Agent Based Risk Management Methods for Speculative Actions
In multiagent systems, a cooperative action requires the mutual agreement of multiple agents which is generally achieved by exchanging messages. Any delay in message transfer will, however, delay the realization of agreement, and this may reduce the effectiveness of the cooperative action. One solution is to use speculative actions, actions taken before agreement is reached with the goal being to ”lock in” the benefits of the cooperative action; its downside is the penalty incurred in unwinding the speculative actions if indeed the agents do not reach agreement. In this framework, we have two risks; the risk of losing the benefits of the cooperative action and the risk of unwinding the speculative actions. It is clear that some form of risk management is needed. In this paper, we propose two risk management methods, the hybrid method and the leveled method, which are viewed as a single agent approach and a multiagent approach, respectively. We discuss their advantages using the meeting room reservation problem.
Unable to display preview. Download preview PDF.
- 1.Burton, F.W.: Speculative Computation, Parallelism, and Functional Programming. IEEE Transactions on Computers C-34, 1190–1193 (1985)Google Scholar
- 2.Halstead Jr., R.H.: Parallel Symbolic Computing. IEEE Computer 19(8), 35–43 (1986)Google Scholar
- 3.Sandholm, T., Lesser, V.: Leveled Commitment Contracting: A Backtracking Instrument for Multiagent Systems. AI Magazine 23(3), 89–100 (2002)Google Scholar
- 4.Satoh, K., Inoue, K., Iwanuma, K., Sakama, C.: Speculative Computation by Abduction under Incomplete Communication Environments. In: Proceedings of the Fourth International Conference on MultiAgent Systems, pp. 263–270 (2000)Google Scholar
- 6.Smith, R.G., Davis, R.: Frameworks for Cooperation in Distributed Problem Solving. IEEE Trans. on System, Man, and Cybernetics SMC-11(1), 61–70 (1981)Google Scholar