Presentation | 2012-03-13 Efficient Task Allocation by Learning and Reorganization of Hierarchical Agent Network Based on Observed Delay Kazuki URAKAWA, Toshiharu SUGAWARA, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | We propose a method for efficient task allocation by changing the network structure among agents to adapt to the environmental changes. As services in a distributed environment like the Internet often consist of a number of service elements, the task for the service can be modeled as a set of subtasks and can be achieved by executing all the subtasks. So they are executed in appropriate agents that have required resources and/or functionalities to In order to realize the corresponding services, this type of problem is formulated as a team formation problem in which all (sub)tasks are allocated to a number of agents (team). A number of studies addressed this issue; proposed a method by adding links between agents based on the amount of unused resources in a task-oriented domain. However, this kind of methods have the drawback that the reorganization stops in an earlier stage of learning. It also retains the generated links, but when the types of the requested tasks change, it could not adapt quickly to the new distribution of incoming tasks. The method proposed in this paper generates a new link that can allocate tasks to unbusy agents and eliminates the link that is hardly used based on the numbers of the processed tasks in each agent. We experimentally show that the proposed method can exhibit higher performance and adapt to the changes of requested task patterns. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Multi-agent reinforcement learning / Distributed cooperative system / Reorganization / Team formation |
Paper # | AI2011-45 |
Date of Issue |
Conference Information | |
Committee | AI |
---|---|
Conference Date | 2012/3/6(1days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | |
Chair | |
Vice Chair | |
Secretary | |
Assistant |
Paper Information | |
Registration To | Artificial Intelligence and Knowledge-Based Processing (AI) |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Efficient Task Allocation by Learning and Reorganization of Hierarchical Agent Network Based on Observed Delay |
Sub Title (in English) | |
Keyword(1) | Multi-agent reinforcement learning |
Keyword(2) | Distributed cooperative system |
Keyword(3) | Reorganization |
Keyword(4) | Team formation |
1st Author's Name | Kazuki URAKAWA |
1st Author's Affiliation | Waseda University Fundamental Science and Engineering() |
2nd Author's Name | Toshiharu SUGAWARA |
2nd Author's Affiliation | Waseda University Fundamental Science and Engineering |
Date | 2012-03-13 |
Paper # | AI2011-45 |
Volume (vol) | vol.111 |
Number (no) | 474 |
Page | pp.pp.- |
#Pages | 6 |
Date of Issue |