KR20210026171A - Multi-access edge computing based Heterogeneous Networks System - Google Patents

Abstract

The present invention provides a heterogeneous network system based on multi-access edge computing, wherein according to the integration of MEC and HetNets, the present invention discloses an architecture of MEC-enhanced HetNets which can mount an MEC server in a flight base station (BS) which supports both wireless and wired backhaul solutions.

Description

Multi-access edge computing based Heterogeneous Networks System

The present invention relates to a heterogeneous network system, and in particular, a flight base station (BS) capable of supporting a wireless and wired backhaul solution required for the integration of multi-access edge computing (MEC) and HetNets (Heterogeneous Networks), and mounting a MEC server. It relates to a MEC-based heterogeneous network system having a.

With the increasing popularity of smart devices, applications serviced in 5G system environments such as online games, augmented reality and intelligent video acceleration are being developed. These new applications require intensive calculations, high energy consumption, and high latency. However, most portable devices cannot adequately process these applications due to battery capacity and low computing power.

Recently, carriers are accelerating the construction of multi-access edge computing (MEC) sites that enable various functions of the cloud to be used directly on mobile access networks to support 5G services. In other words, MEC is an edge computing-based technology that implements ultra-low-latency services by placing nodes as close to the user as possible, and a management function that remotely monitors and controls high-density networks as these new computing locations supporting 5G become online. Is expected to play an important role in maintaining profitability.

Meanwhile, heterogeneous networks (HetNets) composed of macro base stations (MBSs) and small base stations (SBSs) such as pico cells, femto cells, and micro base stations are used as a way to overcome the rapid increase in data traffic and speed requirements. In the heterogeneous network, there are two main types of backhaul including wired and wireless backhaul solutions, and many factors such as implementation cost, traffic load, and quality of service (QoS) must be considered in the selection of the backhaul type.

In addition, although such heterogeneous networks improve network capacity and service range, there is a problem in that network performance is degraded due to severe interference due to high-density cells.

Nevertheless, the above-described MEC and HetNets can provide functions such as low latency, real-time mobility and high computational function (MEC), high spectrum energy efficiency and coverage extension (HetNets) required in 5G networks, which are key in 5G service technology. It is recognized as a technology.

However, the overall network configuration in which HetNets and MEC are integrated has not yet been introduced. That is, although interest in the MEC is increasing, it will be said that most of the studies are still designed and operated for a single server MEC network.

1 shows a configuration diagram of such a single server MEC network, where (a) is a single cell and single server MEC system configuration, and (b) is a multi-cell and single server MEC system configuration. From this point of view, from the perspective of computing offloading and resource allocation, MEC may be a simplified configuration of a network model such as one mobile user (MU), one MEC server, and a dedicated backhaul link with infinite capacity in a small cell. .

However, in such a system configuration, the MEC server is subject to serious overload. Mobile traffic, for example, is expected to increase by about 1000 times over the next 10 years, and by 2020, the number of connected users is expected to reach 50 billion.

In addition, the implementation of HetNets has a problem in data transmission through a backhaul link with limited capacity when considering the ultra-high density HetNet of 5G networks.

For this reason, it was required to integrate the two core concepts of MEC and HetNets. This need is to provide a feasible and realistic solution for 5G networks to take advantage of the individual benefits of MEC and HetNet to meet the critical requirements of applications and services in 5G networks. For example, the latency-sensitive Internet of Things (IoT) is bound to rely on MECs and HetNets for very low latency requirements and large connectivity.

Accordingly, an object of the present invention is to solve the above problems, and to provide a MEC-based heterogeneous network system that supports wireless and wired backhaul solutions required for the integration of MEC and HetNets.

Another object of the present invention is to provide a MEC-based heterogeneous network system with a flying base station (BS) capable of mounting a MEC server.

Another object of the present invention is to provide a MEC-based heterogeneous network system for computation offloading and resource allocation in a single server MEC network.

Another object of the present invention is to design multiple MEC servers so that a mobile user (MU) can select an available MEC server, thereby reducing the load burden of a single MEC server, reducing the consumption of transmission energy, and reducing the execution waiting time. It is to provide a MEC-based heterogeneous network system that can be shortened.

The present invention for achieving the above object supports a wireless and wired backhaul solution required for the integration of multi-access edge computing (MEC) and HetNets, and a MEC equipped with a flying base station (BS) capable of mounting a MEC server. It provides a heterogeneous network system based on.

