High Frequency Communication Network with Diversity: System Structure and Key Enabling Techniques

Kun Xu, Bin Jiang, Zeyou Su, Shengqing Wang, Makun Guo, Xiao Li, Zhiyong Du＊

College of Information and Communication, National University of Defense Technology, Wuhan 430010, China

Abstract: High frequency sky wave communication suffers from poor performance including poor link quality and low link success rate. To enhance performance, diversity technology is proposed in the high frequency communication network (HFCN) in this paper.First, we present the bene fits and the challenges by introducing diversity technology into the existing HFCN. Secondly, to exploit the benefits fully and overcome the challenges, we propose a system structure suitable for deploying diversity technology in HFCN in large scale,based on the cloud radio access network and software defined network. Moreover, we present a general structure for the real-time updating frequency management system that plays a more important role especially when resource consuming (e.g., frequency) diversity technology is deployed. Thirdly, we investigate the key techniques enabling diversity technology deployment. Finally, we point out the future research directions to help the HFCN with diversity work more efficiently and intelligently.

Keywords: high frequency; high frequency communication network; diversity technology;access layer and core layer; frequency management system

I. INTRODUCTION

High frequency (HF) communication, commonly operating on the frequency band 3-30MHz, can realize long distance communication through reflecting signals back to the earth (i.e., sky wave propagation) from the ionospheric layer. This kind of communication possesses advantages like high maneuverability, convenient deployment, strong anti-jamming ability and fast network reconstruction.Due to these merits, it has been widely used in various fields including military, transoceanic and emergency communication [1]-[3]. The basic operating mode of HF is the point-topoint communication, where two distant stations connect each other without help of any other infrastructure but the ionospheric layer.To enhance its performance and ability, a modern trend is to build the HF communication network (HFCN) by connecting existing HF infrastructures (e.g., the transceivers) distributed at different locations through the wired network, e.g., the IP networks [4], [5]. HFCN could provide services, such as random access,automatic alternative routing, anti-jamming,anti-destruction, large coverage, and inter-communication with other kinds of networks.Some existing typical HFCNs are the HF2000 communication network in Sweden, the global HF communication system in the U.S. Air Force, the LONGFISH network in Australia,and the integrated HF communication network in Canada [6]-[8].

As well known, HF communication has two major limitations, i.e., poor link quality and low link success rate, resulting from dif ficulty in finding usable frequencies1The usable frequencies constitute the usable frequency window(UFW), with lower and upper bounds defined by the lowest usable frequency (LUF) and maximum usable frequency (MUF), respectively.[1],[2],[9]. On one hand, the HF channel is both time and frequency dispersive, leading to severe fading of received signal. On the other hand, the usable frequency is determined by the characteristics of the ionospheric layer, which is affected by various factors such as sun activity, season,day, time, and location, and is hard to be modelled or predicted. Thus, finding usable frequencies and transmitting on these frequencies in an effective way are the two most important things in HF communication.

To improve link quality and link success rate, huge amount of techniques have been proposed. For example, in finding usable frequencies, prediction models/methods recommended by ITU (like those embedded in the HF prediction software VOACAP and ICEPAC [10], [11]) are used. To improve performance, advanced coding and modulation schemes [12], [13], cooperative transmission[14], and diversity techniques [15]-[19] are proposed. Among those, diversity is a very suitable and promising one to deploy in the HFCN, where, for example, space diversity can be exploited naturally since access nodes are distributed over a large space and could receive multiple copies of signal simultaneously.What’s more, diversity can help alleviate requirement on finding usable frequencies (we will further discuss this issue in Section II).Therefore, we focus on HFCN with diversity in this paper and present its system structure,benefits, challenges, key enabling techniques and future research directions.

The rest of the paper is organized in the following way. In section II, we describe the existing typical HFCN system structure, bene fits by introducing diversity, and challenges especially for large scale deployment of diversity technology. In section III, we propose a system structure with access layer and core layer to fully exploit the bene fits and cope with the challenges based on the Cloud-Radio Access Network (C-RAN) and Software De fined Network (SDN). The key enabling techniques are identified as well for tackling the challenges resulting from deployment of diversity technology in the HFCN, with emphasis on the HF frequency management system. We further point out the future research direction for enhancing performance of HFCN with diversity in Section IV. The conclusion is drawn in Section V.

