Improving User Experience via Reinforcement Learning-Based Resource Management on Mobile Devices

Mobile devices are required to provide a good user quality of experience (QoE) by ensuring a high performance while minimizing energy consumption. Many efforts have been made to strike a better balance between performance and power usage to achieve this goal. However, little attention has been paid to the potential impact of background applications on QoE. This study found that resource contention between background and foreground applications on mobile devices can lead to a notable decrease in user experience. To address this issue, this paper introduces QoE-Doctor, a background application and computing resource management approach designed to optimize user experience on mobile devices, which identifies bottleneck resources and efficiently limits improper resource utilization from background applications. With a deep reinforcement learning agent, this approach selects the optimal CPU frequency and adjusts the allocation strategy between foreground and background applications, leading to improved QoE on mobile devices. Our evaluations implemented in Google Pixel 4 with various applications show that QoE-Doctor can boost QoE by 1.6 times on average compared to state-of-the-art approaches.