According to the MEC-based heterogeneous network system of the present invention as described above, since it supports the coexistence of wired and wireless backhaul links and provides a MEC-based Hetnet network that can be used as a flying MEC server by deploying an unmanned aerial vehicle (UAV), natural disasters and Active response to emergency situations is possible.

In addition, since a free space optical link (FSO) is used between the UAV and the small cell, the advantages of FSO communication can be applied as it is.

In addition, by supporting the coexistence of the existing central cloud and distributed MEC, the availability of cloud services can be guaranteed.

1 is a typical single server MEC network configuration diagram,
(a) is a single cell and single server MEC system configuration diagram
(b) is a multi-cell and single server MEC system configuration diagram
Figure 2 is a diagram summarizing the existing research on MEC in terms of offload determination and resource allocation
3 is a graph comparing the result of offloading probability and overhead according to the conventional method and the method of the present invention
4 is a block diagram of a heterogeneous network system capable of multi-access edge computing according to a preferred embodiment of the present invention.

The present invention is to propose a MEC-based HetNet architecture that supports the coexistence of wired and wireless backhaul links and enables use as a MEC server in flight by arranging an unmanned aerial vehicle (UAV). It will be described in more detail with respect to the present invention.

First, computing offloading and resource allocation in a single MEC network and a multi-server homogeneous network will be examined.

Computing offloading refers to the transfer of specific computing tasks to external platforms such as xluster, grid, and cloud. Such computing offloading may be necessary due to the hardware limitations of the computer system in processing specific tasks by itself, and intensive computation tasks include artificial intelligence, artificial vision, object tracking, or computational decision-making with computers. It is used in decision making).

Of course, such computing offloading has several challenges such as efficient communication and computing resource allocation, selection of an appropriate MEC server among available servers, and efficient management of simultaneous offloading by multiple MUs.

Two important aspects of computing offloading are actually determining the application model and managing the offloading process. The determination of the application program model refers to the possibility of offloading the application program, determines the amount of processed data, and specifies the dependencies between the offloadable parts, while the management of the offloading process determines the target to be offloaded by the MU. Determining and using channel connections, system bandwidth, and computational resources available on the MEC server.

Therefore, the general goal of computing offloading is to reduce execution latency, reduce energy consumption by migrating computing tasks to the server, and optimize the balance between energy consumption and execution latency. The unique requirements of the target application are taken into account in order to utilize the MEC when designing up to the user-perceived performance at the system level, for example minimizing the latency required for the overall execution, and processing IoT applications. In other words, since batteries are limited to IoT devices, a balance between computation and transmission energy consumption must be achieved.

The MEC network allocates important resources to achieve the goal pursued by computing offloading. Here, the resource includes the calculation speed of the MU, the calculation resource in the MEC server, the transmission power of the MU, and the bandwidth and time allocated to the offloading user.

FIG. 2 is a diagram that summarizes existing studies on MEC in terms of offload determination and resource allocation. For reference,'study' in FIG. 2 indicates related papers, but all descriptions of the papers are omitted in this specification.

On the other hand, as described above, the study on MEC is designed for a single server MEC network as shown in FIG. 1. Looking at this, MEC HetNets in which a plurality of small cells are overlapped in a macro cell, macro BS is equipped with a MEC server, and MU Is a method of offloading a task to the MEC server through a small BS. However, in this case, a single MEC server can be severely overloaded and the load can be concentrated.

Compared to this single server MEC network, multiple MEC servers have the following advantages: First, the MU can migrate computational tasks among multiple available MEC servers, thus reducing the computational burden on a single MEC server.

Second, since the MEC can be within coverage of multiple cells (i.e. multiple MEC servers), the MU can offload calculations to the server with better channel conditions and calculation capabilities. As such, the MU can save transmission energy consumption (i.e., total energy consumption) and shorten execution waiting time.

Third, in the case of partial offloading, the computational work can be divided into smaller parts and offloaded to another MEC server, further reducing the execution waiting time.

Meanwhile, the MEC server can reduce inter-cell interference (ICI) between users offloading from adjacent MEC servers by coordinating information exchange and improve resource allocation efficiency. This collaboration between different MEC servers can potentially help increase cluster resources, thus supporting applications with higher computational requirements.