In this paper, we introduce diversity into the HFCN as performance enhancing technology.

II. SYSTEM STRUCTURE OF EXISTING HFCN

In the section, we first present the system structure of existing HFCN with diversity,and then the bene fits, challenges, and the simulation results on performance enhancement when diversity technology is deployed.

2.1 System structure

The existing typical HFCN system structure is depicted in figure 1, where both the access layer and the core layer are shown. In the access layer, the access node (AN), consisting of several HF antennas and transceivers,transforms the received signal received from the air into digital information in the uplink and transforms the digital information into signal sent over the air in the downlink. Specifically, in the uplink the transceiver converts the signal from the antenna into baseband signal, accomplishes baseband signal processing (including demodulation/modulation and decoding/encoding), and then transforms the decoded information into packets delivered over the wired network. In the downlink, the processing flowchart for the transceiver is in the reverse order. A gateway exists between the access layer and the core layer with function of transforming the data format. In the wired network where the control plane and the data plane of the forwarding device are coupled, the packets are delivered in a traditional way by using, for example, the TCP/IP protocol. The HF frequency management system (HFFMS) is shown as well, which will be discussed in detail in Section III.

To implement diversity technology in such structure, we take the space diversity as an example (shown in the figure). In the uplink, the remote HF user transmits information on one frequency to the HFCN, where HFCN combines the received signals from different paths(here the ANs/transceivers). In the downlink,the information is transmitted by different ANs/transceivers on the same frequency to the user. The task of combining received information from different paths for decoding is done in the diversity processing center (DPC) in the uplink.

Note that (1) different HFCNs have their own specific structural characteristics in the access layer and the core layer. For example,diversity combining can be implemented in the access layer. We only depict a typical HFCN structure in the paper; (2) there are various kinds of diversity techniques other than space diversity, e.g., the frequency, time, and polarization diversity [17]-[19], that can be implemented in the HFCN. For example, the HF user (or AN) can transmit information on different frequencies, in different time slots, or with different polarization ways. However, the space diversity is a natural and resource-saving one since there are many ANs/transceivers distributed over a large space and frequency(or time) diversity wastes resources in essence.

Fig. 1. Existing typical HFCN system structure with diversity.

2.2 Bene fits and challenges with diversity

By introducing diversity technology into HFCN, two major bene fits are obtained as follows.

•Improving received signal quality by combining. This is a widely known fact for diversity technology in mobile communication but is rarely exploited in the HFCN2We note that in most existing HFCNs, the remote user could select an AN with strongest pilot signal for access based on evaluating the pilot signal broadcast by the HFCN. This kind of scheme, seen as selection combining for diversity technology,can also improve the received signal quality.This paper focuses on the more general case where multiple copies of signal from different paths can be combined for further processing.. We note that in some existing HFCNs (e.g., the HF 2000 communication network), there is small scale diversity application in the uplink. As for combining in the HFCN, there are usually two ways, i.e., hard value combining and soft value combining [20], [21]. In the hard value combining scheme, the transmitted signal is decoded with one cyclical redundancy check(CRC) bit indicating success. If there is at least one path successful in decoding, then this path is the output information. Otherwise,the incorrectly decoded signals on various paths are combined further for performance improvement in the DPC. The basic elements of a receiver including filtering at the radio frequency (RF) end, down converter, demodulation, inverse symbol mapping, decoding, and CRC check, is presented in figure 2(a) for hard value combining in the AN. In the soft value combining scheme (shown in figure 2(b) for AN), the baseband signals (i.e., the soft value)are collected together into the DPC, which can further utilize the usable information in the soft value. It’s noted that the hard value combining scheme is easy to be implemented in existing HFCN with only little modification in the transceivers, and the DPC only needs to combine the decoded bit-level signal. To implement the soft value combining scheme,new transceiver should be developed (which will be further discussed later), and the DPC has to process much more soft value signal.

