Edge assisted energy optimization for mobile AR applications for enhanced battery life and performance

Area of Science:

• Computer Science
• Artificial Intelligence
• Mobile Computing

Background:

• Mobile Augmented Reality (AR) applications demand significant device resources, leading to high power consumption and latency.
• Resource-limited portable devices struggle to meet the computational demands of AR, impacting user experience and battery life.

Purpose of the Study:

• To propose an AI-driven edge-assisted computation offloading framework for mobile AR to enhance energy efficiency and user experience.
• To address the challenges of high power consumption and latency in mobile AR applications.

Main Methods:

• Utilized Reinforcement Learning, specifically Deep Q-Networks, for optimal task offloading policy learning based on network status, battery status, and task processing time.
• Implemented Adaptive Quality Scaling to dynamically adjust AR rendering quality based on available energy and computing capabilities.
• Developed a framework for intelligent and timely resource allocation through dynamic offloading deployment.

Main Results:

• Achieved an average of 30% energy saving compared to traditional heuristic-based offloading methods.
• Maintained task success rates above 90% with latency below 80 ms.
• Demonstrated significant improvements in AR task performance, battery endurance, and real-time user experience.

Conclusions:

• The proposed AI-driven framework effectively enhances mobile AR performance by optimizing energy efficiency and reducing latency.
• Reinforcement learning enables dynamic offloading and smart resource allocation, crucial for edge computing environments.
• The research provides a viable approach for efficient and dynamic mobile AR experiences in edge computing settings.