A Novel Approach to Distributed Model Aggregation using Apache Kafka

Abstract

Multi-Access Edge Computing (MEC) is attracting a lot of interest because it complements cloud-based approaches. Indeed, MEC is opening up in the direction of reducing both interaction delays and data sharing, called Cyber-Physical Systems (CPSs). In the near future, edge technologies will be a fundamental tool to better support time-dependent and data-intensive applications. In this context, this work explores existing and emerging platforms for MEC and human-centric applications, and proposes a suitable architecture that can be used in the context of autonomous vehicle systems.The proposed architecture will support scalable communication among sensing devices and edge/cloud computing platforms, as well as orchestrate services for computing, storage, and learning with the use of an Information-centric paradigm such as Apache Kafka CCS CONCEPTS• Networks → Layering; • Human-centered computing → Mobile devices; • Computer systems organization → Architectures.

Artificial intelligence of things at the edge: Scalable and efficient distributed learning for massive scenarios

“…This work has two main objectives: to develop an efficient communication architecture for FL suitable for mobile nodes and to efficiently select clients using this communication paradigm to reduce broker congestion. The proposed infrastructure is based on edge computing to leverage the processing power of edge servers capable of handling distributed tasks such as FL with mobile nodes that require minimizing communication latency with the central server [35,36].…”