|
内容記述 |
In wireless ad hoc networks, all nodes cooperate to provide network services. <br />Due to the limited radio transmission range, data packets are usually forwarded <br />through multiple relay nodes before they reach the destinations. If a node always <br />serves as a relay to transmit the packets, it may quickly use up its own energy and <br />other resources. Therefore, some nodes use a selfish approach: they try to avoid <br />forwarding the packets. Such selfish behavior would probably cause the network to <br />break down. Selfish nodes are common within ad hoc networks because they are <br />managed by different authorities. <br /> In this thesis, the node cooperation problems are analyzed in two steps: 1) <br />a game theoretic analysis is provided to stimulate node to cooperate; 2) a price- <br />demand function based incentive model is proposed to optimize the nodes' service <br />demand and service provision, and encourage the relay nodes to be honest. <br /> Firstly, a game theoretic analysis is proposed to study node cooperation. In the <br />related chapter, a "payment and compensation" scheme is used as a less-aggressive <br />way to avoid nodes' non-cooperative behavior. It is assumed that once a packet is <br />sent from a source node, the packet is associated with a sending fee, i.e, when a node <br />needs sending the packets as a source node, a sending fee is required (e.g. reasonably <br />some money). The fee is adjustable according to the network status, whereas the <br />node can also accept or reject the fee. In order to induce voluntary forwarding, the <br />network will also compensate the nodes who consume their energy in forwarding the <br />packets for others. If I think the sending fee as the penalties to the source nodes and <br />the compensation fee as the encouragement to relay nodes, then local optimization <br />of the node, (the desired performance plus the compensat,ion then minus the cost <br />to be paid) will yield an optimal point. Each node can only select its own packet <br />generation strategy, however the final utility of each node is determined by the <br />strategy set constituted by the other nodes. With the game theoretic analysis, I <br />found that by introducing an incentive pricing policy "payment and compensation"<br />the relay nodes have less motivation to drop the packets. However, I also found <br />that game theoretic literature may not be directly applicable in the scenario where <br />cheating nodes exist and how to reasonably charge the source nodes and compensates <br />the relay nodes.<br /> Therefore, secondly, a price-demand function based incentive model (PDM) is <br />proposed. In the PDM model, the network is modeled as a market, where the pricing<br />is determined by the source node's demahd and the relay node's service supply. The<br />source nodes make use of a price-demand function, which allocates payments to <br />the service provider (relay nodes). The relay nodes are encouraged to cooperate <br />in the PDM model, which is based on the assumption that each relay node wishes <br />to maximize its payoff. Then the source nodes can optimize their prices and the <br />number of sending packets to satisfy the relay nodes' payoff requirement. Once <br />the payoff requirements of the relay nodes are satisfied, the relay nodes have no <br />reason to be selfish. In the PDM model, a source node that has packets to send <br />initially broadcasts RREQ in the network. Once the relay node(s) are selected, each <br />relay node replies to the source node for its forwarding cost. Then the source node <br />calculates the price of the sendihg packets it will pay for each relay node and the <br />number of packets it will send. According to the source nodes' demand, it chooses <br />the route with the lowest payment or the route with the la'rgest number of sending <br />packets. The PDM pricing model seeks to address two main issues: 1) to determine <br />how much to charge the source nodes and how much to compensate the relay nodes; <br />2)to avoid the relay nodes to dishonestly report their forwarding costs. Hence, the <br />contributions are summarized as follows: 1) The relay nodes intend to dishonestly <br />report their forwarding cost to gain a high payoff from the source nodes, which <br />obviously contradicts with the motivation to stimulate cooperation. In the PDM <br />model, however, the relay nodes will have no reason to report a false forwarding cost, <br />since only telling the truth guarantees the relay nodes' final payoff. Such a property <br />is shown by the proof. 2) The PDM pricing model reflects the relationship between <br />the service demand of the source nodes and the service supply of the relay nodes. <br />The PDM model can save money for the source nodes for sending the packets, which <br />is indicated by the simulation results. <br /> |
要旨・審査要旨 (288.6 kB)
