EC-DNNs: An efficient Training Engine for Edge Computing-Based agriculture IoT

Abstract:
As 5G technology takes hold and matures, mobile computing and the Internet of Things (IoT) in agriculture become more prevalent in the real world, billions of agricultural mobile devices and agricultural IoT terminals are connected to the network, generating countless amounts of data at the edge of the network. The need for constant viewing of crop growth and conditions in the farming environment in agriculture results in a large amount of data being generated on mobile devices and agricultural IoT terminals. In this paper, we propose EC-DNNs, a framework where data information is multivariate classified by DNNs, and deep learning collaborative reasoning is performed through end-device - edge node - cloud server - edge collaboration.EC-DNNNs framework embeds DNNs processing algorithms at edge nodes to selectively keep regular data locally and transmit important information to the cloud, reducing bandwidth waste and lowering propagation real delay, thus improving computational performance and achieving balanced load. The experimental results show that the EC-DNNs framework is effective in achieving low-latency edge intelligence for agricultural IoT end devices and provides a good solution for solving the computation of information transmission of agricultural IoT data.