Belief evaluation for cooperation enforcement

Abstract

In autonomous mobile ad hoc networks (MANETs) where each user is its own authority, the issue of cooperation enforcement must be solved first in order to enable networking functionalities such as packet forwarding, which becomes very difficult under noise and imperfect monitoring. In this chapter, we consider cooperation enforcement in autonomous MANETs under noise and imperfect observation and study basic packet forwarding among users using repeated-game models with imperfect information. A belief-evaluation framework is presented to obtain cooperation-enforcement packet-forwarding strategies that are based solely on each node's private information, including its own past actions and imperfect observation of other nodes' information. More importantly, we not only show that the strategy with a belief system can maintain the cooperation paradigm but also establish its performance bounds. The simulation results illustrate that the belief-evaluation framework can enforce cooperation with only a small performance degradation compared with the unconditionally cooperative outcomes when noise and imperfect observation exist. Introduction One major drawback of the existing game-theoretic analyses on cooperation in autonomous ad hoc networks is that all of them have assumed perfect observation, and most of them have not considered the effect of noise on the strategy design. However, in autonomous ad hoc networks, even when a node has decided to forward a packet for another node, this packet may still be dropped due to link breakage or transmission errors.