Abstract

We study a radio resource allocation problem in mobile communication systems. As the distinct characteristic of this problem, a common data transmission rate is used on all channels allocated to a user. Because the channels differ in their quality, for each user the achievable rate varies by channel. Thus allocating more channels does not necessarily increase the total rate, as the common rate is constrained to be the lowest one supported by the allocated channels. Radio resource allocation subject to the common-rate constraint is of practical relevance, though little attention has been paid to modeling and solving the problem. We take a mathematical optimization perspective with focus on modeling. We first provide a complexity analysis. Next, several integer linear programming (ILP) formulations for the problem, including compact as well as non-compact models, are derived. The bulk of our analysis consists in a rigorous comparative study of their linear programming (LP) relaxations, to reveal the relationship between the formulations in terms of bounding. Computational experiments are presented to illustrate the numerical performance in bounding and LP-assisted problem solving. Our theoretical analysis and numerical results together serve the aim of setting a ground for the next step of developing model-based and tailored optimization methods.

Introduction

Mobile communication systems have been evolving rapidly in the past decades. From an optimization standpoint, a key research topic for mobile networks is radio resource management for utilizing the radio resource efficiently, particularly because the amount of data traffic is constantly growing in the current 4th generation (4G) systems, and the trend will remain or even accelerate in the 5th generation (5G) networks. Along with this growth, 5G networks target a plethora of new services (e.g., vehicular communications), making resource management more challenging than before, as pointed out in Calabrese et al. (2018).

4G/5G networks deploy orthogonal frequency division multiple access (OFDMA). In essence, data transmission takes place in two dimensions, namely time and frequency that are divided into time slots and channels, respectively.1 Resource management for channel allocation in a time slot, as well as that across multiple time slots, is commonly called scheduling. In any time slot, a user may be scheduled to receive transmission from its base station (BS) on multiple channels, and a channel may be scheduled for at most one user of this BS. The latter is what the term “orthogonal” refers to in OFDMA. Among the BSs, the channels are reused, i.e., there is generally no dedicated (sub)set of channels for a BS. As a result, a channel is exposed to interference, if the channel is also scheduled in some nearby BSs in the same time slot. Note that even if interference is not present, the channels are still frequency-selective (Tse and Viswanath, 2005), meaning that they vary in quality and hence the achievable data rate from a user’s perspective.

Conclusions and future work

We have studied mathematical modeling of a resource optimization problem with common-rate channel allocation in mobile communication systems. We have demonstrated that the problem admits an array of integer programming formulations. The study shows that modeling does matter, analytically as well as numerically, in terms of performance in bounding and LP-based problem solving. Moreover, there is a clear correlation between the accuracy of LP and the quality of integer solution derived thereby.

As we have observed, straightforward schemes of constructing integer solutions are not practical in solution time, whereas for our resource allocation problem, the ultimate target is real-time optimization. To approach this goal with model-based optimization, we make several observations. First, to deliver a solution rapidly, the amount of computation has to be very small, and thus the LP relaxation may have to be solved approximately, followed by some primal heuristic (both admitting massive parallel computation). Second, non-compact models are also of interest, as restricting the number of columns or rows leads to approximation of the problem (and the LP relaxation). Finally, with presence of rate lower bound, obtaining solution feasibility is of clear significance, and one may need to trade solution quality against feasibility. Developing and implementing algorithmic notions along these lines form interesting topics for forthcoming research.

In the presented work, the performance analysis of an underlay cognitive relay network with a hybrid-decode/amplifyand-forward relaying protocol is undertaken. Since the system is an underlay system, the secondary network will communicate simultaneously with the primary network. Hence interference temperature will act as a constraint to keep the secondary communication from interfering with the primary transmission. To ensure this, secondary transmit powers need to be kept below a pre-defined threshold level. Severe multi-path fading and shadowing or a long distance between the secondary source and the destination make a direct link between the two infeasible. These nodes can only be linked by a relay between the two. All the radio links in the system are characterized by the extended generalized-K fading distribution. The closed form expressions for the outage probability and ergodic channel capacity of the system are derived. The system outage performance and capacity with hybrid-decode/amplify-and-forward relay is compared to the performance with amplify-and-forward and decode-and-forward relays. Optimum relay position for best outage performance is also investigated. Further, the results of the performance analysis of the system are also illustrated by the numerical plots.

