Deep Q Network Based Power Allocation for Uplink 5G Heterogeneous Networks

Abstract

The next generation heterogeneous network (HetNet) consists of multiple technologies for the device to improve their quality of service (QoS) parameters for the ubiquitous connectivity. The key technology has the ability to use machine intelligence in the design of an energy-efficient HetNet. That allows internet of things (IoT) devices to choose which base station (BS) to connect with for optimal performance; the proposed QoS-aware deep Q network (Q-DQN) algorithm adapts an energy-efficient reward function to improve the performance of femto BS IoT devices without deviating from macro BS. The main objective is to ascertain the QoS requirement that should exceed the threshold level. The performance of the proposed work is validated through the QoS, system capacity, and energy efficiency. A dynamic power selection strategy in a HetNet is based on the Q-DQN algorithm subject to network QoS parameters. The Q-DQN power allocation using reinforcement learning in uplink HetNet offers a powerful approach to managing the complex trade-offs between QoS requirements and energy efficiency. By dynamically adjusting power levels based on real-time conditions, the comparative results are evident that the proposed algorithm provides improved capacity and energy efficiency in proportion to the escalating throughput enhancement.