Scope and caution
Automation covers the controls, software, networks and robots that run equipment with limited human intervention. This review looks at themes that were commonly discussed around 2023, concentrating on how control systems and cells were being organized. As with any trends review, talk ran ahead of practice in many places, and no statistics are given here. This is general information, not engineering advice.
Edge computing
A frequently cited idea was edge computing: doing some data processing close to the machine instead of sending everything to a distant server. The reasoning is practical. Machines can produce large volumes of data, such as vibration readings sampled many times a second, and it is often more sensible to filter or summarize it locally, then pass on only what matters. Local processing can also keep a line running if a network link is lost.
Edge devices sit alongside, not in place of, the traditional controllers. Programmable logic controllers still handle the time-critical control of the machine, while edge hardware typically handles analysis, data collection and links to other software.
Flexible automation cells
Another theme was moving from long, fixed lines toward cells that can be reconfigured. A cell might combine a robot, a feeder, a vision check and a small conveyor around one task, and be redeployed as products change. The appeal was handling shorter runs and more variants. Supporting ideas included standardized interfaces, recipe-driven changeovers and simpler programming so that plant staff, not only integrators, could make adjustments.
The trade-offs are real. Flexible cells usually give up some peak speed compared with a dedicated line, and the engineering to keep them reliable across many product variants can be substantial. Collaborative robots were often part of this picture because they can be relocated more easily, though each application still needs a proper risk assessment.
Better industrial networking
As more devices became connected, network design drew more attention. Plants worked on how to arrange control networks, how to segment them, and how to move from older proprietary links to more standard Ethernet-based options. The goal is dependable communication between controllers, drives, sensors and higher-level software. Private 5G networks were one wireless option discussed for large sites or mobile equipment, although for many plants they stayed at the pilot or exploration stage, and wired links remained the usual choice where reliability matters most.
Cybersecurity as a design input
Connecting control systems makes security part of automation design. Industrial cybersecurity commonly includes separating control networks from office networks, limiting and logging remote access, keeping an inventory of devices and planning how to patch or isolate equipment that cannot be updated. Many older systems were built when isolation was assumed, so adding security later takes planning. It is better treated as a continuing practice than a product to buy once.
Data and the supervisory layer
Operators have long relied on supervisory screens to watch processes, and in this period there was growing interest in connecting that layer to wider data tools. Traditional SCADA systems and operator interface screens were being supplemented by dashboards and historian databases that let engineers compare shifts, lines or sites. The value depends on data quality: poorly labeled signals and inconsistent naming are common stumbling blocks, and cleaning them up often takes more effort than expected.
Early, cautious uses of AI
Machine learning appeared in automation discussions, mostly for specific tasks such as visual inspection or spotting unusual equipment behavior. Reports on AI automation challenges at the time tended to note the practical obstacles: needing enough good-quality data, explaining results to operators and validating that a model behaves well when conditions change. In most plants, rule-based control stayed in charge of the process, with AI used as an advisory layer, if at all.
Looking back
The thread that ties these themes together is a move toward layered architectures: dependable real-time control at the machine, local processing at the edge and wider analysis above, all connected through networks that must be secure. It is hard to say how fast plants moved, because progress depended on budgets, existing equipment and skills. In practice, many of the changes arrived as small upgrades to existing lines rather than as sweeping rebuilds.