From Hardware-Independent Control to Real-Time Execution for Physical AI

Software-Centric Control Architecture for Physical AI

For decades, manufacturing equipment control systems have been built around dedicated hardware such as PLCs, motion controllers, and robot controllers. While each controller has reliably performed its designated functions, increasing system complexity has led to more controllers and wiring, as well as greater time and cost required to expand functionality or modify systems.

Movensys’ SDMC (Software-Defined Motion Control) is a control architecture that restructures motion control functions traditionally tied to dedicated hardware around a software-centric approach.

By separating control functions from specific hardware and executing real-time motion control in a general-purpose computing environment, SDMC aims to configure various equipment and control functions for specific purposes within a single software platform.

Hardware-Independent Control

In conventional control systems, equipment functionality and performance have often been determined by the specifications of specific controllers or motion boards. Changing or expanding control functions has typically required adding hardware or redesigning the control system itself.

SDMC reduces this hardware dependency by implementing real-time control functions in software. Instead of configuring equipment around a specific controller, users can select industrial PCs, networks, and servo systems that best fit their required performance and application.

Even when the hardware configuration changes, existing control applications and software assets can be broadly reused, while additional functions can be introduced through software modules and updates. This reduces the burden associated with equipment development and modification while enabling accumulated software assets to be leveraged across multiple types of equipment.

Open and Reusable Control Platform

A key value of software-centric control is that control functions are not locked into a single hardware configuration.

In SDMC, real-time motion control, industrial networking, robot applications, and data interfaces are organized into modular software layers. Each function operates on a common platform and can be combined or reused according to the requirements of the equipment and manufacturing process.

By leveraging general-purpose programming languages and open interfaces, SDMC can integrate not only with existing automation technologies but also with software ecosystems such as ROS 2, digital twins, and AI. This approach brings the development methodologies and scalability of IT into manufacturing environments while maintaining the real-time performance and stability required in OT environments.

As software-based control technologies continue to expand, the competitive advantage of industrial automation is also moving beyond the performance of individual controllers toward software reusability, interoperability, and more efficient deployment and management.

WMX as an SDMC Platform

WMX is a software-defined motion control platform that implements real-time motion control on general-purpose PCs, based on Movensys’ accumulated soft motion technology.

WMX uses a hard real-time control architecture on a general-purpose OS—an architecture designed to ensure that control computations are consistently completed within a defined time regardless of system load—to allocate part of the resources of a multicore CPU to real-time control. This enables high-speed, high-precision control without dedicated motion boards and provides a large-scale, multi-axis synchronized motion framework capable of synchronizing and controlling multiple axes and pieces of equipment from a single PC.

WMX also provides a multi-protocol Soft Master stack optimized for communications, enabling different industrial networks such as EtherCAT, CC-Link IE TSN, and MECHATROLINK-4 to operate within a unified control environment. By separating control modules from communication modules, control applications can be reused even when the network or hardware configuration changes, while hybrid networks combining multiple communication protocols can also be configured.

An open platform layer consisting of APIs, SDKs, and simulation tools further supports the extensibility of WMX. Equipment manufacturers can use general-purpose programming languages and APIs to develop the control functions and applications they need, while validating control logic and interfaces in a simulation environment before connecting to actual equipment.

SDMC is a broader concept that describes these technological strengths of WMX. It goes beyond precise control of individual motion axes to implement a wide range of movements for equipment and robots through software, while connecting different control environments into a unified workflow.

Because the combination of control software and hardware can be designed according to specific purposes, equipment manufacturers can apply accumulated software assets across a variety of equipment. The ability to configure required functions in modular and license-based units, and to continuously improve equipment functionality and performance through post-installation software updates, represents an important value of the WMX-based SDMC platform strategy.

Real-Time Workload Isolation

In PC-based control environments, various workloads—including motion control, robot control, vision, AI, data processing, and user applications—run simultaneously. However, each workload has different requirements in terms of response time and criticality. When real-time control and general-purpose processes share computing resources without separation, system load can affect the consistency and stability of control cycles.

WMX is based on a real-time operating system and assigns one or more cores of a multicore CPU to real-time control. This separates motion control workloads from the load and priority changes of general-purpose processes, creating an environment in which they can execute deterministically according to predefined control cycles.

In such mixed-criticality environments, where tasks with different timing constraints and levels of importance coexist, critical control workloads can use isolated execution resources, while non-real-time applications such as vision, data analytics, and AI can utilize the remaining computing resources.

The goal is not simply to integrate multiple functions into a single PC. Rather, it is to protect the execution environment required for real-time control while allowing different workloads to operate together. Equipment manufacturers can configure computing resources according to control-cycle and performance requirements on a common computing platform, reducing system complexity while adding new functionality.

Note: Mixed-criticality refers to an environment in which workloads with different levels of criticality and real-time requirements execute together on a single system.

Real-Time Execution Layer for Physical AI

Physical AI perceives the real-world environment and plans the movements required to interact with it. However, commands generated by AI or trajectories validated in simulation do not automatically translate into precise movements of a physical robot.

Real-world equipment is subject to various physical variables, including communication latency, friction, vibration, and changes in load. These differences can create a Sim-to-Real Gap between movements planned in simulation and actual physical behavior.

One of the key roles of SDMC is to connect, through software, the higher-level layers where AI and simulation operate with the real-time control layer of physical equipment. It delivers plans and decisions to the motion control layer as executable commands and returns equipment states and motion results to the higher-level system.

WMX R2 is a solution that applies this role of SDMC specifically to robotics and Physical AI. It connects ROS 2-based applications with WMX’s real-time motion control environment, executing tasks and trajectories generated by higher-level systems through deterministic multi-axis control. Joint states and control results collected from physical equipment are then fed back into the ROS 2 ecosystem.

WMX R2 is both an interface connecting ROS 2 and WMX and an example of SDMC in practice, integrating the planning and real-time control layers into a single software workflow while keeping them logically separated.

Through this bidirectional integration, movements planned by AI and simulation can be consistently realized in the physical world, while data collected from the field can be fed back into the decision-making process of higher-level systems.

Through real-time execution layers such as WMX R2, SDMC serves as a core control infrastructure for Physical AI, translating AI decisions into precise physical movements.

Toward Software-Defined Manufacturing

As AI and Physical AI continue to expand, manufacturing environments are evolving from systems that simply execute predefined control logic toward environments where data, software, and physical equipment interact dynamically.

SDMC enhances the reusability of control functions and interoperability between systems by organizing real-time motion control and equipment data around a software-centric architecture. It supports a continuous loop in which decisions made by AI and simulation are translated into physical motion, while data generated in the field is fed back into higher-level systems.

Movensys is expanding the role of SDMC based on the real-time motion control technologies it has developed through WMX.

A manufacturing environment where data becomes decisions, decisions become motion, and the results of motion become data again. This is the direction of Software-Defined Motion Control that Movensys is pursuing.

- Contributor: Hailey, Movensys Business Planning&Strategic