Work

A Green Computation with Quality of Service Optimisation in a 6G Network: An AI-driven Deep Reinforcement Learning Approach Using Information Security

Year: 2024
Type: article
Abstract: The aggregation of connected devices and the development of more complex (data-intensive) applications have resulted in an increase in data traffic, which is a critical challenge for future 6G networks: an escalating energy demand. Green computation has been a desirable algorithmic approach for the deployment of 6G due to its emphasis on the energy footprint of ICT (information and communication technologies), which renders it environmentally sustainable. Therefore, in this paper, we investigate the ability of DRL, a cutting-edge AI technique, to optimise resource allocation and computation, an offloading strategy for green computing within 6G networks that ensures a satisfactory Quality of Service (QoS) for users and also improves the information security of the model. These sources underscore the limitations of classical optimisation techniques, which necessitate simplistic mathematical models that may not be appropriate for simulating the dynamic complexities and heterogeneous requirements of a 6G network. In contrast, DRL offers an alternative solution by allowing agents to learn optimal policies when interacting with the environment. In particular, this paper examines the potential of DRL agents to dynamically determine the timing and location of computing offloading processes on both the MEC servers (computation) and network configurations under stochastic traffic patterns and wireless channel conditions for energy management purposes. This is achieved by incorporating QoS constraints that ensure a latency time bound of 10 m/s is satisfied with a probability more than 95%. The paper will also utilise examples from the sources of DRL applications in systems such as AI-enabled Base Station (BS) administration. This strategy allows for the intelligent activation/deactivation of BSs, which in turn maintains low energy consumption during low-traffic periods. It will also examine the application of DRL to flexibly optimise energy harvesting, security of model, and trading for machine-type communications (MTC), a significant power-limited scenario in 6G. The paper will also address the challenges associated with the practice of DRL, such as the resource-intensive nature of training complex deep reinforcement learning models and the optimisation of reward model efficiency to balance energy conservation and QoS satisfaction. We also consider the challenges that DRL and other AI models face in order to fully realise their potential as green computation strategies when it comes to a final solution at the scale of operating 6G networks.
Language: English
Open Access status: hybrid
OpenAlex ID: W7217377165
FWCI:
Cites: 47
Cited by:
Related to: 10