Resource management issues in wireless sensor networks.

摘要

Sensor networks consist of a number of small sensing devices that are battery-operated and have very limited capabilities. When such sensors are deployed they form a wireless ad-hoc network to communicate with each other and with data processing centers. After deployment, the network is typically required to perform multiple tasks or missions. Because of the limited number of sensors and the possibly large number of missions competition will arise. In such cases, it might not be possible to satisfy the requirements of all missions using available sensors. So, algorithms that decide how the resources are assigned become important. The problem of assigning sensors to missions becomes especially challenging when directional sensors are used as each such sensor can be assigned to at most one mission. Algorithms to solve this problem need to consider the demand and importance for the different missions to decide which missions to fulfill. They also need to assign specific sensors to each mission. This assignment should depend on how suitable a sensor is to the mission and the utility, i.e. amount/quality of information it can provide.;In this dissertation, we study different sensor-mission assignment problems. We design algorithms that attempt to maximize the overall utility of the network and evaluate their performance using simulations on randomly generated problem instances. We propose algorithms that are centralized, in which all assignment decisions are made by a single node that has a global view of the network, and others that are distributed making them more suitable for actual implementations. We consider static environments, in which all mission requests arrive at once, and dynamic environments, in which missions arrive and depart over time. We address different constraints and design specific solutions for each. Namely, we consider energy, budget, computational power and bandwidth constraints. Although most of the problems we consider are NP-hard, our practical algorithms manage to improve the utility of the network and in many cases achieve near optimal performance.