Mobile Edge Computing (MEC) is considered an essential future service for the implementation of 5G networks and the Internet of Things, as it is the best method of delivering computation and communication resources to mobile devices. It is based on the connection of the users to servers located on the edge of the network, which is especially relevant for real-time applications that demand minimal latency. In order to guarantee a resource-efficient MEC (which, for example, could mean improved Quality of Service for users or lower costs for service providers), it is important to consider certain aspects of the service model, such as where to offload the tasks generated by the devices, how many resources to allocate to each user (specially in the wired or wireless device-server communication) and how to handle inter-server communication. However, in the MEC scenarios with many and varied users, servers and applications, these problems are characterized by parameters with exceedingly high levels of dimensionality, resulting in too much data to be processed and complicating the task of finding efficient configurations. This will be particularly troublesome when 5G networks and Internet of Things roll out, with their massive amounts of devices. To address this concern, the best solution is to utilize Machine Learning (ML) algorithms, which enable the computer to draw conclusions and make predictions based on existing data without human supervision, leading to quick near-optimal solutions even in problems with high dimensionality. Indeed, in scenarios with too much data and too many parameters, ML algorithms are often the only feasible alternative. In this paper, a comprehensive survey on the use of ML in MEC systems is provided, offering an insight into the current progress of this research area. Furthermore, helpful guidance is supplied by pointing out which MEC challenges can be solved by ML solutions, what are the current trending algorithms in frontier ML research and how they could be used in MEC. These pieces of information should prove fundamental in encouraging future research that combines ML and MEC.

Mobile devices have several restrictions due to design choices that guarantee their mobility. A way of surpassing such limitations is to utilize cloud servers called cloudlets on the edge of the network through Mobile Edge Computing. However, as the number of clients and devices grows, the service must also increase its scalability in order to guarantee a latency limit and quality threshold. This can be achieved by deploying and activating more cloudlets, but this solution is expensive due to the cost of the physical servers. The best choice is to optimize the resources of the cloudlets through an intelligent choice of configuration that lowers delay and raises scalability. Thus, in this paper we propose an algorithm that utilizes Virtual Machine Migration and Transmission Power Control, together with a mathematical model of delay in Mobile Edge Computing and a heuristic algorithm called Particle Swarm Optimization, to balance the workload between cloudlets and consequently maximize cost-effectiveness. Our proposal is the first to consider simultaneously communication, computation, and migration in our assumed scale and, due to that, manages to outperform other conventional methods in terms of number of serviced users.

Cloud computing is an important technology for bringing a big pool of elastic resources to client devices. Their main drawback has long been the long distance between users and servers, but this has been remedied by Edge Cloud Computing, where the cloud servers are located in the network edge. Edge Cloud Computing is regarded as essential for future networks and consequently, there is plenty of research on how to optimize its operation. However, the vast majority of them ignore the decision of where the edge servers should be deployed, despite how severely this can affect the performance of the system. Furthermore, future networks must also deal with massive amounts of clients and servers, such as the ones characteristic of the Internet of Things and 6G Networks. This demands solutions that are scalable. Given these two points, we propose a Machine Learning-based server deployment policy in 6G Internet of Things environments. Our solution is proven to approach optimality while being feasible. Furthermore, we also prove that our proposal leads to lower latency and higher resource efficiency than conventional Edge Cloud Computing server deployment solutions.

This paper presents a novel method for radio map construction that simultaneously interpolates the received signal power values over space and frequency domains. Radio maps can be used to improve spectrum management and for localization systems, which are related to wireless systems in general such as cellular systems, Internet of Things (IoT) networks, and vehicular communications. Researchers have shown that crowdsourcing using spatial interpolation techniques can be used to accurately construct a radio map to improve these applications; however, because of the limitation in the spatial domain, conventional methods can build a radio map for only those frequencies at which sensing can be performed. Our proposed method focuses on the fact that the shadowing values show strong correlation over a wide range of frequency domains. The main idea is to treat the shadowing values obtained over various frequencies as the ones over target frequency. Using the actual datasets obtained over 870, 2115, and 3500 MHz in a cellular system, we show that the proposed method can accurately generate a radio map, even if no (or few) datasets are available in the target frequency.

Radio maps are a promising technology that can boost the capability of wireless networks by enhancing spectrum efficiency. Since spatial interpolation is a critical challenge to construct a precise radio map, the latest works have proposed deep learning (DL)-based interpolation methods. However, a DL model that achieves enough estimation accuracy for practical uses has not yet been established due to the complexity of radio propagation characteristics. Therefore, we propose a novel DL framework that transforms the spatial interpolation problem into a shadowing adjustment problem suitable for DL-based approaches. We evaluate the performance using real measurement data in urban and suburban areas to show that the proposed framework outperforms the state-of-the-art deep learning models.

The concept of energy harvesting small cell networks (EH-SCNs) has attracted a great deal of attention due to their potential to meet the exponential growth of mobile data traffic with low capital expenditure and operating expenditure. In addition, they are expected to work as a postdisaster wireless access network since exploiting green energy sources to power small cell base stations can overcome the power disruption caused by disasters. However, since the amount of harvested power is not enough for the stable network operation, the networks need to equipped with energy-efficient networking techniques. Additionally, in order to share the accurate postdisaster information, the networks should satisfy a certain quality of experience (QoE) of users in disaster areas. To deal with these challenges, we propose a user location maneuvering method, which advisory controls user's movement to improve the service time of EH-SCNs while guaranteeing a certain QoE level. We also propose an algorithm based on particle swarm optimization, which is configured with appropriate settings to find the best location of users efficiently. Extensive computer simulations show that the proposed maneuvering method can improve the QoE guaranteed service ratio, compared with the existing network operation approaches.