A Heterogeneous Edge-Fog Environment Supporting Digital Twins for Remote Inspections.

Area of Science:

• Computer Vision
• Robotics
• Industrial IoT

Background:

• Digital twins offer advantages for equipment inspection and maintenance but present integration challenges.
• Remote inspection requires synchronizing diverse sensors (stereo vision, SLAM, thermal) and data fusion for comprehensive understanding.
• Existing methods struggle with computational costs and real-time data processing for complex digital models.

Purpose of the Study:

• To propose an optimized edge-fog-cloud architecture for real-time digital twin applications.
• To enhance the efficiency and scalability of remote inspection and maintenance processes.
• To address the challenges of data synchronization, integration, and fusion from multiple sensors.

Main Methods:

• Implemented a publisher-subscriber communication framework across edge, fog, and cloud nodes.
• Utilized edge nodes for initial data preprocessing from stereo and RGB-D cameras.
• Employed fog clusters for point cloud registration, odometry, and filtering, with cloud for final texturing and processing.

Main Results:

• Achieved optimized throughput and reduced computational costs through distributed processing.
• Demonstrated real-time 3D reconstruction with moving cameras, showing improved data acquisition to visualization time lag.
• Experimental results validated precision against ground truth and confirmed scalability with added sensors and algorithms.

Conclusions:

• The proposed edge-fog-cloud architecture effectively manages complex sensor data for digital twins.
• This approach significantly improves the efficiency and responsiveness of remote inspection and maintenance.
• The system is highly scalable, adaptable to new sensors and algorithms, paving the way for advanced industrial applications.