•Improving link success rate (equivalence to less stringent requirement on fre-quency selection).For HF communication,there exists UFW only in which the sky wave transmission is possible. The UFW of a pair of communicating stations depends on factors such as sun activity, season, day, time, and location. When deployed in a large space, the ANs in the HFCN have different UFWs for the same user. This is equivalent to expanding UFW for a user, thus improving the link success rate. Imaging a scenario where there are enough ANs distributed over a large space(e.g., a nation) and the HF user randomly selects a frequency to transmit, it is almost certain that there will be at least one AN succeeding in receiving the user’s signal. Therefore, it is less stringent in selecting appropriate usable frequency for transmission.

Although with such benefits, introducing diversity into HFCN also faces challenges as follows3Almost all these challenges exist for the existing HFCN as well.However, these challenges are more severe when diversity technology is deployed in large scale(e.g., network-wide)..

•How to enhance received signal quality at each path?Although diversity improves signal quality, the combining effect will be better with stronger received signal at each path. On one hand, the distances between the remote user and the ANs are usually on the scale of hundreds of kilometers, not like the cellular communication system where the distances are on the scale of several kilometers.On the other hand, the antenna and power of the user are usually not as that of AN, where large high gain directional antenna and large power transceivers can deploy. With these two factors, the received signal at the AN is weak.Therefore, it’s important to design new methods to improve receiving quality at each path.

•How to access multiple users at AN with reduced cost and power consumption?The existing transceiver is usually expensive and power consuming (such as the widely deployed 400W/125W transceiver system), but with only a 3KHz bandwidth. For multiple users’ access, the traditional way is to add more 3KHz bandwidth transceivers, leading to huge cost and intolerable power consumption. It’s thus necessary to design new transceivers with reduced cost and power.

•How to deliver the received signals forcombining in a timely way?When diversity technology is applied in the whole HFCN,huge numbers of users bring much higher data rate transmission requirement since many copies of same information are delivered in the wired network. It is thus vitally important to design new methods to converge and forward the received signals belonging to the same user in a timely way to the DPC for combining.

•How to manage the HF frequency resources efficiently?The whole HF frequency band (3-30MHz) is narrow and time varying both in its usability and received signal quality. To support network-wide diversity technology deployment, it’s more urgent to first get the usable frequencies and their link quality,and then allocate these frequencies to users efficiently (especially when resource consuming frequency diversity is used).

2.3 Simulation results

To manifest the bene fits brought by diversity technology, simulations are performed here.

Fig. 2. The basic signal processing flowchart of both hard and soft value combining in AN.

First, the simulation is performed on the bit error rate (BER) and frame error rate (FER)performance of both hard and soft value combining schemes. The simulation result is shown in figure 3 for the uplink space diversity scenario. For soft value combining, the performance of selection combining (SC), equal gain combining (EGC), and maximum ratio combining (MRC) are investigated. For hard value combining, the output information is the CRC correctly checked one if there is such one. Otherwise, the frame is delivered to the DPC for further processing, and a frame error claims on the condition that the CRC check still fails after combining. For further processing in hard value combining, three methods are used.The first one is the selecting one (SO) method,where the frame is discarded if all the paths are not checked right. The second one is the large number selection (LNS) method, where the bits of the frame for CRC check are the ones with lager numbers at each same position of all paths. Then the frame is CRC checked.The third one is the exhausted search (ES)method, where the number of inconsistent bits for all paths and their corresponding positions are identified. The number is then compared with a threshold. If it’s less than the threshold,the bits are replaced by combinations of all possible bits and then are CRC checked. Otherwise, the frame is discarded.

The simulation parameters are set as follows. Each transmitted frame includes 45 information bits with 3 zeros padded, and the 12bits CRC check bits are then added forming a frame length of 60 bits. The frame is encoded by a convolutional code with code rate 1/2,constraint length of 9 and generator matrix[1 0 1 1 1 0 0 0 1; 1 1 1 1 0 1 0 1 1]. Viterbi soft decoding is used at the receiver for soft value combining. There are three paths (ANs/transceivers) located at different places in the HFCN. For simplicity but without any effect on manifesting the performance advantages,we adopt independent and identically distributed Rayleigh fading channel models (with zero mean and unit variance) across different paths. It is clear that soft value combining outperforms hard value combining, and the HFCN with diversity outperforms its counterpart without diversity in terms of both BER and FER.

