information sharing can cause non–stabilization of many systems of information agents. We then give sufficient conditions to ensure stabilization. In practice, refreshing information is a common technique to ensure the correctness of information. Our results show that in general, refreshing is not sufficient to ensure the correctness of information.Though push–based communication is used widely in network protocols, emails, etc., it has never got appropriate attention in multiagent research. We Introduce a formal framework for studying stabilization of push–based communication and give sufficient conditions to guar-antee it. We study conceptual relationships between two modes of communication and showthat in practical applications, push–based communication could be conceptually viewed as a special kind of pull–based communication with informationsharing.Unfortunately general sufficient conditions for stabilization are rather strong and so it is nota simple task to ensure stabilization even if agents are fully cooperative, honestly exchange information and every change in the environment is immediately sensed by some of the agents. We formalize a routing system using the Routing Information Protocol (RIP) in our framework as an example to demonstrate the practical importance of stabilization. We demonstrate that an argumentation based semantics of abductive logic programs pro-vides an appropriate platform to model agents and agent communication in this thesis. To Model more complex agents which operate in dynamic environments and perform complicated decision making, dialogs, dispute resolutions, belief revision, etc. and hence their programs may contain rules with defeasible priorities to resolve conflicts, we believe that Extended Argumentation Frameworks (EAFs) introduced by Gabbay where not only arguments but also attacks are allowed to be attacked, and that developed by Modgil for dealing with defeasible priorities and preferences could be useful. Similar to the semantics of ab-stract argumentation, Modgil’s semantics of EAFs is also based on a notion of acceptability of arguments. Though Modgil’s semantics is natural and intuitive, the characteristic function based on it is nonmonotonic. This is a rather intriguing problem as it runs counter to a commonsense intuition that the more arguments one has, the more arguments one is capable to defend. To address this problem, we propose a semantics for EAFs based on an inductive defense relation that preserves fundamental properties of abstract argumentation including the monotonicity of the characteristic function and identify relationships between the new semantics and those of Gabbay and Modgil. We show that any extension of the semantics proposed by Gabbay and Modgil contains a sceptical part being an extension of our seman-tics and a credulous part resulted from its credulousness towards the acceptance of attacks.We further introduce a hierarchical form of Modgil’s extended argumentation which assures that all proposed semantics coincide. |
| Year | 2010 |
| Corresponding Series Added Entry | Asian Institute of Technology. Dissertation ; no. CS-10-03 |
| Type | Dissertation |
| School | School of Engineering and Technology (SET) |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Computer Science (CS) |
| Chairperson(s) | Phan Minh Dung; |
| Examination Committee(s) | Huynh Trung Luong;Honda, Kiyoshi; |
| Scholarship Donor(s) | Austria (ADA); |
| Degree | Thesis (Ph.D.) - Asian Institute of Technology, 2010 |