Data Capsule: Representation of Heterogeneous Data in Cloud-Edge Computing

Abstract

Nowadays, robots (including non-humanoid ones, like self-driving cars) are part of the most promising technologies, and they rise various computing requirements. Some of those requirements came from the fact that the robots are involved in highly dynamic environments and have to execute complex decision algorithms in real-time, while other requirements ask for batch processing of big data compatible datasets. In this paper, we propose a cloud architecture for optimizing data processing using a cloud-edge infrastructure. Besides the computational architecture, we develop a mathematical model for each type of entity in our proposal and a formal description of a data capsule, which represents a generic and flexible representation for unstructured units of data in time series databases. The architecture includes multiple processing platforms. We evaluate the proposed model in edge-cloud computing platforms designed for robots that run machine learning tasks.