A Principal-Agent Theory based Game Theoretic Incentive Method for Packet Forwarding in Ad Hoc Networks

摘要

Dropping packets is one of the most devastating threatens on a performance of ad hoc networks. Among various motivations of dropping packets, selfishness and malicious attack are the main two factors. However, there are still some other cases of dropping packets caused by networking context such as mobility, congestion, and traffic load. Incentive mechanism is a typical method to induce node to participate in relaying packets for other adjacent nodes. However, most of them neglect a noise of dropping packets by networking context or just omit this important factor for brevity. In this paper, a principal-agent theory based game theoretic incentive method is proposed for encouraging each node contributing into packet forwarding, which will result in a high performance cooperative network. Holmstrom-Milgrom model, which is a simplified principal-agent model with a linear reward scheme, is implemented to model the relationship between two adjacent nodes. In each two pair-wise nodes model, the node requests for relaying packets is seen as a principal node while the node forwarding packets is denoted as an agent node. A credit given by principal node to agent node as incentive compensation is the contract, which is designed by principal node based on agent node's effort in forwarding packets. Dropping packets by networking is also taken into consideration as an exogenous uncertainty. It is proven that there is an optimal incentive strategy can encourage packet forwarding even under noise effect. Numerical evaluations are also conducted to evaluate the performance of the proposed method.