Second, to illustrate the effect of expanded UFW with diversity, we use the ICEPAC software package to predict the LUFs and MUFs of three links in one day. The transmitter is located in Wuhan and the three ANs are in Nanjing, Chongqing, and Haikou, respectively. The LUFs and MUFs of three transmitter-AN links are shown in figure 4, where the corresponding parameters are presented as well. The UFWs of each link and the one with diversity combing are summarized in table I.The result clearly reveals the expanded UFW when diversity technology is deployed.

III. PROPOSED SYSTEM STRUCTURE AND KEY ENABLING TECHNIQUES

Fig. 3. BER and FER performance of hard value combing and soft value combining.

In the section, we first present the proposed system structure of HFCN based on C-RAN and SDN, then the general structure of HFFMS and key enabling techniques, and finally the field experimental results on link success rate.

3.1 Proposed system structure

In the existing HFCN, diversity technology is used only in small scale. Due to lack of wideband transceiver and high-speed low-time latency network in HF, it is hard to implement diversity technology in the whole network.Assuming with ideal wideband transceiver and wired network, the system structure to support network-wide deployment of diversity technology is proposed and shown in figure 5. The structure is divided into the access layer and the core layer with a gateway between, but with different constituent components compared with figure 1.

In the access layer, there are antennas and wideband transceivers in the AN. Different with the traditional narrowband (3KHz) transceiver, the wideband transceiver only includes the RF front end, the A/D and D/A converters,and does the work of transforming between the transmitting/receiving signal from the air and the baseband signal. The baseband signal processing (including demodulation/modulation and decoding/encoding) is moved to the DPC. This is analogy to the C-RAN structure,where the wideband transceiver in the HFCN functions as the remote radio head (RRH)[23],[24]. With such wideband transceiver,multiple users could access the HFCN with low cost and low power consumption.

In the core layer, there are DPC and forwarding devices with decoupled control. The DPC functions as the baseband unit (BBU)pool, where demodulation/modulation and decoding/encoding are done. Although both hard value combining and soft value combining can be used, the latter is usually performed in the DPC of figure 5 because of its performance advantages (shown in figure 3). Moreover,since wideband transceiver is supported, the DPC in figure 5 should be more powerful in computing and data storage.

Fig. 4. LUFs and MUFs of links between one user and three ANs.

Another major difference with the exist-ing HFCN is the wired network structure. In the existing HFCN, the information received from different paths for combining converges to the gateway, and then is delivered by the forwarding devices with coupled control(such as the TCP/IP routers) to the DPC in a distributed way. With such structure, it is hard to ensure the information to arrive at the DPC in a timely way and with tolerable time jitter.To overcome this drawback, we adopt an SDN based wired network with the control plane and data plane decoupled. The SDN operates in a centralized and controllable forwarding way, where the forwarding device (corresponding to the data plane) may be the highspeed high-volume hardware switch and the controller (corresponding to the control plane)has the entire network topology and complete control of the forwarding devices [25].

With such centralized structure, the controller communicates with the forwarding devices with decoupled control via, for example, theOpenFlow protocol, to form the flow tables.To ensure high speed and low time latency delivery of information, each remote user can be mapped into a flow entry in the flow tables,with the match field in the entry corresponding to a specific user. Then, the information packets from different paths of the same user are forwarded in the same way to the corresponding DPC. In such way, the information packets could be delivered in a timely way and with tolerable time jitter. An example for uplink and downlink space diversity and its information flow are shown in figure 5.

Table I. UFWs of links with and without diversity.

Fig. 5. Proposed system structure of HFCN for network-wide deployment of diversity technology.

