With the Edge getting rapidly digitized, data acquisition is becoming distributed and processing is becoming disaggregated. Let’s reflect on 2022 realities where Edge Tech is bringing value to industries and changing the way they operate.
Edge Computing is about bringing processing capability close to where data is produced because it is often not feasible to stream data from source points to clouds due to bandwidth constraints. Because edge computing transfers AI-based decision-making ability near to where sensors generate data, it is triggering amazing innovation across industries.
Any device with compute and connectivity can be an edge node. It can range from telco towers and trains to factories and farms. For example, in-car edge software brings your location, smart phone, street maps and fuel aggregators together to offer a driving experience that was not possible just a few years back. Edge processing is revolutionizing rural farming, bringing the ability to realize precision irrigation and spot fertilizer delivery. Edge intelligence is helping mining companies optimize energy usage and reduce water consumption. And edge analytics is enabling airports to improve operational efficiency by predicting counter allocations in real time based on passenger arrival patterns.
However, applications that bring intelligence to the edge have to be carefully designed to constraints that are often alien to traditional software. First, the deployment architecture is hybrid, partitioned across a vast plurality of edge nodes and one or more central sites. Second, edge compute may comprise of thin but specialized hardware, for example, radio units at the telco edge and IoT processors at the industrial edge. Third, far-edge locations usually see limited quality of Internet connectivity. And fourth, security challenges abound in distributed edge deployments because of a profusion of entry points into the network; so the “zero trust principle” that is commonly adopted by enterprise applications becomes harder to conform to at the edge.
Automation and orchestration are technologies on which edge deployments have critical dependencies on. Because the edge implies massive scale, the concept of “zero touch” deployments, sometimes referred to as “extreme automation” assumes importance. When there are 10s of 1000s of containerized edge nodes, for example, in the case of the 5G Radio Access Network (RAN), there are several layers of automation needs at play – deployment orchestration of a mammoth amount of clusters, the ability to remotely monitor, health-check, update and upgrade edge nodes from a central dashboard, and virtual engineers who attempt to self-heal edge site outages.
Computer Vision has become a key sensor for federating AI-based decision-making ability to the edge. Analysis of live video feeds have to be run near the points of capture since vast quantities of data cannot be streamed in real time to Machine Learning (ML) models residing on clouds due to bandwidth limitations. Instead, only the insights discerned at the edge are dispatched to clouds.
A vision-based edge device typically has multiple stereo cameras and special processing units that can run neural networks. It is programmed to arrive at decisions by running ML inferencing algorithms on the captured video feed on the fly. The ensuing system can perform semantic segmentation of various kinds of objects in the camera’s field of vision, extract features of interest, and track objects in 3D with spatial analysis. This cognition is then transmitted to the cloud, where the cross-edge business intelligence (BI) is elicited and visualized.
Applications for such computer vision based edge systems abound, for example, detecting illegal trafficking by monitoring luggage on airport conveyer belts, evaluating the safety of manufacturing shopfloors by tracking the presence and location of dangerous discarded objects, automatically shelfing inventory in warehouses using drones, and estimating queue wait times in retail stores or at highway toll booths.
Edge Tech is thus infusing value that is changing the way that industries operate – through Industry 4.0 and “digital twins” of machines, AgriTech, RetailTech, Construction 2.0 and Connected Vehicles. And Edge intelligence is the enabler for 5G cellular networks. We have now reached a point where Edge Tech has started to accelerate economic growth and improve the quality of life of ordinary citizens!
Written by
Sreekrishnan Venkateswaran, CTO, Kyndryl India
