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

Mobile Augmented Reality (AR) applications have been observed to put high demands on resource-limited, portable devices, thus using up much power besides experiencing high latency. Thus, to overcome these challenges, the following AI-driven edge-assisted computation offloading framework that will provide optimal energy-efficiency and user experience is proposed. Our framework uses Reinforcement Learning/Deep Q-Networks for learning the optimal task offloading policies based network status, battery status, and the tasks' required processing time. Also, as a novel feature, we implement Adaptive Quality Scaling, which leaned from previous strategies managing AR rendering quality in relation to available energy and available computing capability. This one is known to make interaction possible for the handling of call flow to be efficient and at the same time, low energy consumption. Several experiments were conducted on the proposed framework and results show that there are an average of 30% energy saving compared to traditional heuristic-based methods of offloading, and the task success rates are above 90% while the latency is kept below 80 ms. These results support that our method proves to be efficient in improving AR task performance, enhancing battery endurance on the devices, and improving real-time user experience. In addition to this, the system proposed in this paper uses reinforcement learning to dynamically deploy offloading which enhances the resource allocation to be smart and timely. The research given here offers an approach towards ensuring that mobile AR is beneficial in achieving efficiency while addressing the needs of dynamic edge computing.