The IoT boom makes edge computing necessary, but data processing in the cloud is more dynamic and therefore scalable to any extent. A survey among industry experts shows that both technologies have their advantages and can complement each other perfectly.
Dr. Olaf Munkelt, Managing Director, MVTec Software GmbH: Cloud AND Edge Computing for Mutual Support
An industrial application can require high latency and reliability, but it can also require high computing power. As a result, you need to take advantage of both technologies, cloud and edge computing, with the right combination in the Industrial Internet of Things (IIoT). An example from industrial machine vision illustrates how both technologies can work together to promote their mutual strengths. For an industrial inspection task, an intelligent camera records an image of the product in the production line. The embedded processor performs tasks such as alignment, cutting, and preprocessing, while a specialized deep learning computer unit performs a classification from a pretrained network. The vision sensor sends the results to the programmable logic controller (PLC) and the images and telemetry data to a cloud server. The images are used to retrain the neural network overnight and to increase robustness while aggregating the telemetric data for use in predictive maintenance.
Boris Fiedler, Digital Leader of Robotics at ABB: Edge Computing for Data Analysis, Cloud Computing for Training
"Train in the cloud, act locally": this is how the interaction of edge and cloud can work together. A concrete example can be found in the painting sector: A new car is painted every 60 seconds in car body manufacturing. With expensive raw materials and a high energy requirement, painting is one of the most expensive processes in automotive production. To avoid costly reworking, it is important to apply the paint with the highest and consistent quality. ABB has developed the first digitally networked paint atomizer for this purpose. The groundbreaking aspect of the solution is the networking of the sensor-equipped atomizer with the digital ABB Ability platform, which enables intelligent diagnosis in real time and precise paint control. We use edge computing to analyze data during the painting process and to detect process anomalies such as air bubbles in the paint using machine learning algorithms. To develop and improve these algorithms, we upload some of the raw data into the cloud with the consent of our customers and train them thanks to the available storage and computing power.
Bernhard Lusch, Sales Manager CNC, Fanuc Deutschland GmbH: Cloud and Edge Computing as Partners
Cloud and edge computing are clearly partners in IIoT data analysis. This can also be seen in the fact that Fanuc's IoT platform Field System is edge-based, but also has interfaces to cloud systems. The Edge Analyzing Unit module, which can also be retrofitted to existing machines, records CNC and sensor data. The combination of these data and the comparison with target data can be used for preventive maintenance, for example. Comparable information can be obtained using the AI Servo Monitoring software option. Artificial intelligence compares the recorded data to a normality score and suggests appropriate maintenance actions when adjustable limits are exceeded.
Prof. Dr.-Ing. Markus Glück, Research & Development Managing Director, CINO, SCHUNK GmbH & Co. KG: Edge Computing as an Important and Collaborative Complement to Cloud Analytics
We see edge computing as an important and collaborative addition to cloud analytics for the benefit of users. Above all, time-critical data are processed directly in the smart gripper close to the location of the event. For example, quality characteristics of components can be checked during handling and IO/NOK decisions can be made directly in the gripper. The amount of data to be transferred is reduced to a minimum. On the other hand, computation-intensive tasks without real-time requirements take place in the cloud. Smart handling modules, such as the Schunk EGL parallel gripper, create the conditions for a full integration of production plants in the production environment and their connection to cloud-based ecosystems. Each individual process step is monitored in detail and forwarded, for example, to the system controller, to the higher-level ERP system, but also to analysis databases and cloud solutions. Consequently, smart grippers enable closed-loop quality control and direct monitoring of the production process in the production cycle.
Conclusion: Cloud Computing vs. Edge Computing – There Is No Either-Or
Edge computing is especially suited for applications with low latency, and even large amounts of data generated in high-frequency processes can be stored and evaluated close to the process. The cloud is the right place to train machine learning algorithms that will be used later in edge computing; it can be ideal for trend analysis or statistical analysis of process quality.
Both technologies have established themselves in the growing IoT market to manage the storage and processing of data volumes. They have long since ceased to be mutually exclusive, but instead work hand in hand.