This paper proposes a multi-level sleep mode control technique based on safe deep reinforcement learning (SDRL), which integrates long short-term memory (LSTM)-based traffic prediction to reduce the energy consumption of small cell base stations, as they account for a large proportion of heterogeneous networks. To mitigate quality of service (QoS) degradation caused by prolonged sleep states, the proposed framework incorporates safety constraints that proactively avoid excessive sleep continuity by predicting future traffic, while simultaneously restricting unacceptable mode transitions. The framework learns to balance these objectives through a reward structure that explicitly captures the trade-off between energy saving and QoS, and is implemented using a deep Q-network (DQN). Simulation results show that the proposed method ensures efficient and stable operation in dynamic traffic environments, achieving reduced energy consumption while stably satisfying QoS requirements compared to benchmark methods.

Solar energy harvesting offers an energy-efficient solution to reduce the growing power consumption of mobile networks. However, the intermittent nature of solar energy and the spatio-temporal variability of traffic demand make stable operation and resource management challenging. To address these issues, we propose a prediction-based distributed deep Q- network (DQN) approach that dynamically controls transmission power based on predicted traffic and harvested energy. Simulation results demonstrate improved energy efficiency and network performance over benchmark methods.

Heterogeneous multi-processing (HMP) is rapidly emerging as a promising solution for high-performance and low-power computing. Despite extensive prior work, system-software support for self-adaptive multithreaded applications has been little explored in the context of HMP. To bridge this gap, we propose HARS, a heterogeneity-aware runtime system for self-adaptive multithreaded applications. HARS continuously monitors the application performance and dynamically adapts the system state to enhance the performance/watt of the target self-adaptive multithreaded applications on HMP systems, while satisfying the user-specified performance goal. We quantify the effectiveness of HARS by demonstrating that HARS achieves significantly higher efficiency than the baseline version with the Linux HMP scheduler and comparable efficiency with that of the static optimal version.