Of course, in a multi-server MEC network, the offload decision problem is complicated due to the availability of multiple servers. In addition, since the trunk line phenomenon exists between cells, the problem of resource allocation in a multi-server MEC network is much more difficult than that of a single server MEC network.

So, to solve this problem, spectral resources within each cell can be divided into orthogonal sub-channels (for example, sub-channels that the MU should use to offload the computational task to the MEC server), which must be efficiently allocated to the MU. . Network performance can be further improved by optimizing the subchannel allocation in common and adjusting the transmit power in the MU.

Such MEC servers may be deployed in a predetermined area in a large amount, and the association between the user and the MEC server largely depends on the location of the MEC server. That is, a specific area can be densified by a large number of MEC servers, but only one MEC server can be deployed in another area. In this case, the MU can run applications and services with different priorities and security levels. For example, some services can only run on private MEC servers due to security considerations.

In addition, the calculation task may have different input data size, calculation workload, and execution deadline. Therefore, the problem of common resource allocation should take these characteristics into account.

The present invention proposes distributed computing offloading and resource allocation in a multi-server MEC network in consideration of technical features such as computing offloading or multiple MEC servers. In order to enable distributed computing offloading, matching theory, which is a powerful tool for designing a distributed algorithm of multiple resource allocation problems in wireless communication, is adopted.

3 is a graph comparing the result of offloading probability and overhead according to the conventional method and the method of the present invention. From this, it can be seen that the system (JCORAMS) proposed in the present invention is superior to the alternative technique in terms of offload probability and computational overhead.

In other words, as the number of MUs increases, the probability of using the MEC server and sub-channel that each MU prefers decreases, and the inter-cell interference (ICI) among MUs becomes more severe, so the offload probability decreases and the total computational overhead increases as the number of MUs increases. You can see that there is.

Here, looking at (b) of FIG. 3, it can be seen that it is not always good to offload all calculation tasks. That is, when the number of MUs is sufficiently large, performance may be worse than that of the local only method, because a limited number of radio and computational resources must be shared by many MUs.

Meanwhile, the present invention provides a MEC-based HetNet architecture that supports coexistence of wireless backhaul links and deploys an unmanned aerial vehicle (UAV) to use it as a flight MEC server.

As is known, there are many advantages, but it is costly and time consuming to provide wireless backhaul connectivity to ultra-dense small cells. There are also many factors to consider for the deployment of wired backhaul, such as the location of small cells (SBS) and QoS requirements of mobile users.

The wireless backhaul can use several frequency bands. For example, 6GHz to 60GHz microwave band, millimeter wave band including V-band (57 to 76GHz), E-band (71 to 76, 81 to 86, 92 to 95GHz), W-band (92 to 94, 94.1 to 100, 102 to 109.5 and 111.8 to 114.25 GHz) and D-band (130 to 134, 141 to 148.5, 151.5 to 164 and 167 to 174.8 GHz), sub 6 GHz band, satellite frequency band, and TV blank band.

In addition, free-space optical communication is a method that can efficiently deploy high-capacity wireless backhaul in ultra-high-density HetNets networks.

Therefore, the achievable capacity of wireless backhaul will reach 10Gbps or more, and wireless backhaul can be a practical solution in high-density HetNets networks. In addition, wireless backhaul can provide wireless access to some areas, such as hard-to-reach rural areas, while reducing maintenance costs.

In addition, 5G network abnormalities are expected to support emergency services and disaster situations. Wired backhaul failures seriously affect network response and stability in that fast repairs are difficult, while wireless backhaul links can solve these problems. .

Due to this problem, it is to provide an unmanned aerial vehicle (UAV) support radio system as an alternative solution to meet temporary or unexpected demands when ground communication is not satisfactory. Recently, unmanned aerial vehicle (UAV) has been used in a variety of applications. This means that UAVs can be deployed as an airplane base station (BS), which is used to provide communications to ground users and increase network coverage and capacity, thanks to UAVs' continued cost savings.

In addition, UAVs can be quickly deployed in disasters, emergencies or unexpected situations. For example, if the terrestrial network infrastructure does not provide stable communication and recovery is delayed, a UAV can be deployed for internet connection.

In addition, network performance can be improved by jointly optimizing the UAV's state and communication. End-user QoS requirements can be met, for example, in that the energy consumption of UAVs can be improved by optimally adjusting flight speed, direction and acceleration, and reliably connected with potential users by adjusting altitude and mobility. Can be set.

