Dynamic QoS/QoE-aware reliable service composition framework for edge intelligence.

Area of Science:

• Artificial Intelligence
• Edge Computing
• Internet of Things (IoT)

Background:

• Edge intelligence offers advantages over cloud computing but faces challenges integrating AI with resource-constrained edge devices.
• Existing solutions struggle with the efficient deployment of AI methods on edge hardware.
• Smart systems like automated driving and connected healthcare require robust edge AI solutions.

Purpose of the Study:

• To address the challenge of adopting AI methods on resource-constrained edge devices.
• To introduce the Edge devices as a Service (EdaaS) concept.
• To propose a quality of service (QoS) and quality of experience (QoE)-aware framework for dynamic and reliable AI subtask composition.

Main Methods:

• Development of a novel QoS and QoE-aware framework for AI subtask composition on edge devices.
• Introduction of the Edge devices as a Service (EdaaS) paradigm.
• Evaluation using three meta-heuristics: black widow optimization (BWO), particle swarm optimization (PSO), and simulated annealing (SA).

Main Results:

• The proposed framework demonstrates applicability for AI subtask composition in connected healthcare scenarios.
• Black widow optimization (BWO) significantly outperformed particle swarm optimization (PSO) and simulated annealing (SA).
• BWO showed 95% efficiency over PSO and 100% efficiency over SA, prevailing in most experiments.

Conclusions:

• The EdaaS concept and the proposed framework effectively manage AI adoption on resource-constrained edge devices.
• Black widow optimization is a highly efficient meta-heuristic for AI subtask composition in edge environments.
• The framework offers a reliable and dynamic approach to enhance AI capabilities in edge computing applications.