Introduction

Due to the increased data applications and services worldwide, global data traffic and volume has increased exponentially, giving rise to spectrum scarcity. This scarcity is more due to the under utilization of the available spectrum than due to actual lack of it. As a remedy to this problem, the cognitive radio networks came into existence in late twentieth century [1],[2]. These networks provided a means to alleviate the problem of spectrum scarcity. The said networks constitute multiple user networks that enable unlicensed (secondary) users to access a channel belonging to a licensed (primary) user. The access may be through one of the three laid down methods viz. overlay, underlay or interweave [3]. The channel has to be free before the secondary user (SU) can access it in overlay system, [4]. The SU has to vacate the channel if the primary user (PU) needs access to the channel as the PU will always have a higher priority. Underlay systems can give both PU and SU access to the channel simultaneously as in [5], provided the SU does not interfere with the PU signal while communicating. For this the powers transmitted by the secondary nodes need to be kept below the pre-defined threshold values. The secondary user in the interweave systems, employs some spectrum awareness methods to explore and then opportunistically utilize the unused spectrum, also called the spectrum holes as explained in [6]. In this work, the ultra wideband antenna configuration is used for spectrum sensing and the reconfigurable antenna is used for communication for secondary transmission on the radio channel.

Abstract
This paper analyzes the economic feasibility of a market for data-based services, which is a scenario envisioned in sectors that are adopting Internet of Things and/or Big Data technologies. A business model for the provision of a data-based service is proposed. In this model, service providers build services from data that is collected and sold by data providers in a market, and provide services to final users. Service providers compete strategically against each other in the data market and in the service market. The strategic interaction between the service providers is analyzed by means of non-cooperative game theory. A thorough mathematical analysis of the Nash equilibria is carried out and existence and uniqueness conditions are proved and discussed.

We conclude that if the sensitivity of users to the data-rate-to-price ratio is above the unity and the number of service provider does not exceed a given limit, a unique and meaningful equilibrium exists. We also conclude that the intensity of competition is beneficial for the users and detrimental for the service providers, while for the data providers, maximum surplus is obtained in an intermediate intensity of competition. Finally, an increase in the users’ sensitivity to the data-rate-to-price ratio causes a shift on the surplus from both the service providers and data providers to the users.

1- Introduction

2- Experiment

3- Results and discussion

4- Conclusions

References

Abstract

In this paper, we present a detailed investigation of low frequency noise (LFN) for different n-MOSFET devices dedicated for memory applications. We investigate the impact of the gate oxide thickness (GOX) on LFN. We analyzed how the position, the existence and the composition of Lightly Doped Dopant (LDD) implant in the source/drain region affect the LFN performance of the device. The results demonstrates that the thinner gate oxide and the device without LDD improved the noise performance compared the devices with thick GOX and with LDD implants. On the other hand, the absence of LDD implant on one side of the MOSFET didn’t reveal a global trend for all measured devices. Finally, the different LDD implant composition resulted in different LFN performance which is gate area dependent. These results can be used from both process and design engineers to improve the LFN of n-MOSFET.

Introduction

Today, it is well known that LFN can be used as a characterization tool for the quality and the reliability of the devices [1-2]. It is obvious that the level of noise has a direct impact on the device quality. In addition, LFN can influence the design of an electronic circuit since it can limit the overall performance and operation. Thus, the noise investigation of special architecture devices is of paramount importance for device and design engineers. Single transistors designed for logic NVM of SONOS type memory applications [3] can be used for LFN comprehension analysis due to the special architecture they have. In this experiment five single cell n-MOSFET devices were designed for memory applications and used for noise analysis. In literature exists a number of publications referring to LFN characterization for logic NVM of SONOS type [4-7]. According to our knowledge, we could not find a prior work that used these type of devices for LFN improvement purposes and not for simple characterization. We tried not only to identify the main source of LFN in these devices but also to use this info for the improvement of LFN performance in CMOS technology. In MOSFETs, it is generally accepted that the flicker (1/f-like) noise originates either from carrier number fluctuations (CNF) (Eques 1, 2 Table 1) [1] or from Hooge mobility fluctuations [8]. The CNF noise is due to carrier exchange between the near-interface gate dielectric traps and the channel. The charge fluctuations in the gate dielectric could also induce fluctuations of the carrier mobility, giving rise to the so-called correlated mobility fluctuations (CMF) (Eques 3, 4 Table 1) [9-11].

