Game Theoretic Approaches For Conflicts In Complex Communication Networks

Wireless sensor network and multihop wireless network are both complexcommunication networks, and using self-organized method to maintain continu-ous operation. Nodes in these networks is unable to get the information aboutthe whole network, thus there must be interactions or even con?ictions betweenthe decisions/behaviors of nodes. It is of great signi?cance to learn how the nodesinteract with each other and then we can make proper scheduling accordingly,which are the right focuses of this paper.First, we develop a key-level control mechanism for a kind of sensor net-works with hierarchical architecture, where the key level is made up of nodes onehop away from the sink node, which are named OHSs. To achieve more e?cientenergy consumption and more reliable communication, a special Medium AccessControl protocol is added for the key level to schedule OHSs, then we also employpower control and admission control that compose the key-level control. After-wards, we model the key-level problem as a non-cooperative game, and furtherapply a double pricing scheme in the de?nition of player’s utility so that NashEquilibrium can be obtained. Based on the game theoretic analysis, a distributedalgorithm is proposed, and the later simulation shows its e?ectiveness.Second, we formulate a network congestion game for the routing congestionproblem in a more generalized multihop wireless network. Our goal is to assignthe source nodes to the routs as evenly as possible so as to reduce the whole la-tency. Dynamic Pricing (DP), a distributed algorithm based on pricing scheme,is presented to compute the Nash Equilibrium in the game, which assumes thateach source node needs to pay its using rout for the routing service. We ?nd theexecuted time of DP a loose upper bound, and the simulation results demon-strate that it performs much faster in practice than the upper bound. Besides,a comparison between DP and a centralized algorithm based on Cost ReducingPath is made, and the result shows that more than 90% of the simulation runscan get optimal assignment.