Design and Evaluation of an Edge-Enabled Wireless Sensor Network Using Heterogeneous Multi-Core Architecture on Raspberry Pi

This paper designs a framework based on EdgeSched-H, in which it enables and uses Wireless Sensor Networks (WSNs) these networks use Raspberry pi as their device. The need of these is highlighted through the literature review on why the traditional methods only depends upon either the decentralized CPU and barely depend upon the GPU models as well. So the methodology used in this paper integrates dynamic CPU-GPU task scheduling, along with MQTT as the communication medium. For the allocation of works based on the weightage of the tasks its uses DVFS and for latency requirements as well. The framework has a sensor acquisition and adaptive scheduling. Heavyweight tasks and the parallel tasks are performed by GPU. This framework shows significant improvement over cloud computing which mostly depend upon CPU based models. The results include significant Latency reduction and the power consumption being reduced and accuracy is also well maintained at a higher level.