We present only one DPC to mean that less DPC is needed in figure 1, while multiple DPCs are deployed in figure 5. This is because in existing HFCNs only small scale diversity application with narrowband transceiver is supported, and one DPC will be enough for combining. However, it is not the case for HFCN equipped with wideband transceivers supporting network-wide diversity technology. In such case, the data rates and combining processing burden are much heavier than that of existing HFCN, and it’s thus necessary to deploy multiple DPCs to share the processing task. Moreover, with multiple DPCs, it’s indispensable to develop the SDN based wired network for high speed and low time latency delivery.

It’s noted that due to the sky wave propagation of HF signal, the ANs hundreds kilometers apart may receive the signal transmitted by the same user while this is not possible in cellular communication system. Hence, to combine the signals received from two distant ANs, there is more stringent requirement on time latency and time jitter in the SDN based wired network in the HFCN compared with in the cellular communication system.

3.2 HF frequency management system

When diversity technology is deployed network-wide in the HFCN, it’s more urgent to design efficient HFFMS to allocate resources(e.g., transceiver and frequency), especially when resource consuming (e.g., frequency)diversity technology is deployed. It’s intuitive that with less resources meeting users’ demands, not only resources are saved but also interference is reduced.

The aim of HFFMS is to ensure that the real time users’ demands are evenly matched with HF channels as much as possible. However, the HF channel is very dynamic in terms of the bandwidth available and its use. Therefore, the HFFMS needs to manage the demand for bandwidth in real time as much as possible and spread its traffic load over all the service designated bandwidth with minimum interference to other users, when adapting to congestion (or interference) as well as changing propagating conditions [26].

To achieve the objective, a general structure of HFFMS is presented in figure 6 based on our previous proposed HF network situation awareness system [27]. The main function of HFFMS is to present frequency support schemes (FSS) responding to user’ demands and frequency situation in real time with the support of external input information, including frequency quality and noise information4Frequency quality and noise information mainly includes frequency prediction, real-time probing frequency,historical frequency communication quality,and local noise power information.,SunSpot Number (SSN), GeoMagnetic activity Index (GMI), position information, and equipment state information.

The FSS is a scheme including how many ANs and frequencies should be used to meet users’ demands, e.g., frequency coverage, link success rate and service type5There are two major types of service in HF,i.e., date and voice.There are also many metrics measuring the performance of HF communication, such as BER,FER, link success rate,frequency coverage, etc.In the paper for the FSS,we only take frequency coverage, link success rate and service type for brevity.. It also includes which AN and frequency and what kind of diversity (i.e., space, time, or frequency diversity) should be used.

The HFFMS consists of five modules, i.e.,position information management module(PIMM), network resource management module (NRMM), frequency prediction module(FPM), frequency situation generating module (FSGM), and frequency planning module(FPLM). The functions of the components and the working process to present frequency situation and FSS are summarized as follows.

•PIMM:the module maintains the geographical position information of users and the ANs, where the position information is obtained from the positioning system (e.g., GPS or Beidou Navigation system) or the HF users when they register in the network for the first time. The position has great effect on UFW and link quality.

•NRMM: the module maintains the equipment information of antenna, transceiver, and users in the network. Information includes antenna type, antenna gain, transit and receiver power, users’ ability, etc.

•FPM: the module produces the long term and/or short term predicted UFWs and their link loss with the support of PIMM and NRMM. When the SSN/GMI is available,the prediction process is straightforward with mature software like VOACAP or with some neural network models [28], [29]. When the SSN is unavailable, the frequency quality from limited number of ANs and users can help in predicting UFWs [30], [31]. The noise power information is at last added to form the link quality measured in SNR.

•FSGM: the module collects all the information at hand to generate the frequency situation of the whole network (for example with a method named two-dimensional interpolation and extrapolation in the space and time dimensions). Specifically, the link quality of limited links is used as baseline. By inserting in the space, the link quality of other links without real-time or historical quality data is obtained. In the time dimension, future usable frequencies or link quality can be predicted based on real-time or historical quality data.

Fig. 6. General structure of HFFMS.

•FPLM:with the users’ demands and the frequency situation at hand, it is possible to produce the FSS, and this is usually modeled as a multi-objective optimization problem with multiple constraints [32].

