Similarity-based data transmission reduction solution for edge-cloud collaborative AI

Abstract

Edge-cloud collaborative processing for IoT data is a relatively new approach that tries to solve processing and network issues in IoT systems. It consists of splitting the processing done by a Neural Network model into edge part and cloud part in order to solve network, privacy and load issues. However, it also has it shortcomings such as the big size of the edge part's output that has to be transmitted to the cloud. In this paper, we are proposing a data transmission reduction method for edge-cloud collaborative solutions that is based on data similarities in stationary objects. The performed experiments proved that we were able to reduce 62% of the data sent.