| Co-Investigator(Kenkyū-buntansha) |
YOKOO Makoto Kyushu University, Graduate School of Information Science and engineering, Professor, 大学院システム情報科学研究院, 教授 (20380678)
MATSUBARA Shigeo Kyoto University, Graduate School of Informatics, Associate Professor, 情報学研究科, 助教授 (80396118)
IWASAKI Atsushi Kyushu University, Graduate School of Information Science and electrical engineering, Research associate, 大学院システム情報科学研究院, 助手 (30380679)
|
|---|
| Budget Amount *help |
¥3,400,000 (Direct Cost: ¥3,400,000)
Fiscal Year 2006: ¥1,800,000 (Direct Cost: ¥1,800,000)
Fiscal Year 2005: ¥1,600,000 (Direct Cost: ¥1,600,000)
|
|---|
| Research Abstract |
This research aims to evaluate the market-based fair and efficient protocol, in order to apply it to the allocations of networked resources for its future applications to this domain. In ad hoc networks used in P2P and the decentralized sensor network, for example, individual nodes are owned by different persons and designed based different specifications. In this case, it is necessary to consider the incentives, that is, reward, to transmit data to each node appropriately. The auction protocol is often used for the decision of this reward. However, by using the fake (false-name) node or by conspiring with other nodes, a certain node can acquire the reward illegally in conventional protocols. We theoretically showed that this type of illegal behaviors cannot be prevented even in Vickrey-Clarke-Groves protocol (VCG) in the research period by this grant.
We then proposed Reserve-Cost protocol (RC), which is the extension of VCG by introducing the penalty proportional to the number of agen
… More
ts (nodes) who manage the network route. We also clarified that the RC is false-name proof, that is, the fairness of RC protocol is not influenced by the false-name bids. In addition, we also showed that RC is more efficient than VCG about 60-80% by small-scale network simulation.
Moreover, it is necessary for agents to decide, by using some protocols such as auctions, where to receive/send data based on locally available information in an actual network. This corresponds to the selection of an awarder to some degree when multiple bidding agents (this corresponds to servers in this case) are identified as the appropriate for awarders.
In this research, we investigated and analyzed the phenomenon occurring when such a resource allocation protocol was used in a large-scale multi-agent system such as network.
In this type of systems, many demands like the resource allocation on the network occurs simultaneously from many different agents independently, thus the entire efficiency falls down. We also identified that a little bit of fluctuation can significantly improved the entire performance by avoiding concentration. Less
|
