Collaborative processing among sensors to fulfill given tasks is a promising solution to save significant energy in resource - limited wireless sensor networks. Quality of Service such as lifetime and latency is largely affected by how tasks are mapped to sensors in a network. Due to the limitations of wireless sensor networks, existing algorithms cannot be directly used. This paper presents an efficient allocating algorithm that allocates a set of real-time tasks with dependencies onto a sensor network. The proposed algorithm comprises linear task clustering algorithm and sensor assignment mechanism based on a task duplication and migration scheme. It simultaneously schedules the computation tasks and associated communication events of real time applications. It reduces inter-task communication costs and moderates local communication overhead incurred due to communication medium contention. Performance is evaluated through experiments with both randomly generated Directed Acyclic Graph (DAG) and real-world applications. Simulated results and qualitative comparisons with the most related literature, Multi-Hop Task Mapping and Scheduling (MTMS), Distributed Computing Architecture (DCA), and EnergyBalance Task Allocation (EBTA), demonstrated that the proposed scheme significantly surpasses the other approaches in terms of deadline missing ratio, schedule length, and total application energy consumption.
展开▼