Software-Defined Product (SDP) is a methodology that utilizes software-based approaches to dynamically define, model, and adapt product architectures. The methodology aligns effectively with the demands of Industry 5.0, characterized by the need for flexibility, adaptability, and intelligent system control. As markets increasingly require rapid adaptation, SDP separates the core control functions of products from physical hardware, using modular digital twins in combination with advanced decision-making algorithms. The strategic setup significantly reduces dependence on physical prototypes, facilitating more efficient evaluation of diverse design scenarios. However, successfully achieving SDX through SDP fundamentally requires traditional products to be completely redesigned, clearly distinguishing between hardware elements and software-driven functionalities.

A core strength of SDP lies in the strategic use of modular digital twins, which enable efficient exploration of different design options. Complex products can be decomposed into separate functional modules that can be flexibly combined or rearranged without incurring significant additional costs. Additionally, modular platforms further enhance efficiency because the platforms provide standardized methods for reusing previously validated modules across different products. The platform also improve prediction accuracy and overall resource efficiency [12].

Furthermore, Generative AI (GenAI) offers an effective extension to SDP through innovative design options informed by existing data and market insights. GenAI models leverage patterns identified from internal datasets and external sources, such as industry trends, market feedback, and user-generated content [13]. Thus, companies can make various design strategies for a product aligned with market demands and stakeholder expectations. When integrated with modular digital twins, GenAI provides a powerful tool for enhancing the flexibility and market alignment of SDP-driven product development.

SDPr can be characterized as a paradigm shift in the design, monitoring, and optimization of manufacturing processes. Conventional manufacturing systems rely on fixed operational sequences with limited capacity for immediate adjustment. In contrast, SDPr introduces a dynamic framework that allows the reconfiguration of operational workflows, resource allocation, and production parameters as required [14]. This capacity for rapid adaptation to variations in product mix and volume is achieved by decoupling process definition and control from the physical machinery. Consequently, equipment may be reprogrammed or redeployed without requiring extensive hardware modifications.

In practical implementations, SDPr systems depend on the integration of data from sensors, machines, and other production assets into a centralized software platform. The software platform employs rigorous analytical methods and AI techniques to detect process bottlenecks, forecast potential downtime, and optimize scheduling [9]. Continuous real-time machine or product level feedback is used in two ways. The primary function of real-time feedback is first used to control or optimize the current manufacturing process. For instance, if real-time measurements indicate that a specific station is approaching capacity or that a robot shows signs of imminent failure, the SDPr system may reassign tasks or adjust operational parameters to prevent disruptions. The real-time feedback can also be stored in either on edge device or a cloud data lake that can later be used to train AI models for higher precision and accuracy. In this manner, SDPr transforms the manufacturing environment into a highly adaptive system with minimal impact on overall throughput and product quality [2].

SDF extends the principles of SDX to the entire manufacturing facility, orchestrating production activities across multiple lines, cells, or plants. Traditional factory systems are characterized by static layouts and predetermined production flows. In contrast, SDF facilitates the reconfiguration of plant layouts, the optimization of resource allocation, and the synchronization of production schedules based on software-driven decision making [10,11]. This comprehensive approach employs digital simulations and advanced optimization algorithms to evaluate various factory configurations under diverse demand scenarios and operational constraints [15].

Within SDF, real-time monitoring of machinery, robotic systems, and workforce availability is integrated with predictive analytics to reduce downtime and enhance equipment utilization [1]. Fig. 2 illustrates a digital twin of an example scenario where DRL is used to maximize throughput in cellular manufacturing by optimizing product scheduling, AMR routing, and machine repair prioritization. Moreover, standardized interfaces and communication protocols are implemented to ensure interoperability between heterogeneous systems, ranging from legacy controllers to contemporary IoT-enabled devices [5]. This robust connectivity enhances system resilience, as rapid reconfiguration helps maintain production continuity even in scenarios with supply chain disruptions, equipment failures, or labor constraints [16]. Thus, the SDF architecture provides a foundation for an intelligent and responsive manufacturing ecosystem that meets the stringent demands of Industry 5.0.

A Digital Twin is a real-time virtual replica of physical assets and processes, serving as a foundational component of the SDF. The digital twin continuously receives sensor and machine data from the factory floor, mirroring machines, robots, and processes in a dynamic model to allow real-time adjustments [17]. The digital twin transforms the physical factory into a digital domain that can be analyzed, programmed, and optimized with agility [18]. A typical digital twin in SDF includes IoT-connected physical devices such as machines, conveyors, robots, and sensors. A communication layer streams this data to the platform, which holds virtual models of equipment and processes synchronized with reality. An analytics layer runs AI/ML algorithms and simulations to generate insights [19]. Among the essential technologies for SDF are applications and dashboards that provide real-time 3D visualization and KPI tracking. Any sensor changes update the digital twin instantly, and optimized planning and scheduling from the twin flow back to the shop floor, forming a closed loop. The data-driven system monitors the entire factory, consolidates information into a single source of truth, and enables predictive maintenance. Through sensors and software, potential failures are anticipated and mitigated.

