FOR most of the industrial era, automation has relied on some form of logic. Ford’s conveyor assembly relied on a logical workflow called manu-mation (manual automation).
Then the Japanese semiconductor plants introduced what is known as islands of automation — separate automation systems that feed semi-finished products from one “island” to the next. Full integration followed when these islands were connected to each other by programmable automation, one system taking over from the other until a finished product was made.
Since automation is a series of systems engineered to perform predefined sequences with precision and repeatability, this approach has delivered safety, throughput and consistency, but it is constrained by its rigidity.
At the 2025 Automation Fair in Chicago, Rockwell Automation Chairman and CEO Blake Moret presented a detailed technical roadmap for what he called “industrial autonomy”— a decisive shift away from traditional programmable automation toward systems capable of real-time adaptation, prediction and decision-making without continuous human intervention.
Moret, in his keynote, framed this evolution not as an incremental upgrade but as a redesign of the fundamental architectures that underpin modern manufacturing. This is a step into what futurists call Industry 4.0, where smart machines are instructed rather than programmed to complete tasks.
This autonomy replaces fixed instruction sets with control systems that continually adjust their behavior based on environmental feedback, sensor-rich operational data, and predictive models trained on both real and simulated scenarios. Implementing this shift requires the integration of several maturing technologies into a unified operational stack.
“We are building a system that learns and improves permanently,” Moret emphasized.
Factory-Scale Emulation
The cornerstone of this architecture is a new level of virtual modeling capability built around dynamic digital twins. The Rockwell chief emphasized that these twins are no longer static design artifacts but continuously synchronized representations of physical systems, updated in real time and accurate enough to predict performance outcomes. Integrating its own Emulate3D software with Nvidia’s Omniverse APIs enables physics-based simulation engines that can model mechanical behavior, thermal characteristics, electrical loads, and multi-system interactions at high fidelity.
The Result is What Moret Termed “Factory-Scale Emulation.”
Instead of modeling isolated machines, engineers can now simulate complete production environments — multi-vendor equipment interactions, material flow dynamics, automated storage and retrieval systems, network communication patterns, and load-dependent behavior across entire process lines. This virtual environment becomes both a development platform and a validation tool.
Machine learning models can be trained against thousands of simulated operating conditions, including rare failure modes that are impractical or unsafe to test in the physical world. Proposed control changes can then be evaluated against predicted outcomes prior to deployment, reducing commissioning time and de-risking plant modifications.
Software-Defined Automation
Achieving autonomy also requires a decoupling of application logic from hardware — a concept Rockwell describes as Software-Defined Automation (SDA). In conventional architectures, control programs are tightly bound to specific PLC (Programmable Logic Controller) families, firmware versions, and I/O (input/output) mappings. SDA abstracts this relationship, allowing application logic to be developed, tested and executed independently of the physical controller platform.
This separation creates a more flexible lifecycle for deploying advanced algorithms. Machine learning models and updated control strategies can be pushed to runtime environments without requiring hardware replacement. Moret highlighted the integration of Nvidia’s Nemotron Nano, a compact language model, directly into Rockwell’s FactoryTalk Design Studio.
Embedding a small language model within engineering tools enables natural-language generation of control logic, accelerates development cycles, and reduces the manual coding burden for increasingly complex systems. This AI integration is crucial to industrial autonomy, with each component of the network “self-thinking.”
Mobile Sensing Platforms
Autonomous mobile robots (AMRs) like those from OTTO Motors that are used for material movement are being reconceived as data acquisition platforms that extend the reach of fixed sensors. Outfitted with higher-level sensors, these AMRs capture spatial datasets — wireless signal maps, thermal gradients, airflow characteristics, and particulate concentrations — that static infrastructure cannot easily measure.
As these robots navigate facilities during routine tasks, they generate comprehensive environmental datasets that feed into machine learning models and real-time optimization engines. This enhances situational awareness for autonomous control systems and fills observational gaps in complex, high-variability production environments.
Simplify the Complex
Industrial autonomy, in Moret’s view, is not about eliminating human expertise but amplifying it. He described autonomous systems as cognitive augmentation tools that handle high-frequency control loops, data-intensive monitoring, and anomaly detection while elevating strategic decisions to human operators. As he put it, the aim is to give workers “superpowers”—the ability to manage environments too complex to oversee unaided.
For autonomy to be viable, the components must function as a coherent system. Predictability, deterministic behavior, and verifiable safety remain non-negotiable. The challenge, and the opportunity, lies in building architectures that retain the reliability of traditional automation while adding the adaptive intelligence required for next-generation manufacturing.
“Our goal is to simplify the complex. We want to be able to use AI and bring everything together in as simple a form as possible,” Moret said. With the complexity of manufacturing now exceeding humans’ capability for direct supervision, requiring hundreds if not thousands of interacting variables, manual optimization has become increasingly impractical, and industrial autonomy the only practical solution.
Originally published on The Manila Times