Smartphone Architecture for Edge-Centric IoT Analytics

Area of Science:

• Computer Science
• Electrical Engineering
• Data Science

Background:

• Traditional Internet of Things (IoT) architectures rely heavily on cloud/core for data analysis and storage.
• Edge computing is increasingly recommended for IoT solutions to enhance efficiency and reduce latency.
• Smartphones possess significant computational capabilities suitable for edge processing in IoT.

Purpose of the Study:

• To propose an edge-centric IoT architecture specifically designed for smartphones.
• To introduce a novel architecture supporting comprehensive machine learning model lifecycle management on the device.
• To develop a real-time model evaluation and updating mechanism for sustained edge performance.

Main Methods:

• Formulation of a novel smartphone IoT architecture (SMIoT).
• Implementation of functionalities for data capture, preprocessing, model deployment, evaluation, and updating.
• Development of a real-time scheme for model validation and adaptation to changing data patterns.

Main Results:

• The proposed SMIoT architecture effectively supports edge computing tasks on smartphones.
• The novel evaluation and updating scheme ensures real-time model validation at the network edge.
• The architecture demonstrates sustainability and reliability by automatically adjusting to IoT data subspace changes.

Conclusions:

• The SMIoT architecture offers a viable edge-centric approach for smartphone-based IoT applications.
• Real-time model management at the edge enhances the performance and adaptability of IoT systems.
• This research contributes to the development of more efficient and responsive IoT solutions.