What is Edge Computing and Edge AI?

Introduction

 
In this article, I will talk about IoT Edge Computing and why we are so into Edge Computing, and also why it is expected to replace cloud computing shortly. And also I will provide a sneak-peek into Edge AI.
 

Edge Computing

 
'Edge' implies processing technology near to the data source, i.e. data will be processed at the node from where data has originated, unlike client-server architecture, where data is processed at the server and then returned to the client. This is the centralized method for the transmission of data as close to the origins of the data as possible. This technology allows for effective usage of services that cannot be linked constantly to a network such as computers, smartphones.
 
Edge Computing encompasses the broadest spectrum of infrastructure, including cellular sensor networks, ad hoc, mutual peer-to-peer, decentralized cloud/fog, virtual edge computing, data storage, and recovery, autonomic self-healing networks, distributed cloud applications, augmented reality and much more.
 

Edge Computing vs Cloud Computing

 
The edge implies local (or near-local) service rather than going somewhere in the network. It would be an individual physical computer such as an autonomous refrigerator or server positioned as close the source as possible (i.e. servers situated in the area rather than the opposite end of the globe).
 
When low latency is required or the network itself might not always be accessible, the edge can be utilized. This may be accompanied by a need to make choices in other systems in real-time.
 
Most cloud apps receive data locally, transfer the data to the cloud, process, and return it. The edge implies that the cloud needs not to be delivered, therefore it is better and less impactful (depending on the reliability of the edge device). Edge AI algorithms in the cloud can still be equipped but operated at the edge.
 
AI Application Edge or Cloud Computing
Voice Assistants  Cloud
Self-Driving Cars  Edge 
Insights from Millions of Sales Transactions Cloud
Remote Nature Camera Edge
 

Why Edge Computing?

  1. Connection with the Network can be costly (bandwidth, energy use, etc.) and often challenging (think about distant areas or natural disasters)
  2. For systems such as self-driving vehicles, real-time computation is needed that can not accommodate delay for critical decisions
  3. Edge applications can use data that is vulnerable when transmitted to a cloud, such as health data
  4. Specially built for different devices, optimization applications may help achieve great performance with edge models

Why Not Edge Computing?

  1. The attack vectors will increase. Bringing more 'intelligent' systems into the mix, for example, edge servers and IoT networks with stable, embedded processors, offers malicious actors additional ways to exploit such machines.
  2. Further local equipment is required. For instance, if an IoT camera requires an embedded machine of transmitting its raw data to a web server, it will require a much more complex, computing efficient machine to operate its movement detection algorithms. 

Real-Life Applications of Edge Computing

  1. Self Driving Cars
  2. Detecting Irregular Heartbeats
  3. Robots Used in Medical Surgeries
  4. Tracking Animals which are in danger of Poaching
  5. Predicting how would a person react based on the words he or she types
  6. The reaction in case of a medical condition by the robot assistant
  7. Firewall's action when a threat is detected
  8. Checking if a person is answering the question rightly or not 

Key Terms

  1. Edge
    This depends heavily on the cases in usage. It may be a mobile phone or a cell tower, like in telecommunications. Similarly, it may be a car in the automotive case. It may be a computer in engineering and a laptop in information technology.
     
  2. Edge Devices 
    An edge device is one that collects and delivers data like machines and sensors
     
  3. Edge Gateway 
    It is a buffer that handles edge computing. The gateway is the opening in the field outside the network boundary.
     
  4. Fat Client 
    A program that stores data on edge computers opposite a thin client, who moves data just marginally.
     
  5. Edge Computing Equipment
    Devices such as sensors and computers may be fitted with Internet connectivity to operate in edge computing environments.
     
  6. Mobile Edge Computing
    In communications networks including 5 G situations, this signals the rise of modern computer technologies.

Types of Edge Computing

  1. Device Edge            
    Edge Computing is included in this paradigm of the existing environments of consumers. For instance, Microsoft Azure IoT Edge and AWS Greengrass. To read more about Microsoft Azure IoT Edge, visit
     
  2. Cloud Edge 
    This Edge Computing platform is a public cloud extension. Content Distribution Networks are classic examples of this topology that caches and delivers the static contents across a location at the geographical borders.

Edge AI

 
With Edge AI, AI algorithms are locally processed without contact on the hardware computer. It uses data generated from the system and processes it in less than a few milliseconds for real-time insights.
 
Example
 
By design, a portable power tool is on the network edge. The Edge AI program framework operates in real-time from the power tool data on a microprocessor. In the Power Tool, the program Edge AI produces data and stores the data on the computer locally. The technology machine links to the Internet during operating hours and only transfers the data to the cloud for retrieval and analysis. Long battery life is among the main characteristics of the illustration above. The battery will be exhausted in no time if the power machine continuously downloaded data into the cloud.
 

Advantages of Edge AI

  1. Reduced Cost 
    Edge AI can reduce the efficiency of connectivity and network transmission with fewer networks being sent. Costs of AI development in the cloud are often significantly more expensive owing to the cost of AI hardware.
     
  2. Security 
    Information becomes relevant for users by utilizing AI in situations such as surveillance cameras, individual vehicles, aircraft, etc. Since Edge AI manages data locally, streaming will prevent the issue without storing a ton of cloud data that renders your privacy vulnerable. Edge AI is very easy to process with few milliseconds, which significantly decreases the likelihood that data may be corrupted during the transit. Also, improved protection apps may be used to render Edge AI systems secure.
     
  3. Highly Responsive
    Edge AI systems can handle data very quickly compared to centralized IoT models, as you already know. These require real-time operations such as data development, decision-making, and intervention, as observations are transmitted directly inside the same hardware and ideal for applications in which milliseconds matter like self-driving cars.
     
  4. Easy to Manage 
    Edge AI systems are self-contained and do not need to be managed by data scientists or AI developers. Information and observations are distributed either automatically or available on-site through highly visual interfaces or dashboards.

Real-World Applications of Edge AI

  1. Surveillance and Monitoring Systems
  2. Autonomous Vehicles
  3. Smart Speakers or Smart Assistants
  4. Automating an Industrial Factory

Commercial IoT Devices using Edge AI

  1. NVIDIA® Jetson Nano™ Developer Kit
     
    jetson
     
  2. Sipeed Maixduino Kit for RISC-V AI + IoT
     
    maixduino
     
  3. Sipeed MAix BiT for RISC-V AI+IoT
     
    maix
     
  4. Raspberry Pi 4 Computer Model B
     
    raspberrypi
     
  5. Coral Dev Board
     
    coral
     

Conclusion

 
In this article, we had a look at Edge computing and also we tried to see why we can bet on Edge Computing in both IoT and Artificial Intelligence.



C# Corner
MVP Program Director