With the exponential growth of computation- intensive and latency-sensitive applications in 5G, it is hard to satisfy the heterogeneous requirements for increased data traffic with limited on-board resources. Mobile Edge Computing (MEC) has been considered as potential solution to offload computation task from User Equipment’s (UEs) to network edge in order to address certain challenges such as intolerable delay, high cost of resource utility in terms of energy and bandwidth. In this paper, we model the task offloading which aims at minimizing the overall energy consumption in task computation and latency requirements from both communication and computation aspect in MEC scenario. We first formulate the task offloading as classification problem while considering energy and latency constraints and then propose novel supervised learning-based classification techniques for classification of task, whether to offload or not, from UE to edge network. The numerical results demonstrate the capability of proposed offloading decision solution set to guarantee Quality of Experience (QoE) and offloading utility in terms of accuracy score for low- and high-energy devices.

Due to exponential growth in demand for radio spectrum for wireless communication networking, the radio spectrum has become over-crowded. The fixed spectrum allocation policy of the radio spectrum leads to inefficient utilisation of the available spectrum, which diverted the attention of researchers towards different intelligent techniques to access the spectrum dynamically and efficiently. The concept of Cognitive Radio (CR) has been considered as a promising technology to solve the problem of spectrum scarcity through the utilisation of various unutilised spectrum bands. In a future network deployment, multiple radio access networks may coexist having different characteristics. Hence, it becomes a challenge for CR networks to select the optimal network out of available networks. For efficient realisation, CRs requires intelligent spectrum management techniques for Dynamic Spectrum Management (DSM). Till now, there does not exist a literature survey that addresses the spectrum management with machine learning techniques in an intelligent manner. Hence, this paper presents the detailed classification and comprehensive survey of various machine learning techniques for intelligent spectrum management with their paradigms of optimisation for cognitive radio networks. The paper also provides new directions and open issues for the research community to work further in CR networks.

The rapid development of wireless traffic pushed the wireless community to research different solutions towards the efficient utilization of the available radio spectrum. However, a recent study shows that most of the dynamically allocated spectrum bands (radio frequency resources), experience significant underutilization as cognitive radio (CR) technology still lacks intelligence. An intelligence in CRs can be incorporated with machine learning algorithms. Further, the perfect channel state information (CSI) is hardly obtained and CSI imperfections play a crucial role in Dynamic Spectrum Management. Thus, for efficient utilization of available spectrum, a decentralized Multi-Agent Reinforcement Learning based resource allocation scheme has been proposed. A robust resource allocation scheme is proposed which integrates machine learning and CR technology into a sophisticated multi-agent system (MAS). Moreover, assisted with cloud computing which provides a huge amount of storage space, reduces operating expenditures, and provides wider flexibility of cooperation. Hence, to foster the performance of the proposed scheme, a cooperative framework in MAS is introduced which enhances the performance of the proposed scheme in terms of network capacity, outage probability, and convergence speed. Numerical results verify the effectiveness of the proposed scheme and show the non-negligible impact of imperfect CSI, thus highlighting the importance of robust designs that maintains users’ QoS in practical wireless networks.

The explosive growth of heterogeneous mobile data traffic with stringent Quality of Service (QoS) requirements put significant pressure on the existing network infrastructure and also present an important challenge even to very anticipated Cognitive Radio (CR) networks. In this perspective, to meet the demand of upcoming QoS requirements tied to their distinctive needs for resources, an intelligent solution is required which offers higher flexibility and to handle both current and future upcoming QoS challenges. The conventional resource allocation strategies face significant problems in meeting the increasing demand for bandwidth-hungry services due to stochastic behavior of wireless networks. To solve this problem, an efficient resource allocation scheme is proposed which maximizes energy efficiency while maintaining QoS requirements for all users. The Reinforcement learning based Q-Learning (Q-L) scheme is found most suitable for green resource allocation based on current network conditions that manages the QoS provisioning for heterogeneous traffic even under dynamic environmental conditions. Further, to enhance the performance of such sophisticated scheme, cooperative framework is introduced to improve the convergence speed. Experimental and simulated results demonstrate the effectiveness of proposed cooperative scheme. The proposed scheme outperforms current benchmark schemes in terms of meeting the energy-efficiency and stringent heterogeneous QoS requirements.

The most prominent challenge to the wireless community is to meet the demand for radio resources. Cognitive Radio (CR) is envisioned as a potential solution that utilizes its cognition ability intended to enhance the proper utilization of available radio resources and improves energy efficiency. However, due to the co-existence of Primary Base Stations (PU-BSs) and Cognitive Base Stations (CR-BSs) in CR networks, the problem of aggregated interference occurs which poses a critical challenge for resource allocation in CR networks. Moreover, in practical scenarios, it is difficult to form the correct network model due to complex network dynamics beforehand. Therefore, this work presents Multi-Agent Model-Free Reinforcement Learning schemes namely Q-Learning (Q-L) and State-Action-Reward- (next) State- (next) Action (SARSA) for resource allocation which mitigates interference and eliminate the need of network model. The proposed schemes are implemented in a decentralized cooperative manner with CRs act as multi-agent, forms a stochastic dynamic team to obtain optimal energy-efficient resource allocation strategy. Numerical results reveal that: 1) proposed cooperative scheme 1 (Cooperative Q-L scheme) expedites the convergence; 2) proposed cooperative scheme 2 (Cooperative SARSA scheme) achieves significant improvement in network capacity. Both the proposed cooperative schemes demonstrate its effectiveness by providing significant improvement in energy efficiency and maintain users' QoS.