It’s noted that the frequency situation is dynamically updated when input information changes. Thus, it can adapt to congestion as well as changing propagating conditions, and work in an intelligent way to ful fill users’ demands in real time as much as possible.

3.3 Key enabling techniques

Although there is small scale application of diversity technology, it’s necessary to develop new techniques to fully exploit the benefits and overcome the challenges before diversity technology can be deployed in the HFCN in network-wide scale. Here, major key techniques are listed as follows.

•Wideband, omnidirectional, and high gain antenna design.High gain antenna plays an important role in improving received signal quality. However, the antenna is often directional and huge (e.g., the dipole antenna or the log-periodic antenna). It’s thus hard to direct the radiated signals to and receive the signals from multiple remote users simultaneously with high gain antenna. Additionally, when wideband transceiver is deployed, it’s also required that the antenna operate in the wide frequency band (best in the whole band) without performance loss, such as the nonlinearity loss, in some bands. Wang et al. has proposed to form an omnidirectional and high gain antenna by placing three twenty meters long dipole antennas reasonably [7]. However, it is still hard to design wideband, omnidirectional,and high gain antenna design with small size.An alternative is to use directional and high gain antenna to realize omnidirectional coverage by mechanical rotating or phase control.However, this method needs external position information support.

•Wideband transceiver design.Standard like MIL-STD-188-110C has specified the wideband HF transmission up to 24KHz with continuous spectrum [33]. There is also research on wideband transmission with non-continuous spectrum [34]. Efforts have been made in design more wider (even the whole HF band) transceivers as well [35].However, it is difficult to design practical wideband transmitter under out of band emission constraint for non-continuous spectrum,due to the limitation of amplifier and oscillator [36]. It is still unknown whether there are practical amplifier and oscillator supporting transmitter with bandwidth larger than 24KHz.

•High speed and low time latency wired network design.The SDN structure with decoupled control and data plane is suitable in realizing such network in data center [37].By using hardware switch and a controller with network-wide control, the packets received from multiple paths can be delivered in a timely way to the DPC. However, further research on time delay and time jitter should be done to fully enjoy the benefits of SDN since the distance of ANs (and thus the delay) is much larger than its data center counterpart.

•Efficient HF link quality prediction.Obtaining the usable frequencies and their link quality is the most challenging thing in HF communication. In the HFCN, it’s not possible to have the link quality for users and ANs at any arbitrary position in real time. Therefore,it’s needed to monitor and then predict the HF link quality, including both the link path loss and noise power. At the user side, cognitive radio concept has been applied in the HF transceiver for noise monitoring and prediction [38]-[40]. At the network side, predicting the link quality, both long term and short term,of users at any arbitrary position is of vital importance [41]-[44]. When diversity (especially the frequency diversity) is introduced, efficient prediction is more urgent in selecting usable frequencies for transmissions, allocating limited available frequency resources to more users, and limiting interference to other systems.

3.4 Field Experimental Results

We conduct a field experimental test to show the improvement of link success rate when diversity technology is deployed in the HFCN.The tested HFCN consists of three ANs located in Changsha, Nanning, and Guangzhou,and four users located at Chengdu, Xi’an,and Guangzhou (with two users denoted by Gz1 and Gz2), respectively. All the received signals from ANs are converged to the DPC deployed in Guangzhou for combining. The prototype wideband transceiver we developed in the test covers the whole HF band but could decode at most eight 3KHz narrowband signals simultaneously due to hardware limitation. Three kinds of service, i.e., short message service (SMS), voice SMS, and real time voice, are supported, all with a transmission rate of 600bps. In the uplink, space diversity is used when frequency diversity is employed in the downlink (with different ANs transmitting on different frequencies to the same user). The test is conducted in days and nights during March 28th to April 16th, 2017. In each trial,the lengths of one SMS and one voice SMS are 60 bytes and 700 bytes, respectively, and the real time voice lasts about 15 seconds. Figure 7 summarizes the test result, where the total number of trials for each kind of service is listed at the top right corner of each sub figure.The link success rate is defined as the ratio between the number of trials successful in the uplink and downlink simultaneously and the total number of trials. It’s obvious that the link success rate with diversity can be improved to be above 90%. We note that the link success rate for HF communication without diversity is usually around 50%-80% at best. Therefore,diversity is very promising in improving link success rate performance.