In addition, it is possible to improve the signal from the UAV to the ground user through the visible range communication link, thereby reducing the effect of signal shading and blocking.

Of course, due to the limited weight and size, the on-board energy of the UAV will be limited, and the performance of the UAV-enabled wireless system will be greatly affected.

Therefore, in order to maximize the use of UAV in wireless communication, air-to-ground channel modeling for UAV-BS communication and ground-to-membrane channel modeling for BS-to-Air communication A plan like this should be prepared. For example, it is necessary to propose an optimal modeling through UAV communication, optimal UAV placement, energy recognition operation of UAV, interference recognition resource management of multiple UAV systems, and performance analysis of UAV supported wireless networks.

Meanwhile, 5G networks support not only communication but also computation, caching, and control (4C), which requires a new architecture that can integrate UAV communications, edge computing, and heterogeneous networks.

In the past, there have been studies in which UAVs were integrated into HetNet. However, due to the high mobility of the UAV and the dense deployment of HetNet, there were some obstacles to designing an effective interference management system. For example, power failure due to damage and malfunction of BS, mobility and dynamic location of UAV, transmission of burst data due to failure or disaster, high QoS and low Latency requirements, etc.

In order to solve this problem, efforts have been made to minimize the maximum distribution delay of the UAV and the overall deployment delay. That is, representative scenarios have been presented to show the potential applications of UAV-supported HetNet, including UAV as a flying base station (BS), UAV as mobile relay, UAV-assisted energy transmission, and UAV-assisted caching. It also showed superior performance of UAV-supported HetNets network compared to static networks, and showed coordination between UAVs, energy limitation of UAVs, flying ad hoc networks and UAV deployments.

As another example, a two-layer air-ground mobile edge network architecture (MEN) has been proposed. MEN is a scheme that combines the three basic components of MET, including network densification, moving edge caching, and MEC.

As another example, architectures such as network interworking, MEN performance evaluation methods, prediction and optimization of communication (ie, access, front haul and backhaul) links, and software-defined networking-based control and communication systems are also presented.

However, these schemes are a configuration in which the air and ground layers are connected to each other.

The network architecture proposed in the present invention is proposed in terms of MEC and HetNets, and each MU may be provided by SBS, MBS or flight UAV supporting BS. And communication between the two SBS and MBS is connected through a wireless backhaul or a wired link.

Given the importance of MEC and HetNets, the 5G network can be said to be an architecture that can perform the integration of MEC and HetNets, coexistence of wireless and wired backhaul for MEC deployment, and UAV support for disasters and emergencies. That is, it is an architecture for MEC-supported UAV-supported HetNets.

The configuration proposed by the present invention will be described with reference to FIG. 4. The present invention is a heterogeneous network system capable of multi-access edge computing, and a MEC-supported HetNets architecture designed in consideration of various communication and computational resources.

The MBS 110, which is a wireless base station shared by various network operators to physically deploy the base station, is provided. MBS 110 may be a wireless cellular tower, and is shared among multiple network operators to physically deploy individual BSs. Therefore, small cell network operators may be different. The MBS 110 is linked to one or a group of MEC servers 120 through a dedicated fiber link in order to provide a service to the MU.

As described above, the integration of MEC and HetNets supports the coexistence of conventional wired and wireless backhaul so that small cells can be flexibly adapted to the actual network conditions and deployed. In the embodiment, a total of three cells are provided. Cells, for example, in hard-to-reach locations, small cells 1 (SBS 1) 130 and 3 (SBS 3) 150 are connected to MBS 110 by a wireless backhaul link, whereas small Cell 2 (SBS 2) 140 is configured to transmit and receive data from the MBS 110 through a conventional wired backhaul (wire link) connection.

The calculation of the MU in the small cells 1 (SBS 1) 120 and 3 (SBS 3) 150 is first offloaded to the MBS 110 through a wireless backhaul link, and then to the MEC server 120 through an optical fiber connection. It is off-road. On the other hand, since the small cell 2 (SBS 2) 140 is connected to the MBS 110 through a wired backhaul, the MU associated with the small cell 2 (SBS 2) may ignore the offloading time through the wired backhaul.

