Abstract
In the era of the Industrial Internet of Things (IIoT), the demand for high-precision, low-latency quality inspection is rapidly increasing. This paper presents a Software-Defined Edge Intelligent Controller (SD-EIC) architecture that integrates operational technology (OT) and information technology (IT) on a single platform. By utilizing virtualization via a hypervisor, the SD-EIC consolidates real-time motion control and AI-driven machine vision, reducing hardware silos and communication latency. The system employs semantic digital twins for flexible reconfiguration and a cloud-edge collaboration model for continuous self-learning. Qualitative analysis suggests significant improvements in resource efficiency and system adaptability for modern smart factories.
Introduction
The transition to "Smart Factories" necessitates advanced quality inspection systems. Traditional approaches often rely on manual labor or fragmented machine vision setups. These existing implementations suffer from three primary challenges: high cloud latency for real-time processing, hardware silos where PLCs and AI servers operate independently, and a lack of flexibility when switching production lines. To address these issues, this research proposes the SD-EIC, a unified framework designed to bridge the gap between high-level AI inference and low-level industrial control.
The SD-EIC Architecture
The core of the proposed scheme is the SD-EIC, which utilizes a heterogeneous hardware architecture (CPU, NPU, and FPGA) managed by a hypervisor layer. This virtualization allows the hardware to be partitioned into distinct functional units.
As shown in the architecture diagram, Virtual Controllers run Real-Time Operating Systems (RTOS) to manage high-precision motion tasks, such as controlling robot arms. Simultaneously, Virtual Edge Computing Nodes run Linux-based environments to execute AI models for defect recognition. This consolidation eliminates the network delays typically associated with inter-device communication by using shared memory between virtual machines on the same chip.
System Deployment and Cloud-Edge Collaboration
The SD-EIC serves as the central hub for multiple workstations, facilitating a seamless flow of data from terminal devices to the industrial cloud.
The deployment model leverages semantic digital twins using the Resource Description Framework (RDF) to model terminal devices. This allows the system to dynamically reconfigure parameters via software when a new product type enters the line. Furthermore, a cloud-edge collaboration loop ensures that while inference happens at the edge for speed, model retraining occurs in the cloud using new defect data collected from the production line, creating a continuous self-learning cycle.
Key Insights and Results
The SD-EIC architecture effectively eliminates the need for separate hardware for AI and motion control. This IT/OT convergence transforms industrial logic into flexible software containers. Key results and advantages include:
- Resource Efficiency: A single SD-EIC can replace multiple dedicated hardware controllers, significantly reducing the bill of materials for factory automation.
- Multi-Station Support: The platform can manage different inspection tasks across parallel virtual environments simultaneously.
- Reduced Complexity: The architecture simplifies the terminal-edge-cloud hierarchy, making it easier for factory engineers to maintain the system without needing deep expertise in both AI and PLC programming.
Conclusion and Future Work
The proposed SD-EIC framework offers a robust solution for flexible, software-defined manufacturing. While future work is needed to quantify hardware scaling limits and minimize hypervisor overhead for ultra-hard real-time tasks, the current scheme provides a clear path toward more integrated and adaptable industrial quality inspection. Future extensions may also include non-visual sensing modalities such as acoustic and infrared sensors for more comprehensive quality checks.
Summary updated