Vidal's library
Title: Solving the Auction-Based Task Allocation Problem in an Open Environment
Author: David Sarne and Sarit Kraus
Book Tittle: Proceedings of Twentieth National Conference on Artificial Intelligence
Pages: 164--169
Year: 2005
Crossref: aaai05
Abstract: In this paper we analyze the process of allocating tasks to self-interested agents in uncertain changing open environments. The allocator in our model is responsible for the performance of dynamically arriving tasks using a second price reverse auction as the allocation protocol. Since the agents are self-interested (i.e. each agent attempts to maximize its own revenue), previous models concerning cooperative agents aiming for a joint goal are not applicable. Thus the main challenge is to identify a set of equilibrium strategies - a stable solution where no agent can benefit from changing its strategy given the other agents' strategies - for any specific environmental settings. We formulate the model and discuss the difficulty in extracting the agents' equilibrium strategies directly from the model's equations. Consequently we propose an efficient algorithm to accurately approximate the agents' equilibrium strategies. A comparative illustration through simulation of the system performance in a closed and open environments is given, emphasizing the advantage of the allocator operating in the latter environment, reaching results close to those obtained by a central enforceable allocation.

Cited by 3  -  Google Scholar

@InProceedings{sarne05a,
  author =	 {David Sarne and Sarit Kraus},
  title =	 {Solving the Auction-Based Task Allocation Problem in
                  an Open Environment},
  booktitle =	 {Proceedings of Twentieth National Conference on
                  Artificial Intelligence},
  crossref =	 {aaai05},
  pages =	 {164--169},
  year =	 2005,
  abstract =	 {In this paper we analyze the process of allocating
                  tasks to self-interested agents in uncertain
                  changing open environments. The allocator in our
                  model is responsible for the performance of
                  dynamically arriving tasks using a second price
                  reverse auction as the allocation protocol. Since
                  the agents are self-interested (i.e. each agent
                  attempts to maximize its own revenue), previous
                  models concerning cooperative agents aiming for a
                  joint goal are not applicable. Thus the main
                  challenge is to identify a set of equilibrium
                  strategies - a stable solution where no agent can
                  benefit from changing its strategy given the other
                  agents' strategies - for any specific environmental
                  settings. We formulate the model and discuss the
                  difficulty in extracting the agents' equilibrium
                  strategies directly from the model's
                  equations. Consequently we propose an efficient
                  algorithm to accurately approximate the agents'
                  equilibrium strategies. A comparative illustration
                  through simulation of the system performance in a
                  closed and open environments is given, emphasizing
                  the advantage of the allocator operating in the
                  latter environment, reaching results close to those
                  obtained by a central enforceable allocation.},
  keywords =     {multiagent auctions task-allocation},
  url = 	 {http://jmvidal.cse.sc.edu/library/sarne05a.pdf},
  cluster = 	 {2581919643919100741}
}
Last modified: Wed Mar 9 10:16:28 EST 2011