Meanwhile, the MBS 110 and the MU 1 160 are directly connected. Therefore, the computing time includes offload time (transfer time from itself to MBS) and time due to limited computing resources of MEC server 120. Therefore, the offload time from the MBS 110 to the MEC server 120 over the optical fiber link can be completely ignored.

Much of the existing literature on MEC networks relies almost entirely on the assumption of infinite backhaul capacity, which is unsuitable for existing multi-cell networks, but it can be further exacerbated in HetNets with MEC rights. Due to this limitation of backhaul capacity, resource management issues require not only to consider other functions such as working links and heterogeneous computational functions, but also to allocate the transmission bandwidth over the access link and backhaul connection.

The architecture of the present invention supports coexistence of an existing central cloud and a distributed MEC to ensure availability of cloud services in the entire network. Unlike the MEC server 120, the central cloud has a much more powerful computational function, allowing a large number of signal tasks to be centrally processed.

Due to the existing centralization of cloud computing, some limitations such as high latency and a burden on the front link link may occur. However, decentralized MECs are more suitable for new applications and services such as real-time video games, IoT, and mobility-related and location-aware applications.

According to the present invention, the MBS 110 is connected to a centralized cloud 190 (Clon Computing Center) through an optical fiber backbone network. In the C-RAN (Cloud Radio Access Network), the BS consists of a distributed remote radio head (RRH) and a centralized base band unit (BBU) responsible for baseband signal processing functions.

The multi-tier computing architecture can be designed in a C-RAN where the cloud center 190 is deployed in a central BBU and edge servers are deployed in a distributed RRH (i.e., macro RRH and small RRH). Through this hierarchical structure, the MU can utilize various computing servers with heterogeneous computing functions, and can perform calculations to improve computing performance by cooperating with computing servers of the same or different tiers.

And the central server can be integrated into the cellular network, but can be standalone such as AmazonWeb Service, Microsoft Azure and Google Cloud Platform.

In order to reinforce the calculation function of HetNet supported by the MEC server 120 and to ensure coverage and services for natural disasters and emergency situations, a UAV-mounted BS 170 with computing resources is provided. That is, the present invention can provide the following scenario using UAV-capable MEC-enhanced HetNets.

First, because traditional terrestrial communication infrastructure can be damaged or partially broken after natural disasters such as earthquakes, UAVs are used to (i) collect information such as building damage and factory photos and transmit site information remotely, ( ii) the composition of finding people in the earthquake-prone area, and (iii) providing basic services to the MU.

In addition, during virtual reality or large-scale games, the existing MEC server 120 is limited to computing and may not be able to provide additional services, so more UAV-equipped MEC servers 170 are used to (i) monitor illegal activities. And (ii) make it possible to create and display a 360° soccer game in virtual reality.

Similar to offloading work from MU 1 160 to MEC server 120 as described above, MU 2 162 may directly or indirectly offload to the UAV-equipped MEC server 170. For example, MU 2 (162) does not have a small cell or MBS (110) to connect, so the UAV-equipped MEC server (170) adjusts the altitude and position close to that of MU 2 and calculates the MU 2 (162). Process. In addition, MU 2 162 offloads the calculation work to the UAV-equipped MEC server 170.

Recently, free space optical (FSO) communication is very attractive for a wide range of applications including backhaul/front hauling for wireless networks. The main advantages of the FSO communication are ease of construction, long-distance transmission of up to several kilometers, high data rates (more than 10 Gbps), robustness against electromagnetic interference, enhanced security, and full-duplex operation.

Therefore, in the present invention, the fronthaul link between the UAV-mounted MEC server 170 and the small cell 150 is connected using FOS. The UAV-mounted MEC server 170 can be used for various purposes in that it can be disposed at different altitudes.

Since the UAV-mounted MEC server 170 is difficult to use for a long time due to limited on-board energy, it cannot perform sufficient work processing. Accordingly, the UAV-equipped MEC server 170 can offload the computationally-intensive work to the ground MEC server.

In addition, the UAV-equipped MEC server 170 may set a ground control station (GBS) 180 and an FSO backhaul link to receive control signals and feedback collected information and statistical data.

The present invention is a MEC-enhanced HetNets architecture proposed by integrating UAV considerations with wired and wireless backhaul due to the potential function of MEC and HetNets, and may have additional problems.

