Record:   Prev Next
Author Dorsey, John G
Title Game-theoretic power management in mobile ad hoc networks
book jacket
Descript 362 p
Note Source: Dissertation Abstracts International, Volume: 65-08, Section: B, page: 4103
Adviser: Daniel P. Siewiorek
Thesis (Ph.D.)--Carnegie Mellon University, 2004
This thesis applies game-theoretic mechanism design to the management of energy consumption in mobile ad hoc networks. In this setting, routes must be selected based on the private energy information held at each node. The problem is to develop a negotiation procedure that incents the nodes to truthfully reveal their preferences over network configurations
Popular spread-spectrum wireless network interfaces, such as 802.11, experience high energy consumption while in the idle state. This induces energy complementarity across concurrent traffic flows at a node, in which the marginal energy costs of servicing additional flows are small. A strategy-proof mechanism for this environment must account for this cost behavior when routes overlap. No previous mechanism for network routing satisfies this requirement
We present Exchange Power Management, a complement to on-demand source routing protocols which enables route selection via a strategy-proof combinatorial exchange. This is the first application of mechanism design to real wireless protocols, and the first to be evaluated in a realistic wireless network simulation environment. Experimental results show a reduction in energy variability by a factor of 5 relative to 802.11 power management. When unaffordable routes are used, this improvement increases to a factor of 12. Average-case energy consumption is competitive with previous work. We also show a technique for reducing route setup latency under power management by up to a factor of 16
This work is the first to take a systems view of mechanism design application to ad hoc networks. The results of this research characterize the kinds of energy performance improvements that could be expected from negotiation-based power management. Future work will refine the fault-tolerance and scalability of the distributed protocol, increase the sophistication of agent valuation functions, and examine application awareness of exchange-based route selection
School code: 0041
Host Item Dissertation Abstracts International 65-08B
Subject Computer Science
Engineering, Electronics and Electrical
Alt Author Carnegie Mellon University
Record:   Prev Next