I- Introduction

II- Ofets Fabrication Process

III- DC Characterization

IV- AC Characterization

V- Conclusions

References

Abstract

Thanks to recent progress in terms of materials properties, polymer field-effect transistors (FETs) operating in the MHz range can be achieved. However, further development toward challenging frequency ranges, for a field accustomed to slow electronic devices, has to be addressed with suitable device design and measurements methodologies. In this letter, we report n-type FETs based on a solution-processed polymer semiconductor where the critical features have been realized by a large-area compatible direct-writing technique, allowing to obtain a maximum frequency of transition of 19 MHz, as measured by means of ScatteringParameters (S-Parameters).This is the first report of solution-processed organic FETs characterized with SParameters.

INTRODUCTION

ORGANIC electronics underwent an impressive development and thanks to its peculiar properties, e.g. high degree of tunability of electronic properties and thin molecular films deposition with scalable processes at low temperature on a wide variety of substrates, represent one of the most promising candidates in the path to ubiquitous, flexible and lightweight low-cost electronics. To date, organic light emitting diodes (OLEDs) [1] have reached commercial maturity, and significant progress has been achieved for organic solar cells [2], photodetectors [3], biosensors [4] and circuits based on organic transistors [5]. In the case of field effect transistors (FETs), the use of polymer semiconductors has the advantage of achieving superior mechanical properties and easily enabling large-area fabrication techniques, such as printing. Excellent DC performances have been recently achieved owing to an improvement in the understanding of device physics [6], and steady increase in charge carrier mobility [7] and charge carrier injection [8]. Yet, progresses in DC performances are not usually matched by corresponding improvements in AC properties, since typical FET structures devised to test new polymer semiconductors are not optimized for frequency operation. It is instead highly demanded to enhance polymer FET maximum operational frequency, as this would enable a much broader application of this technology. For example, higher operational frequencies are crucial in drivers for high resolution flexible displays [9], radio frequency identification (RFID) systems for smart items identification [10] or wireless communication [11].

I. INTRODUCTION

II. PRELIMINARIES AND SYSTEM MODEL

III. SIMULATIONS

IV. FUTURE RESEARCH DIRECTIONS

V. CONCLUSION

REFERENCES

Being increasingly spectral and energy efficient, massive multiple-input multiple-output (MIMO) is envisaged as a potential technology for fifth generation (5G) wireless communication networks. Radio spectrum has become a scarce resource in wireless communications and consequently imposes excessive cost on the high data rate transmission. Several linear and non linear detection techniques such as Zero-Forcing (ZF), Minimum Mean Square Error (MMSE), and Vertical Bell-Labs Layered Space Time (VBLAST) have been introduced. The purpose of such schemes is to mitigate the signal detection problems which are based on trade-offs between the bit error rate (BER) performance and computational complexities. The challenge in the design of massive-MIMO systems is developing less complex and efficient detection algorithms. The problem in building a receiver for massive-MIMO is to de-correlate the spatial signatures on the receiver antenna array. In this paper, we propose a novel algorithm viz: Hybrid n-Bit Heuristic AssistedVBLAST (HHAV) to perform an optimum decoding. We have simulated this structure in dynamic Rayleigh fading channel. We have also evaluated the AMP algorithm with two threshold functions which include AMP with ternary distribution (AMPT) and AMP with Gaussian distribution (AMPG). Numerical results confirm that HHAV algorithm performs significantly better than the in vogue aforementioned detection systems as introduced in recent years.

