Towards 5G, Artificial neural networks and EDGE computing Part-2

Towards 5G, Artificial neural networks and EDGE computing Part-2
December 12, 2019

In the first part of this blog “Towards 5G, artificial neural networks, and EDGE computing- Part 1”, we discussed the basic understanding of 5G, EDGE computing, and ANN (artificial neural networks). In this blog, we will see how 5G, along with ANN and EDGE, can play a pivotal role in the evolution of communication.

5G network connectivity requires higher bandwidth and lower latency which is not supported by current internet infrastructure because of the processing and forwarding nature and network congestion. So, to overcome this, the 5G network needs to process huge bytes of big data at edge and have better storage capability, as well as advanced CPU and GPU computing for intelligent decision making. These requirements can be fulfilled by a combination of edge computing and ANN-based machine learning algorithms. Most of the 5G applications like augmented reality/virtual reality (AR/VR), AV, eHealth, and V2X cannot withstand high delays across the internet or cloud networks, so an intelligent edge process can reduce the data traffic in the core network which will have a huge impact on congestion in the cloud network and across the internet. Also, the 5G network slice aided by machine learning will play a crucial role in dynamically adjusting network traffic based on the different network elements and end applications. The existing LTE and 5G networks can easily coexist due to the spectrum sharing of frequencies by both standards.

Following is an example scenario of a combination of the technologies described above. In this diagram, the source will be a combination of multiple types of applications, services, and devices. These include thousands of machine learning-based IoT devices, different sensor nodes, autonomous vehicles, vehicles, augmentation, virtual reality-based devices, and eHealth devices, among others. The source device can be a cluster based on the data requirement and application pattern, then the edge-based intelligence can be applied for network slicing.

Figure 1: Intelligent EDGE-based 5G communication representation

The network slice can vary based on the number of devices connected at the source and its corresponding end application and service requirement. Real-time big data needs to be processed at the edge to create effective network functioning. Here, machine learning and deep neural algorithms help to identify the appropriate actions on the incoming data at edge and help to create the network slices based on intelligent analysis.

Real time big data needs to be processed at edge to create effective network functioning

In the future, edge will replace the intermediate and end-networking nodes. It will support intelligent 5G applications and services to reside at the edge and provide a better and faster response to the more energy-efficient, low latency-based devices, and associated applications.

Technologies keep evolving, so in the future, dynamically collaborating edge-cloud integration looks possible, without the remote monitoring and hardware requirements needed as of now. The next-age neural networks with intelligent networking measures like software-defined networking (SDN) will change the continuous need for remote infrastructure, and processing power efficiently. The combination of 5G, edge, and deep intelligence aided by machine learning provides a massive advantage for the next level of a super-connected world and its big-cloud network with massive IoT devices.