ukANIARA

Connectivity between devices in a manufacturing chain is key to enabling Industry 4.0 and Smart Manufacturing by closing feedback loops and increasing flexibility and adaptability of automation cells. In a contemporary factory, the connections between sensors, robots, end-effectors, PLCs, CNC machines and other devices are made via cables operating Fieldbus protocols. Achieving the same with a wireless technology opens innumerable possibilities by connecting more devices, faster, and in a more flexible manner, effectively giving manufacturers greater control over their connectivity.

5G promises a number of useful properties for industrial communications in manufacturing. The first is often referred to as Ultra-Reliable Low Latency Communications (URLLC). Ultra-reliable means that the probability of a message not reaching its target is between one in a million and one in a billion (10-6 -- 10-9). Low latency refers to millisecond latencies, ensuring that the time between a message being sent and actually entering the network is five times lower than 4G and WiFi. 5G also allows massive connectivity, supporting one million connected devices per km2. Together these properties put 5G on a par with contemporary industrial communication protocols.

To actually achieve these properties in a network requires advanced mechanisms and algorithms to make complex, real-time decisions about the orchestration and management of the network resources and nodes. The focus of the UK cluster of the European ANIARA project is to develop 5G edge and distributed solutions for intelligent control, monitoring and performance enhancement of industrial manufacturing assets, process flows and in-factory product optimisation. This entails creating a dedicated network slice, or a private network, for a manufacturing facility where network and radio resources can be optimised to provide the required reliability, low latencies and predictable connectivity for industrial operation. ukANIARA will develop edge and semi-distributed AI techniques, most notably data-driven deep neural networks, combined with traditional model-based approaches to handle real-time resource management and orchestration.

The network itself will be supported by a flexible, cloud-native, micro-service architecture with defined application programming interfaces (APIs) to facilitate orchestration and ensure scalability and modularity for the addition of new industrial applications. An edge-cloud architecture will support dynamic management of underlying resources to provide the single-digit millisecond latencies and ultra-high reliability required of 5G for factory scenarios.

The project will demonstrate the possibilities of 5G in manufacturing on controlled-industrial site networks and results will be disseminated in high-calibre industrial and academic events.