NESL Technical Report #: 2003-9-3
Abstract: Tasks for sensor networks often have constraints such as lifetime and latency. Performance of a sensor network like energy consumption and latency is largely affected by how the task is mapped to the nodes in network. This paper presents an energy-efcient task assignment and migration framework for sensor networks. With proposed framework, optimal task transformation and assignment is sought so as to minimize given cost function. Cost function reects total energy consumption in a network, maximum energy consumption among nodes and maximum latency. Simulated annealing method is used to solve the task transformation and assignment problem. For run-time support, we developed distributed task migration method. While executing tasks in a node, if the remaining energy is less than threshold level, such tasks are migrated into neighbor node which is healthier. With demonstrative examples, we evaluate our task assignment framework and distributed task migration features through Sensorsim simulation which is an extension of ns-2 simulator. We demonstrate that our framework can handle node heterogeneity and various cost function requirements. The experimental results show that elaborate assignments can save more energy than simple assignment of aggregation functions and help in improving performance.
Publication Forum: CENS Tech Reports
Public Document?: Yes
NESL Document?: Yes
Document category: ReportBack