Monitoring schedules for randomly deployed sensor networks

Abstract

Given n sensors and m targets, a monitoring schedule is a partition of the sensor set such that each part of the partition can monitor all targets. Monitoring schedules are used to maximize the time all targets are monitored when there is no possibility of replacing the batteries of the sensors. Each part of the partition is used for one unit of time, and thus the goal is to maximize the number of parts in the partition.We present distributed algorithms for Monitoring Schedule under the following assumptions: 1) identical sensors can each monitor all targets within a certain radius, 2) the n sensors are randomly distributed uniformly in a large square containing the targets, 3) the number of sensors is high enough given the area the square, and 4) the communication range is twice the sensing range (thus any two sensors which can monitor the same target can communicate in one hop). Our results hold with high probability. With the further assumptions that the sensors are capable (for example, by GPS) of knowing their exact geographic position, and targets fill out the square, our schedule has at least (1-e) opt parts, where opt is the optimum solution. Without geographic position we show that a previously proposed distributed algorithm can be modified to achieve a constant approximation ratio. Our algorithms run in a polylogarithmic number of communication rounds, with the exact running time depending on assumptions on the information a sensor receives when packets collide.