IV. FUTURE RESEARCH DIRECTIONS

To further enhance performance of HFCN with diversity and to make it work more efficiently and intelligently, we believe the following directions deserve further research.

•High performance RF component design.The design of wideband, omnidirectional, and high gain antenna with small size is very important, especially at the user side,where the large antenna is impossible to deploy. Furthermore, amplifier with high range linearity and high stable oscillator are needed to limit the intermodulation interference and phase noise, to make the radiated out of band interference as small as possible.

•AN and DPC deployment position planning.Unlike in the cellular communication system, the ANs in the HFCN are far apart and the positions of ANs will in fluence the usable frequencies and link quality. It is therefore more necessary to plan the AN deployment positon reasonably, according to the users’demands. Moreover, the positions of DPCs should be planed accordingly as well since their positions have huge effect on the hops(and thus the time delay and jitter) of informa-tion delivery in the wired network. It’s pointed out that the AN and DPC deployment position plan is easier in scenarios where the users’ positions and demands are relatively fixed.

Fig. 7. Experimental link success rate with diversity.

•Virtualization technology and network slicing.Although position plan can help in receiving, it’s impossible to meet all the users’ demands when considering factors like randomness of use’s access, randomness of user’s transmission requirement, uncertainty of user’s position distribution, and unknown number and positons of ANs in receiving the signals of the same user. To overcome such limitations, it is suitable to utilize virtualization technology [45]. By treating the ANs and DPCs as network resources, it’s possible to meet the varying demands of different groups of users through dynamic resource allocation and flexible scheduling in the network slicing way.

•Intelligent frequency prediction and allocation.Although existing HF frequency (long term and/or short term) prediction models/methods [46]-[51] could predict the link quality to some extend of accuracy, the prediction models/methods are not strong enough to utilize all the available information including prediction information, real time probing frequency information, historical frequency quality information, and local noise power information. On the other hand, the predictability depends on the already obtained data (available information) in the entropy view [52] and/or the correlation in the time,frequency and space dimensions [53]-[57].New intelligent models/methods for frequency prediction and allocation [58],[59], especially those incorporating deep learning [60] and entropy analysis, should be developed to fully utilize the available information in a comprehensive way and with deep mining. With accurate prediction, it’s possible to allocate the frequency in an efficient way to meet users’demands with less resource and at the same time limit the interference to other users.

V. CONCLUSION

In this paper, we introduce diversity into the HFCN as performance enhancing technology.Its benefits (improving link quality and link success rate) and challenges are presented first. The existing HFCN system structure is shown to be unable to support large scale diversity application. Therefore, we propose a new system structure of HFCN suitable for large scale deployment of diversity technology to fully exploit the bene fits and overcome the challenges. The proposed system structure refers to the C-RAN and SDN concepts in cellular communication system. Moreover, we present a general structure of HFFMS that can adapt to congestion and changing propagating conditions in real time as much as possible,which is more necessary when resource consuming (e.g., frequency) diversity technology is deployed. The key techniques enabling network-wide deployment of diversity technology are identified as well.

Both simulation and field experimental results confirm the performance advantages of diversity technology. We believe that the HFCN with diversity is a very promising technique for enhancing HF communication performance with the support of efficiently designed HFFMS. We also point out the future research directions for further performance enhancement.

ACKNOWLEDGEMENT

The work is supported by the National Science Foundation of China under Grants No.61801492 and No. 61601490, and a national major specific project governed by the national development and reform commission of China. The authors acknowledge Musheng Li,Xiaotian Chen, Jun Lu, Yongtao Cao, Qifeng Guo, Xufu Zhang, and Zhehai Yang from Guangzhou Haige Communications Group Incorporated Company for their help in executing the field experimental test and recording the test data.