In HetNets, when the inter-cell interference (ICI) increases, the wireless backhaul capacity is greatly reduced. Wireless backhaul can support speeds of several Gbps, which is much more stringent than that of conventional wired backhaul, which can provide speeds of several Gbps. On the other hand, although the wired backhaul capacity is fixed, it is very flexible to allocate bandwidth resources for wireless backhaul.

Therefore, the MBS 110 provides a wireless backhaul to a number of small cells 130, 140, 150 and simultaneously divides the bandwidth for radio and access transmission, and a different small cell that shares the same backhaul as the MBS 110. Allocates resources to different wireless backhaul connections, such as time division for (130, 140, 150).

In MEC-enhanced HetNets, the transmit power and bandwidth division factors of the SBS jointly affect the wireless backhaul capacity.

On the other hand, offloading is determined according to the characteristics of the computing task, the transmission power of the SBS (130, 140, 150), the bandwidth allocation factor, and the maximum computing power in the MEC server 120. That is, in order to improve network performance, there are problems with interference management, offloading decisions, and resource (bandwidth and calculation) allocation.

The inter-cell interference phenomenon (ICI) is a major obstacle to obtaining high performance in the existing multi-cell network as well as HetNets with MEC authority, so this is the following for mitigating the inter-cell interference phenomenon (ICI) between adjacent MEC servers 120 We need the same solution.

First, it is Fractional Frequency Reuse (FFR) that divides the cell space into several regions and allocates a unique frequency band to each region. By using this, the cell-edge user does not interfere with the cell-centered user and the cell-edge user of the adjacent cell may not be disturbed, and thus the quality of the received signal can be improved. Inter-cell and intra-cell interference can be greatly reduced through optimal FFR.

In addition, the transmission power of the MU is adaptively adjusted (power control). That is, if the transmission power of the offload MU is increased, the transmission speed and offload time can be increased.

Another is 3D beamforming. Since most studies on MEC simply assumed that the output result size was small and the downlink transmission power from the MEC server 120 was high, the downlink latency and energy consumption for returning the calculation result back to the MU are negligible. Multiple MUs can immediately request computational offloading, e.g. in an augmented reality application, multicast the output from the MEC server 120 for a group of MUs, whereas MUs can be used in urban areas with many skyscrapers as well as horizontal spaces It can be distributed in the vertical direction, such as.

Therefore, 3D beamforming can dynamically control radio patterns in both vertical and horizontal dimensions. For example, it is possible to adapt different down tilts for the center zone and edge zone MU. Here, joint optimization of the downlink and the uplink may be considered by optimally controlling the transmission power of the down-tilt of the MEC server 120 and the antenna.

And by optimizing the UAV trajectory, the UAV can establish a line-of-sight connection to the relevant terrestrial BS (cellular connection mode) or terrestrial MU (UAV support mode) and at the same time avoid interference to the ground MU. This can provide significant gains, such as increasing transmission speed, and can provide terrestrial MU benefits due to computing offloading.

That is, due to the dense arrangement of the MEC server 120, the UAV can be connected to the terrestrial server, and since the transmission power of the UAV can be optimized, inter-cell interference (ICI) is alleviated and UAV durability is improved.

Another problem may be a problem related to the offloading decision.

Small cells 130, 140, 150 are connected to MBS 110 via a separate backhaul solution. For small cells 130, 140, 150 with wireless backhaul, in addition to the bandwidth dividing factor, the transmit power is very important to determine the backhaul capacity. In addition, small cells 130, 140, 150 are densely deployed in a 5G network, and MUs exist within the coverage of a number of small cells.

Therefore, each offload MU has to select a preferred SBS (130, 140, 150) in relation to the allocation amount, channel condition and backhaul quality of different cells, and in order to find the offloading decision of the MU in the cell, the corresponding BS The bandwidth is optimally partitioned so that the backhaul capacity constraint is not violated and the computational resources in the MEC server 120 are significantly shared among offloading users.

Here, in the centralized optimization of the offloading decision, since all MUs must report local information to a central entity, the overall computational overhead of the network may increase. In addition, heuristic methods are often adopted to solve centralized computing offload, and central entities in wireless networks are not always feasible.

Therefore, the distributed approach can be said to be a scheme in which each MU can independently determine an offloading decision using local information or limited information exchange. These include matching theory, game theory, and machine learning, and one can be adopted.

