Resource efficiency in ad hoc wireless networks: problems and solutions

Ad hoc network is promising due to its versatility and convenience. Resource scarcity is one of the characteristics in ad hoc networks. This thesis studied the problems of making efficient use of energy resource and radio spectrum. A broadcast session traffic uses a broadcast tree to disseminate information. To be energy efficient, the broadcast tree construction needs to consider the transmission power needed by each transmitting node. We study the problem of finding the tree that minimizes the maximum transmission power, and in case that each node has different energy reserve, we find a tree that balances the energy consumption. Ad hoc networks are infrastructureless. However, network topology is an important virtual structure above which the upper level routing protocols are implemented. The underlying topology will affect the performance of upper level applications. To improve energy efficiency, we can control the link topology by adjusting the node transmission power. This thesis studied the topology control problem to achieve the basic network connectivity. We studied the problem for different objective functions, such as minimum total energy, or balanced energy distribution. Media access control is another important issue in wireless networks. This thesis investigated channel assignment problem in ad hoc wireless networks such that interference between users are eliminated and the total number of channels is minimum. Interference can be caused by either direct collision from users that can hear each other or indirect collision from stations that are not in the range of each other but simultaneously transmit to the same destination. We modeled the wireless networks using a new class of disk graphs, and we used graph coloring approach to handle collision avoidance. Sensor network is an application of ad hoc technology in large scale networks. Sensors in a network can cooperatively gather information from a specified region of observation and relay this information to a base station. For stochastically deployed sensor networks, some area is redundantly covered. We divide sensors into mutual exclusive subsets and put them alternatively in active mode and powered-down mode, such that there is only one set of sensors active at any time and the set of active nodes can fully cover the area. We studied the complexity of the maximum disjoint cover problem and provided heuristics to maximize the number of disjoint covers, and this effort significantly extended the sensor network lifetime.