Selfish Agents and Economic Aspects

Abstract

Routing algorithms in wireless networks can be developed using different techniques. In a network consisting of selfish nodes, traditional approaches will fail, since nodes are not interested in forwarding foreign data and might refuse to cooperate. In this chapter we discuss algorithmic mechanism design, which provides a possibility to deal with selfishness of network nodes.

Nodes in a network are called selfish, if they are not providing own resources to other nodes for free, which is usually bandwidth for transit data transfers. Forwarding foreign data in a wireless network means for a batterypowered node spending valuable energy, which can be considered as an expence for a node. We will call such expences costs. A rational node will provide a service when offered compensation for its costs, wich could be some kind of virtual currency. The value of such a transaction to a node is the difference between the amount of compensation and the own costs of providing this service. In this chapter we will not consider malicious nodes, for example, nodes with the intent of jamming or tapping the network.

Selfish entities are called agents and are assumed to be rational and responding to incentives when asked to cooperate. For each network model and given routing problem, the aim is to design a cheating-proof mechanism, which computes outcome (the route) and payoff function in polynomial time and minimizes, if possible, the overall route cost.