Multi-agent search in a set: distribution of effort and estimating the efficiency

Ilya A. Chernov

IAMR KRC RAS

Abstract: We consider a search model on the set, in which each point is a search location for valuable objects. Searchers of different efficiencies are distributed across a multitude, choosing locations based on the apriori idea of how promising each location is. Those who choose the same location compete with each other. We show that in quite free assumptions the distribution of agents coincides with that of the prospects. This allows us to estimate the specific flow of results, which is constant on the set. In order to more accurately predict this flow, which depends on the performance of individual agents, we offer a policy of bids: a reward for winning one. Cases of various distributions of uncertainty in the agent assessment of their own productivity are considered and the difference between the declared productivity and the average one is estimated.

Keywords: volunteer computing, desktop grid, throughput prediction, distributed search.

UDC: 519.83
BBK: 22.176

Received: 03.10.2019
Revised: 17.01.2020
Accepted: 20.05.2020