(An) energy-efficient base station sleep scheme for periodic uplink transmission of industrial iot in private 5G networks

Abstract

In 2023, the number of connected Internet of Things (IoT) devices is around 16.7 billion, and it is predicted that by 2025, the global number of connected IoT devices will exceed 30 billion. Connected IoT enables better insights, automation, and decision-making through data collection and exchange in industrial and everyday life. The continuous increase of IoT devices in the industrial sector underscores the need for 5th generation (5G) communication technology to meet high-performance requirements like massive connectivity, high-speed data processing, ultra-low latency, and enhanced security. The proliferation of these IoT devices plays a crucial role in the advancement of industrial processes, enhancement of automation, and improvement of efficiency, for which 5G private networks are being introduced. These networks, isolated from public networks, reduce interference from other networks and enhance performance by densely deploying small base station (BS) to increase network coverage and capacity. However, BS energy consumption accounts for $60~80\%$ of the total network energy consumption, even consuming over $50\%$ of this in standby mode without transmitting traffic, thus increasing operational expenses (OPEX). Therefore, industrial operators must consider the balance between improved network performance and increased OPEX due to BS energy consumption. Sleep mode methods for BS, as an effective approach to solving the problem between energy savings and network performance optimization, support the high QoS requirements of 5G while reducing energy usage. 3GPP proposes extending the signaling period to enable deeper sleep modes, but deciding on the sleep mode for each BS in an Ultra-Dense Network requires considerable computational operation. To address this, research based on Reinforcement Learning, considering energy efficiency and QoS, has been conducted to find the optimal sleep mode for BS. However, previous studies have focused on reducing energy consumption for downlink traffic from servers to IoT devices. Recently, most industrial IoT applications involve periodic uplink data transmission from IoT devices to servers for industrial environment monitoring. This dissertation proposes a method for transitioning to sleep mode between uplink data transmission of industrial IoT devices connected to BSs in order to minimize energy consumption in industrial IoT without compromising network performance. To address the complexity of deciding on sleep modes in a dense BS environment, the dissertation considers the optimization problem between the energy efficiency of the BS and network performance, applying a Proximal Policy Optimization (PPO) based method to reduce computational complexity. This dissertation defines an optimization problem to minimize BS energy consumption while meeting the performance requirements of IoT services, determining the optimal sleep mode for BS. The dissertation builds a simulation environment referencing 3GPP 5G industrial IoT standard documents, analyzing performance in terms of network throughput, energy consumption, and energy efficiency. Additionally, this dissertation introduces an energy-conscious BS sleep mode control through collaboration with IoT devices in industrial 5G networks, which comprehensively considers the increase in energy consumption that occurs during the handover process of IoT devices as a result of the BS's transition to sleep mode. This research extends the PPO-based optimization problem to lower computational complexity, simultaneously considering the energy efficiency of both BS and IoT devices in dense networks, thus enhancing overall energy efficiency. This method demonstrates superior performance in terms of network throughput, energy consumption, energy efficiency, and expected lifespan of IoT devices compared to other algorithms. This significantly contributes to the development of an energy-efficient BS sleep method and verifies the performance in depth to increase the overall energy efficiency of the private 5G industrial network.