Simulation is another key function. Engineers can explore “what-if” scenarios such as the reconfiguration of production lines or schedule adjustments without any physical changes [20]. The simulation enables quick decisions; if a machine shows anomalies, the twin suggests maintenance. The digital twin also allows flexible line setups, shorter ramp-up times, and enhanced quality control, which helps maintain agility. Automakers rely on it to optimize assembly lines, and semiconductor manufacturers refine processes to boost yields, which establishes the twin as the digital core of SDF.

Digital manufacturing helps conventional factories into flexible, software-driven environments. Unlike traditional setups, it is capable of real-time adaptation to customer-specific needs through digital processes. Digital manufacturing is achieved via modular digital twins, AI-driven optimization, and automated production cells that adjust dynamically [21]. A key advantage of the digital manufacturing is its ability to customize products within a single production cell. Manufacturers can modify production parameters instantly for efficient mass customization. The digital framework integrates software-controlled fabrication techniques, reducing the need for complex reconfiguration. However, digital configuration alone is not sufficient for full implementation. The digital manufacturing can be realized through Additive Manufacturing (AM) or 3D printing, which bridges virtual designs with physical production. AM improves production responsiveness with real-time modifications, rapid prototyping, and on-demand manufacturing [22]. The digital manufacturing empowered by AM allows industries to shift from rigid setups to fully adaptive production models. The combination of digital control and AM ensures manufacturing systems are software-driven, and capable of executing complex, customized production strategies efficiently [23].

Robotics forms a crucial pillar in the transition toward SDF, as modern automation solutions rely on the adaptability and precision that robots provide [24]. Industrial robots, employed for tasks such as welding, painting, and material handling, are evolving into more flexible agents that can be rapidly reprogrammed to accommodate changing production requirements [8]. The flexibility is enhanced by collaborative robots (cobots), which are designed to work safely alongside human operators by sharing tasks that demand both dexterity and efficiency. By decoupling robotic control from fixed hardware configurations, manufacturers can deploy new product lines or variations without substantial retooling.

Mobile robots also play a crucial role in SDF environments. Automated Guided Vehicles (AGVs) and Autonomous Mobile Robots (AMRs) transport materials and subassemblies across the shop floor, reacting in real time to dynamic scheduling commands [1]. This is especially beneficial for just-in-time (JIT) production scenarios or when managing high product mix with smaller batch sizes. In addition, the integration of AM systems within a robotics-driven process flow allows for rapid prototyping, customized parts production, and on-demand manufacturing [14]. By centralizing the control logic in a software layer, these various robotic subsystems can be orchestrated with minimal manual intervention, leading to improved responsiveness and reduced downtime [16].

In addition to the industrial, mobile, and cobots, humanoid robots are emerging as another key asset in the SDF. With an anthropomorphic design that mirrors human dexterity and mobility, the humanoid robots are engineered to operate in spaces traditionally configured for human workers. Humanoids can perform complex tasks—such as detailed assembly, intricate inspection, and adaptive maintenance—in environments where traditional robots might struggle, with advanced sensor arrays, AI-driven decision-making, and flexible limb articulation [9]. Moreover, humanoids’ humanlike interaction capabilities facilitate more intuitive collaboration with human operators, easing the integration of legacy systems and fostering a more cohesive, agile production process.

AI highlights many of the adaptive capabilities inherent in SDF. Traditional automation systems rely on predefined rules or human intervention to adjust production parameters, but AI-driven methods introduce a level of autonomy and insight that allows for real-time, data-driven decision-making [7]. One of the most important applications is predictive maintenance, where machine learning algorithms analyze sensor data—such as vibration signals, temperature readings, or motor current profiles—to detect early signs of mechanical or electrical failures [2]. By identifying anomalies early, factories can schedule maintenance proactively, thereby reducing unexpected downtime and extending equipment lifespans.

Another key area where AI stands out is process optimization. Techniques ranging from classical operations research to advanced reinforcement learning can be used to dynamically allocate resources, balance production lines, and select optimal process parameters. For example, an AI system can detect when a particular product variant experiences a sudden increase in demand that requires more frequent setup changes on specific machines. Rather than waiting for human intervention, AI can autonomously redistribute workloads or adjust sequencing to maintain overall throughput. Additionally, AI-driven quality inspection—using computer vision and anomaly detection—enables in-situ monitoring of parts and assemblies, catching defects early in the process and reducing rework or scrap. The level of oversight and adaptability becomes integral to the continuous improvement ethos that defines an SDF environment.

