PRESS RELEASE: In September 2022, China Mobile and ZTE jointly completed the world’s first pilot project of the Cloud SPN computing-network convergence solution in Changzhou, Jiangsu. This innovative solution is mainly applicable to the scenario where the traditional quality inspection is upgraded to the AI-enabled one in industrial manufacturing. It allows the cloud-edge collaboration and the new algorithm model of AI intelligent inspection to help customers with the automatic quality inspection of production-line spray codes. It adds AI boards to the SPN product platform in the existing commercial network, loads AI algorithm models to upgrade current product capabilities, and offers hyper-edge ubiquitous computing services.

Prior to this, there were a variety of “pain points” for quality inspection in some industrial parks: lack of intelligent measures, reliance on manual inspection and massive consumption of human resources; costly dedicated devices deployed for automatic inspection, lots of inspection sample variations and inflexible customized algorithms for dedicated devices, which make it hard to upgrade the inspection system; low accuracy for partial automatic local inspection, real time tuning of system parameters and high service costs. The Cloud SPN computing-network convergence solution by China Mobile and ZTE makes great use of existing SPN capabilities to provide local lightweight computing resources for the park and consolidate edge computing, cloud computing and network capabilities into the edge-cloud collaboration solution. Applied to an industrial park where SPN devices are added with AI boards to deploy local applications rapidly, it extends the public cloud to the edge by virtue of low latency of the transport network to collect data in parks and process real-time services locally. The early demand analysis and preliminary test results are used to verify the methods of solving the three major problems:

  1. Low-latency metro transport networks supporting fine-granularity slicing private lines ensure secure data isolation in ToB application access to cloud, extend the public cloud to the park, and make the public cloud intelligence move down to the park.
  2. The production line in the park is deployed with transport access devices integrating machine vision chips and applications. Based on training samples, they can use AI models quickly for local real-time inspection and make user data kept in the factory and remain confidential.
  3. Wide-coverage transport access devices can integrate computing power for ubiquitous deployment. The cloud-edge collaboration architecture contributes to lightweight edge computing, and the intrinsic computing power of network devices remarkably cuts deployment and O&M costs of hyper-edge computing services.

For example, in a technology quality inspection scenario. a specific controller and an industrial camera work with the PLC (Programmable Logic Controller) for the original on-site quality inspection. Several problems emerge in the process: There is no service data association between quality inspection workstations on the production line, and a set of specific controllers and industrial cameras need to be deployed separately at each workstation. The controller is complex in quality inspection process or standard changes, which makes it hard to upgrade. The service response capability of the controller agent cannot keep pace with the requirement, and the service localization is weak. After the Cloud SPN computing-network convergence solution is launched for a pilot trial, the ZTE HEC (Hyper-Edge Computing) device (SPN access device 6180H integrating AI boards) is deployed on the production line to inspect the quality of spray codes at the eighth workstation and make the joint debugging test to address the requirements of the customer for spray codes. This is the first time to provide local intelligent computing services for industry customers on SPN devices. In the second phase, in-depth cooperation will continue. One HEC device can gradually take over the inspection tasks of seven other workstations on the production line. The advantages of the overall solution stand out:

  1. Data between procedures can be analyzed and managed in a centralized manner, and new services such as full-process fault prediction of the production line can be offered in the future.
  2. Even if a process or standard is changed at a workstation, software and algorithm model can be trained to put the new inspection procedure into use quickly, which makes it easy to upgrade.
  3. In addition to ZTE’s advantages in localization services, a model can be trained in the mobile cloud later through edge-cloud coordination, and an inference can be made on the local SPN device. The model can be delivered to the HEC device in the factory via the SPN MAN channel to load and update online.
  4. The Cloud SPN computing-network convergence solution reduces the cost more than the previous industrial park controller solution.

The solution upgrades the traditional quality inspection to the AI-enabled high-precision one. Leveraging a wide range of SPN devices invested in the existing networks of China Mobile, it generates a new model of edge-cloud collaboration for innovative applications, validates the feasibility of hyper-edge ubiquitous computing applications, and set a basis for the operator to boost ToB services as a result of scenario duplicability.