Effect of Mobility on the Sensing and Communication Performance of Mobile Sensor Networks
Veria Havary-Nassab · TSpace (University of Toronto) · 2015
We study the effect of mobility on the sensing and communication performance of a sensor network with mobile nodes. Communicating the sensed data clearly necessitates that the sensor node is able to connect to adjacent (sensor or sink) nodes to be able to exchange its information. Same node to node links can also be used in the sensing phase if a cooperative sensing scheme is in place. Assuming that the phenomenon being sensed generates a sparse and spatially correlated signal across the sensors, the theory of distributed compressive sensing suggests that cooperative sensing can facilitate the sensing process. We study the effect of mobility on the probability that a sensor node can connect to its neighboring nodes either to communicate its sensed data or to initiate a cooperative sensing procedure. We show that, by increasing the probability of such connections, mobility enhances both communication and sensing performance of mobile sensor networks.We first study how mobility of the nodes affects the communication performance of a mobile network. Based on a Wiener process mobility model, and in a network with mobile sensor and sink nodes, we study a repetition based approach for sensor node to sink node communication, where sensor nodes attempt communicating their sensed data with sink nodes repeatedly between two sensing instants. Using a stochastic framework, we prove that the average number of successful communications is an increasing function of the sensor node's mobility.Then, we study the sensing performance of the above network based on a cooperative sensing scheme. In this scheme, sensors share their samples and perform the sensing cooperatively, reducing the required number of samples per sensor node. The reduction in the number of samples each sensor needs to obtain, namely the sensing gain, depends on the level of correlation between such samples. Adopting a joint sparsity model for the correlation between the samples of the sparse signal, we show that the mobility of the sensor nodes can increase the chances of higher correlation between the samples, leading to a reduction in the number of requires samples.