Data acquisition scheduling for wireless sensors networks

Wireless sensor networks have been increasingly studied by industry and academic institutions due to their potential application in environmental and geothermal monitoring, security enhancement among others. Such technology would play an important role in future societies and would be the key tool to make improvements in productivity and efficiency fast and on a large scale. In this work it is proposed to use Divisible Load scheduling Theory (DLT) as a mathematical tool to study data load distribution on wireless sensors networks under time and monetary cost constraints. Pursuing this aim, five different studies have been conducted. The first one considered the total monetary cost optimization, and a sensitivity analysis in a single level tree network. The problem of load distribution sequencing for optimizing monetary cost in a single level tree network is reviewed and simulation results showed that the monetary cost function depends on the fractions of loads assigned to each processor. In addition, a sensitivity analysis performed on the monetary cost function suggested a complex relationship between the cost function and the network parameters. The second study explores such relationship proposing a strategy to study the monetary total cost in a single level tree star network as a function of a non-linear load parameter. The third study investigated three scheduling load protocols on wireless sensors networks. A novel closed form solution was found for each of the protocols presented for optimum finish, reporting and pre-processing time for a single level tree sensor network including an immediate measuring data feature. In the fourth study an adaptable data scheduling methodology is presented. The schedule protocol presented here takes into account the amount of load processed from previous load assignments based on distributing the new incoming load to the sensor network aiming to minimize the total finish time of processing the entire Divisible Load Job Task (DLJT) and reducing the idle state of each processor. Finally, the performance in terms of speedup for a single level tree network with multiple links and cores is investigated.