When designing a computing offloading policy, network dynamics such as channel quality that changes over time, dynamic job arrival rate, number of users requesting offloading, and number of MEC servers 120 must be considered. In addition, the integration of communications, computing, caching and control is essential to the successful implementation of many applications and services in new 5G networks.

For this reason, the MEC-enhanced HetNets of the present invention are required to optimize a vast amount of variables in the long term. Therefore, long-term approaches such as the Markov decision process and machine learning are particularly suitable. For example, the MU makes decisions about offloading and choosing a server in the current network state through interactions and past experiences with other MUs. And, using machine learning, the MEC server 120 can predict the operation of the MU. Allocating computational resources efficiently, and finally, UAVs can use machine learning to optimize trajectories and optimization variables such as transmit power and allocated computational resources.

Next, there is a problem related to hierarchical and collaborative computing offload. The development of 5G mainly consists of three categories: mobile broadband, large-scale IoT and mission-critical services. .

The first class is aimed at very high data rates and large capacity, and the second class is to serve a vast number of IoT devices that meet the requirements of ultra high density, ultra low energy and ultra high scalability. It is suitable for critical applications with high requirements for strong security, ultra high reliability and ultra low latency.

In addition, the vast and heterogeneous MUs of emerging 5G networks are clearly different in battery size, computational capabilities and target applications. For example, in a network with two IoT devices, one runs an application that requires approximately 1 ms end-to-end latency, and the other continuously observes environmental conditions such as temperature, humidity and photography, and an aggregator for processing raw data. If passed to (aggregator), the latter MU can generate big data with large computational requirements for data processing. In this case, the latter IoT devices using computationally intensive and low-latency applications prefer to offload computational functions to higher-level servers.

Conversely, IoT devices that use latency-critical applications are recommended to use low-tier MEC servers.

In addition, the MEC server 120 has a size and calculation function similar to that of a desktop computer. Therefore, some computational tasks may not be strictly handled by only one MEC server. This motivates different servers to cooperate to run the same calculations to improve practical computing performance.

Hierarchical and collaborative computations become much more complex when jointly optimized with user connections and resource allocation for wireless backhaul. Further, compared to MUs with latency-sensitive applications, MUs with latency-sensitive applications have a higher priority in accessing MEC server 120 with higher computational resources and better connection quality.

Next is the selection problem of the MEC server 120. A plurality of MEC servers 120 in which the MBS 110 is a multi-operator wireless base station may be physically located in the same area.

In this case, in addition to the user connection problem due to various backhaul solutions, there is a very need for an efficient system for selecting the MEC server 120 suitable for offloading users.

For single-user networks, MUs are recommended to connect to small cells with wired backhaul and best channel quality and then offload to MEC servers with the highest computational capabilities.

However, the actual network scenario is multi-user, and small cells are usually limited in a few user quotas. As a result, it can be designed so that each MU can act on its own interests and make independent offloading decisions, such as an uncooperative game. Here, each MU selects the best small cells 130, 140, 150 and MEC server 120 to always minimize its own benefits, such as minimum energy consumption and computation time due to limited power.

The following is a MEC network supporting non-orthogonal multiple access (NOMA). MEC servers were originally proposed to improve the computational and computational capabilities of end-devices with limited battery capacity to run a variety of computationally intensive and latency-critical applications.

These MEC servers will have more than 50 billion devices connected to the Internet by 2020, and more than hundreds of millions of sensors will be deployed around the world. Thus, it can be expected that a large number of devices will simultaneously connect to a single MEC server for computationally intensive tasks.

In this case, in order to provide services to multiple MUs, bandwidth and time resources can be divided into orthogonal parts and allocated to other MUs for computation offload.

In addition, some studies on non-orthogonal multiple access (NOMA) capable MEC networks have recently reported that the combination of NOMA and MEC can reduce energy consumption and latency. However, as mentioned above, the existing work is limited to a specific scenario of a single server MEC network. For example, considering an offload user set without inter-cell interference (ICI), user clustering per cluster, resource allocation, and transmission power control are jointly optimized to minimize the energy consumption of offload users.

However, as more and more MEC servers are deployed, inter-cell interference (ICI) is regarded as a clear obstacle because inter-cell interference (ICI) severely limits the performance of cell edge users. So, user clustering, power allocation, and user scheduling (UCPAUS) are important issues to achieve the benefits of NOMA in MEC-powered HetNets, but ME servers are combined and solved.

