Abstract
Our world is increasingly shaped by digital systems, from vast industrial processing plants down to household devices. Industries today are undergoing a rapid transition towards the Fourth Industrial Revolution (Industry 4.0), characterized by real-time automation, machine-to-machine communication, and highly flexible production. Imagine a factory where production lines continuously adapt to change in production demands, robots coordinate tasks autonomously, and precision industrial machines respond instantly to sensor inputs. In such scenarios, reliability and latency matter, even minor delays can lead to significant losses or hazards to human operators and equipment. This presents the underlying network infrastructure with unprecedented challenges as the industry adopts larger automated- and interconnected systems.
Cyber-Physical System (CPS) are integration of computation (the cyber) with physical processes in a distributed and networked system. The computational part monitors and reacts to the physical world to provide continuous feedback to in real-time. Low-Latency Cyber-Physical System (LLCPS) denotes systems attached to highly dynamic processes where the response must occur within a short and well-defined timespan.
When designing, developing, and testing LLCPS, it is crucial to shield the behavior of one component from the rest of the system, and vice versa. This notion of composability must be applied to all levels of the system, including the network. Designing and verifying a LLCPS requires tools and methods to configure the network, implement distributed logic, and evaluate the real-time performance.
Time Sensitive Networking (TSN) is a set of IEEE standards that amends standard packet-switched Ethernet to become more deterministic and resilient to network faults. Through a multipronged approach, several aspects of ordinary Packet-Switched Network (PSN) non-determinism are addressed, from time accuracy to delivery guarantees. Although TSN greatly improves the reliability and usefulness of PSN in and industrial context, it also increases the complexity. For greenfield and legacy systems alike, lowering the initial threshold of deploying TSN is crucial for successful TSN adoption.
This thesis addresses the critical demand for ultra-reliable, low-latency wired networks that underpin CPS in industrial environments. By addressing the question of composability, it shows how deterministic networks are both a sufficient and necessary element to scale distributed systems as envisioned by Industry 4.0. The contributions describe how TSN is well suited for LLCPS.
A key element of this work is Composable Network Channels (NetChan), a library that provides deterministic network channels as an abstraction to TSN streams. Much of the complexity required to configure and operate a TSN-capable application is handled seamlessly by NetChan. The NetChan network primitives have been shown to be a powerful abstraction that gives a clear and concise way to express critical network traffic. It has evolved incrementally from a basic Proof of Concept (PoC) into a library capable of managing multiple, high-rate critical streams of low-latency control traffic. It allows for systematically capturing detailed data, enabling precise assessment and verification of distributed real-time systems’ real-time performance. This makes NetChan well suited for researchers and developers to test and validate ideas in the context of LLCPS, thus accelerating innovation and easing the transition towards Industry 4.0.