Generative AI based on large language models (LLMs) and multimodal architecture provide a distinct dimension of adaptive capability within the SDF. LLMs integrate live telemetry, historical maintenance logs, and engineering manuals to generate fault diagnosis reports and recommend repair sequences [25]. Retrieval augmented generation pipelines connect sensor anomalies with relevant domain knowledge and facilitate root cause identification as well as maintenance schedule formulation [26]. Generative adversarial network-based data synthesis creates realistic failure signatures that expand sparse sensor datasets and increase prediction accuracy. LLM-based agents also support rapid construction of process and equipment models. The agents transform unstructured engineering text into executable simulation code and accelerate digital twin development and scenario evaluation. Agentic AI frameworks with reasoning engines evaluate alternative production schedules and set machine parameters in advance of demand shifts or disturbances [27]. Therefore, generative AI enhances predictive maintenance, process optimization, process modeling, and knowledge transfer across the factory ecosystem.

High-speed, low-latency communication networks are essential for achieving the real-time responsiveness envisioned by SDF architectures [3]. The transition from wired industrial networks to wireless technologies like 5G and emerging 6G standards facilitates connectivity among cyber-physical systems (CPS), including sensors, robots, and other smart manufacturing assets [11]. Advanced wireless infrastructure supports machine-type communications and factories to collect and process data from thousands of nodes simultaneously. The capability is especially valuable when implementing advanced AI and big data analytics, as these approaches require near-instantaneous feedback loops.

In parallel, the standardization and integration of communication protocols, such as OPC-UA, MQTT, and Kafka facilitate the interoperability of heterogeneous systems [5]. However, different messaging protocols used by different machines often generate a problem of miscommunication between systems. Hence, a unifying software layer can bridge the differences among machines or robots designed for various protocols to achieve consistent data exchange. The unified approach streamlines process control and supports centralized AI-driven orchestration, where decisions regarding scheduling, routing, or resource allocation can be delivered instantly to the manufacturing environment. Ultimately, robust communication infrastructures allow SDF to function as an integrated entity, rather than a collection of isolated production cells.

An emerging initiative in this area is Universal Automation, which is an independent non-profit association comprised of diverse manufacturing vendors [28]. Universal Automation promotes the separation of traditional manufacturing equipment’s hardware from its control software. In this approach, the hardware provides only a runtime environment while the control logic is standardized and unified through IEC 61499-based control program applications. The hardware-software decoupling breaks hardware vendor dependency and subsequently enhances the reusability of control logic, which is crucial for achieving a flexible, software-defined process in SDF.

Cloud computing offers the scalable computational resources and storage capacities of SDF, particularly when running complex simulations or big data analytics are required [25]. A hybrid cloud approach—combining public and private cloud resources—allows manufacturers to optimize both security and performance. Sensitive data, such as proprietary product designs or strategic production plans, can be stored and processed on private cloud servers to comply with data governance regulations and protect intellectual property. Meanwhile, less sensitive tasks, including large-scale simulations for factory layout optimization or machine learning model training, can leverage public cloud services to access virtually unlimited computational power.

In many cases, edge or fog computing layers further enhance the cloud by handling latency-sensitive tasks near the production environment [26]. The distributed architecture is particularly advantageous for real-time control loops, where even minor network delays could impact product quality or process stability. Data collected at the edge can be filtered and processed before being sent to the cloud for deeper analytics to reduce bandwidth requirements and ensure that only relevant information is transmitted. Insights derived from historical data in the cloud’s data lake feed back into real-time edge systems that dynamically adjust production parameters to form a closed feedback loop within this hybrid cloud ecosystem.

As manufacturing systems become more software-driven and interconnected, data security emerges as a critical challenge [10]. SDF relies heavily on real-time data exchange between machines, robots, and cloud-based services, which can be vulnerable to potential security risks susceptible to cyber threats. To mitigate these risks, strong encryption methods are employed to secure data both in transit and at rest. Furthermore, secure communication protocols and authentication mechanisms restrict network access and system configuration changes exclusively to authorized devices and users.

Beyond the technical measures, an effective security strategy also includes policies and governance frameworks that clearly define access rights, logging, and auditing procedures. For instance, engineers responsible for product design may have permissions to modify digital twin models, but not to alter factory layout algorithms. Similarly, role-based access control limits the ability to override specific parameters during critical production runs exclusively to select personnel or systems. By embedding security considerations into the design of SDPr and SDF architectures, manufacturers can minimize disruptions, protect intellectual property, and maintain customer trust.

In summary, these seven technologies—Digital Twins, Digital Manufacturing, Robotics, AI, Wireless Network and Data Protocol, Hybrid Cloud, and Data Security—collectively establish the foundation for an SDF environment. Table 1 presents a consolidated overview of the three SDX layers (SDP, SDPr, SDF), their key characteristics, the primary technologies required, and representative applications within smart manufacturing.