One solution is to consider only one cluster in each cell and then allocate the transmit power of the MU. However, this solution has enormous complexity because the signals of all the same cell MUs are superimposed on the transmitter side, and the heavily superimposed signals are decoded at the receiver side.

Another solution is to have multiple clusters in each cell, assign each cluster to a sub-channel, and adjust the power allocation of each cluster. This reduces complexity and hardware requirements compared to the single cluster approach. However, the performance is still not optimized.

The joint UCPAUS problem is very complex in MEC-enhanced HetNets. First, each MU must determine the offload in case of binary offload or the percentage of offloadable calculations in case of partial offload. Second, it is necessary to determine the optimal number of clusters and how to select the MU to create the cluster. However, this is not easy. For example, a clustering scheme can allocate two latency-critical MUs to a cluster and two other latency-resistant MUs to another cluster. This does improve network performance. However, in the MU where the waiting time is important, the completion time for executing a predetermined operation may not be satisfied.

Therefore, one suitable solution is to group one latency-sensitive MU with one latency-sensitive MU.

Finally, the performance of the MEC network is greatly affected by the allocation of computational resources in the MEC server 120.

As a result, studies of offloading decisions, user clustering, and resource allocation are needed to further improve the performance of NOMA-capable MEC networks that require an efficient but close system for successfully deploying MECs and NOMAs in new 5G.

As described above, in the present invention, in the MEC-enhanced UAV-supporting HetNets, the MU can select and offload a preferred small cell, so that the user connection is the most important. And it is motivated by a heterogeneous wireless backhaul solution between different small cells. In addition, enough attention should be paid to the allocation of computing resources, as work requirements and computing overhead are greatly affected by the allocated resources. In addition, the present invention is to provide an architecture of MEC-enhanced UAV support HetNets.

Although it has been described with reference to the illustrated embodiments of the present invention as described above, these are only exemplary, and those of ordinary skill in the art to which the present invention pertains, without departing from the gist and scope of the present invention, various It will be apparent that variations, modifications and other equivalent embodiments are possible. Therefore, the true technical protection scope of the present invention should be determined by the technical spirit of the appended claims.

110: MBS
120: MEC server
130, 140, 150: first to third SBS
160: MU1, 162: MU2
170: UAV
180: ground control station (Ground BS)
190: Cloud Computing Center

Claims (5)

A wireless base station (MBS) connected to the MEC server through a dedicated fiber link to provide services to mobile users;
Two or more small cells connected to the wireless base station by a wireless backhaul link or wired backhaul;
A centralized cloud center connected to the wireless base station (MBS) through an optical fiber backbone network;
A flight base station equipped with computing resources that can be deployed at different altitudes to enhance the computational function of heterogeneous networks supported by the MEC server and to ensure extended coverage and service; And
And a ground control station (GBS) connected to the flight base station through an FSO backhaul link to receive control signals, information collected from feedback, and statistical data,
The flight base station is a multi-access edge computing-based heterogeneous network system that is connected to any one small cell and a fronthaul link using a free space optical link (FSO). The method of claim 1,
The flight base station,
A heterogeneous network system based on multi-access edge computing that offloads to the MEC server to perform computation. The method of claim 1,
The calculation of the mobile user in the small cell connected by the wireless backhaul link,
Multi-access edge computing-based heterogeneous network system performed by being offloaded to the wireless base station through the wireless backhaul link and then offloaded to the MEC server through an optical fiber connection. The method of claim 1,
The inter-cell interference phenomenon between the MEC servers is a multi-access edge controlled by using one or more of FFR (Fractional Frequency Reuse), mobile user transmission power adjustment, 3D beam forming (Beam froming), and flight trajectory adjustment of the flying base station. Computing-based heterogeneous network system. It supports multiple wireless and wired backhaul solutions required for the integration of MEC (multi-access edge computing) and HetNets (Heterogeneous Networks),
Further comprising a flight base station (BS) using a MEC server to control the MEC,
The flight base station is in the form of a UAV equipped with a computing resource function to enhance the calculation function of the HetNet and ensure coverage and services for natural disasters and emergency situations,
Any one small cell is connected to the FSO fronthaul link, and the ground control station (GBS) is connected to the FSO backhaul link to receive control signals, feedback collected information, and statistical data. Multi-access edge computing-based heterogeneous network system, characterized in that.

Images