Introduction to 5G Enabled Digital Services

Published on 1 March 2023 by Johan Krebbers, Head Emerging Technology Strategy

Introduction

 

5G will power a new generation of Enterprise Digital Services. Digital Nasional Berhad (DNB) is working to catalyse the local ecosystem by identifying promising enterprise solutions which have the potential to be augmented by 5G.

DNB is mapping out key challenges in various industries to see how they could be solved with 5G-enabled Enterprise Digital Services. These end-to-end services address notable pain points across a growing number of industries. Mobile Network Operators (MNOs) will play a critical role in bringing these services to the business community at large.

In this article, we seek to explain the “5G Enterprise Digital Services” as we see it, identify portions of the value chain that we be filled by the business community and MNOs at large, as well as the portions for which DNB is responsible.

   

5G Enterprise Value Chain Visualised

Step 1: Devices

 

Any Digital Service starts with Data and therefore you need to first determine what data is needed to be collected. Sensors are used to collect this data, note that there is broad array of sensors for many different purposes and many different data types:

Keep in Mind:

 
  • Cost of sensors: There are many suppliers and options for sensors, and thus significant price competitiveness. As a business that might be looking to retrofit your operations with sensor technology, it is advisable to shop around for the best possible price points.
  • Be well aware that 5G allows for the use of far greater volume of sensors (and therefore much more data collected) than 4G. Businesses must be mindful of this at the design phase to avoid bottlenecks in the data loading part.
  • You might have heard the term ‘’sensor dust’’, which means that sensors get that so small (nano) and low costs that it is possible to have millions of sensors. Indeed, this appears to be the  of innovation in the sensor space.
 

Step 2: Connectivity

 

Once a company has settled on the type(s) of sensors it wants to use in Step 1, the next key consideration is connectivity. This includes connectivity as between the sensors, as well as connectivity of these sensors to Step 3 below.

There are multiple interfaces to consider for sensors, including LoRa ( LoRa PHY | Semtech ), BLE (BlueTooth), LTE, 4G, 5G, etc, etc. As the Company responsible for Malaysia’s 5G rollout, while we may suggest adopting a 5G-enabled solution for Step 2, this is not a given. In deciding this, consider first the extent of your company’s data collection needs.

Businesses must understand their connectivity needs. Does your company have large amounts of data (enhanced mobile broadband connectivity, or eMBB) and / or where real-time responses and feedback (ultra-reliable low latency communication, or URLLC) is crucial? If so, then a 5G-enabled connectivity solution is necessary. 

If for example, your company requires a new batch of cameras that will be taking video / images for which analysis will be done in near-real time, it’s likely that native 5G interfaces will be necessary. The “real time analysis” element of the operation will likely have relatively higher bandwidth demands, since it will likely have to stream the high-definition audio and visuals back to the human operator and / or machine visual engine for analysis.

If your business does not require near instantaneous processing or analytics in key processes, then a LoRaWAN interface should suffice.

Importance of Quality of Data:

 
  • The quality of data has a significant influence on the AI model's outcomes. The better the data captured, the more effective the AI analytics model becomes. It's rare for businesses to begin with faultless data capture and analytics capabilities. However, it's important to remember that data capture is an iterative process, and making small, steady enhancements over time can lead to improved capabilities.
 

Step 3: Edge Compute

 

We use sensors to collect data, and when decisions require real-time processing, we store the data on Edge devices, which have computing power close to the sensors. If edge computing capabilities are necessary, we recommend using 5G connectivity as the interface because of its low latency. In such cases, AI elements such as Machine Learning or Machine Vision also run on the Edge device. We will delve into Edge devices a separate blog since that is a topic on its own.

In all cases, the AI process runs on the Edge device and provides advice to the Decision Maker for decision-making (Step 6) or takes the decision autonomously.

 

Step 4: Cloud Compute & Storage

 

Data collected by sensors are best stored in a cloud-based data repository to ensure that it is immediately accessible to data owners and users.

Why is data ownership important? While conventional wisdom suggests data collected is owned by the company that generates it, AI routines used by a business are typically third-party IPs that belong to other suppliers. These AI routines run on a Platform-as-a-Service (PaaS)/Software-as-a-Service (SaaS) basis. Once the proprietary data is shared beyond the data owner's platform, there is a risk of diluting its value.

To ensure data ownership and accessibility, it is advisable to use a cloud-based data storage system that is under the control of the data owners. Third-party AI routines can request access to the data from this central data store, which serves as the master copy. This approach ensures that the data owners have complete visibility and control over external requests for access, while also maintaining the security and integrity of the data.

 

Step 5: Application & Services

 

Now that key data collectors (sensors) and storage systems (cloud storage) are in place, the next obvious step for a business is to sweat these proprietary data sets for insights. This entails running these data sets through some combination of:

 
  1. Third party (PaaS / SaaS) AI-based analytics software;
  2. In-house AI analytics;
  3. Human operators with the ability to gather insights.

AI can encompass various modalities, such as:

 
  • Machine Learning (ML): Sensor data (time series, etc.) can be used to predict future failure.
  • Machine Vision (MV): Image data (from cameras) can be used to locate problems with the images such as causes of illness; Quality issues of products; etc.

Look as well at AI tooling (assisting in developing AI models) developments to assist such as AutoML.

Workflows are Fully Automated:

 
  • In most cases these are fully automated workflows where (sensor) data is collected, sent to the data store, extracted by AI, executed by AI and decisions being made either autonomously or passed on to a decision maker for action.

Role of SME and Data Scientist:

 
  • These AI routines can be developed in-house or procured on the market as PaaS or SaaS services. To be able to do this in house you need access to both Data Scientists and Subject Matter Experts (SMEs) since both are needed to get to an AI model. Often the SME comes from the company owning the business process.
 

Step 6: Inform Decision Makers

 

Once the AI routines have been executed, we can inform the decisions makers (Step 6) in multiple ways such as:

 
  • An simple e-mail or message.
  • Speech interface.
  • (Speech) + Bots interface.
  • XR (Virtual and / or Augmented Reality): You can expect that that XR gets more and more important (Enterprise and Consumer) with new form factors coming to the market such as from Apple, Google, Meta, etc.. Thus, it's crucial to have access to staff who can develop these user interfaces.
  • Etc.
 

Step 7: Support Decision Making

 

Here, the actual decision will be made. Of course, in cases of autonomous activities the decision will be made at Step 3 (time critical) or at Step 5.