Xentara Brings Physical AI Closer to Industrial Control
Industrial automation is entering a new phase as artificial intelligence moves closer to machines and control systems. Xentara, developed by embedded ocean GmbH, aims to support this transition through a runtime platform for physical AI. The company recently briefed ARC analysts on its approach to next-generation factory automation.
Unlike conventional industrial AI applications, physical AI connects intelligence with real-world machine behavior. Systems must interpret sensor data, make decisions, and execute actions within defined timing constraints. Therefore, industrial AI needs more than powerful models. It also requires real-time execution, secure connectivity, and consistent access to machine-level data.
Why Industrial AI Needs a Dedicated Runtime Layer
Many manufacturers already use machine vision, digital twins, simulation tools, historians, and cloud-based analytics. However, these applications often operate separately from deterministic control systems. Legacy architectures, proprietary technologies, and complex integrations can prevent AI applications from influencing machine operations directly.
Xentara addresses this challenge by combining machine connectivity, real-time control, and edge AI within one runtime environment. This approach targets the operational gap between AI-generated insights and physical machine actions. Moreover, it aligns with the broader shift toward software-defined industrial automation, where software increasingly determines how machines process data and execute control tasks.
Xentara Combines Edge AI, PLC Connectivity, and Real-Time Computing
Xentara runs on commercially available x86, AMD64, and ARM hardware. Its architecture brings together semantic data models, timing management, security mechanisms, and industrial communication interfaces. This combination aims to provide a consistent execution environment for machine data, control logic, and AI inference.
The platform supports technologies across several layers of industrial automation. These include Linux with PREEMPT_RT, OPC UA, Time-Sensitive Networking (TSN), WebSockets, MQTT, REST interfaces, EtherCAT, PROFINET, and Modbus. It also supports Siemens S7 and Beckhoff ADS connectivity, alongside database integration and microservices.
In addition, Xentara supports the Functional Mock-up Interface (FMI), Functional Mock-up Units (FMUs), C++, and ONNX-based machine learning inference. These capabilities can help developers integrate simulation models, existing PLC systems, and AI applications within a common runtime.
Nevertheless, interface availability alone does not guarantee deterministic performance or seamless interoperability. Engineers must validate timing behavior, supported devices, network configurations, and application requirements during system integration.
Brownfield Modernization Offers a Practical Entry Point
Replacing an entire industrial control system can introduce substantial engineering costs and operational risks. Many OEMs already have proven machine mechanics, HMI designs, application logic, and process expertise. Their existing equipment may depend on obsolete real-time kernels, aging motion cards, or restrictive communication architectures.
Xentara presents an alternative approach to this modernization challenge. Machine builders can explore replacing outdated runtime components while retaining valuable application knowledge and established machine behavior. The proposed architecture combines Linux with PREEMPT_RT and EtherCAT-based motion communication.
This strategy could help OEMs extend the useful life of existing equipment while adopting more flexible software architectures. However, migration requires careful validation of control-cycle timing, motion performance, device compatibility, functional safety boundaries, and recovery behavior. Manufacturers must also assess how the new runtime interacts with existing PLCs, drives, HMIs, and supervisory control systems.
EtherCAT Drive Communication Expands Machine-Level Data Access
One technical opportunity involves replacing conventional analog drive commands with digital fieldbus communication. Analog setpoints can control motion effectively, but they provide limited visibility into internal drive states. Engineers often need separate interfaces to access detailed operating parameters and diagnostic information.
EtherCAT can provide richer process data within cyclic communication. Depending on the drive and its configuration, this data may include position, torque, following error, operating parameters, and diagnostic status. Consequently, the drive can contribute more information to the wider machine control and monitoring architecture.
This additional visibility creates opportunities for condition monitoring, quality assurance, and process optimization. For example, engineers could correlate drive behavior with production quality or identify deviations during repetitive motion sequences. These applications still require suitable sensors, validated data models, and application-specific analysis.
Moreover, access to drive data does not automatically produce actionable AI insights. Developers must establish meaningful process relationships, define appropriate thresholds, and validate algorithms against actual machine behavior. Even so, richer drive communication can provide a stronger foundation for data-driven factory automation.
Software-Defined Automation Connects Control Systems with Physical AI
Xentara positions itself within the emerging software-defined automation market rather than as a conventional PLC or DCS replacement. Its approach focuses on creating a runtime layer that connects industrial communication, deterministic execution, and AI applications.
This distinction matters because manufacturers rarely replace every control component simultaneously. Instead, many modernize selected machines, introduce edge computing, and connect previously isolated data sources. A flexible runtime could support these incremental changes while preserving established control architectures.
However, industrial adoption depends on more than architectural concepts. OEMs and end users need documented performance, proven compatibility, cybersecurity controls, long-term support, and successful deployment references. They must also understand how the runtime handles system failures, software updates, and interactions with existing control infrastructure.
What Xentara Could Mean for the Future of Factory Automation
Physical AI could expand the role of industrial software from monitoring and analysis toward direct operational decision-making. To achieve that transition, platforms must combine machine-level connectivity with predictable execution and appropriate security controls.
Xentara offers one approach to this challenge by integrating real-time computing, industrial protocols, and edge AI capabilities. Its focus on brownfield modernization may also appeal to machine builders seeking alternatives to aging runtime technologies.
Ultimately, the platform's market position will depend on real-world performance and deployment results. If Xentara demonstrates consistent timing, practical integration, and measurable benefits, it could become a useful runtime option for OEMs and manufacturers.
For industrial automation teams, the broader lesson is clear: successful physical AI requires more than machine learning models. It also depends on the control infrastructure, data access, and execution environment that connect digital intelligence with physical operations.