INTRODUCTION

Nowadays in wireless communication systems, massive MIMO has become the subject of an extensive research due to its potential to support higher data rate as compared to their counterpart single-input single-output (SISO) [1]–[3]. Hence, massive-MIMO systems with tens to hundreds of antennas at the base-station (BS) have gained the attention of researchers [2], [4]–[11]. A few hurdles, however, remain to be overcome to achieve optimum advantages offered by massive MIMO systems. One of them is to model a less computational and reliable detection system without compromising performance. The motivation to consider massive MIMO systems is their potential to meet the growing demands for higher throughput as the volume, velocity, and variety of data due to exponential augmentation in mobile users and communication networks [12]. The ever-increasing number of users in various mobile networks along with the additional enabling connectivity of the mobile services [13], [14], has necessitated the 5G technologies [15]. The aim of 5G is to provide the much-needed capacity, increased data rates, reduced latency, and improved quality of service as compared to previous technologies. In massive-MIMO, detection is a complicated process. It requires an extensive search over the space in order to find the closest received symbols in terms of euclidean distances for an optimal solution. However, these procedures are rendered unfeasible for larger systems such as massive MIMO systems.

Inspired by their tremendous success in optical image detection and classification, convolutional neural networks (CNNs) have recently been used in synthetic aperture radar automatic target recognition (SAR-ATR). Although CNN-based methods can achieve excellent recognition performance, it is difficult to collect a large number of real SAR images available for training. In this paper, we introduce simulated SAR data to alleviate the problem of insufficient training data. To address domain shift and task transfer problems caused by differences between simulated and real data, we propose a model that integrates meta-learning and adversarial domain adaptation. We use sufficient simulated data and a few real data to pre-train the model. After fine-tuning, the pre-trained model can quickly adapt to new tasks in real data. Extensive experimental results obtained in the real SAR dataset demonstrate that our model effectively solves the cross-domain and cross-task transfer problem. Compared with conventional SAR-ATR methods, the proposed model can achieve better recognition performance with a small amount of training data.

Introduction

Synthetic aperture radar (SAR) is an active sensor mounted on moving platforms such as aircraft, satellites, and spaceships. SAR provides two-dimensional high-resolution images by receiving the electromagnetic echoes of targets. Benefiting from its unique imaging mechanism, SAR can operate day and night, independent of weather conditions, and has specific surface penetration capability. The SAR system has unique advantages in many applications, ranging from disaster monitoring and resource exploration to military inspection, and it plays an unreplaceable role in both military and civilian fields. Automatic target recognition (ATR) is an essential topic in the field of SAR application research. According to different implementation methods, classic ATR methods can be classified into feature-based and model-based approaches. Feature-based methods extract discriminative features, such as binary regions [1], target contours [2], monogenic signals [3], [4], projection features [5], [6], and tensor decomposition features [7] from images. Classifiers such as K-nearest neighbor (KNN) [8], support vector machine (SVM) [9], the Bayesian classifier [10], and the sparse representation classifier [11] have been developed to classify the extracted features. Both feature extraction and classification require careful selection by experienced researchers. Model-based methods [12]–[14] focus on the electromagnetic scattering features of a target, which are related to the physical characteristics of the target.

A reconstruction technique of the antenna pattern for correcting the polarization measurement bias is proposed in this paper. In the case of the prior knowledge of patterns at some scanning angles, the pattern at other beam position can be estimated by application of the minimum norm least-square solution. Once the pattern is reconstructed, the corresponding projection matrix could be obtained in the presence of the mutual coupling and edge effects in the polarimetric phased array antenna. Thereby, the correction matrix for the polarization scattering matrix is repaired. A microstrip patch array antenna is designed to demonstrate the validity of the proposed method, and the performance of bias correction is manifested using several polarimetric variables. Specifically, the bias of differential reflectivity (Z b DR), the bias of linear depolarization ratio (L b DR) and the phase difference are less than 0.14 dB, 46.85 dB, 3.31◦ in a wide scanning angle range, respectively.

Introduction

Phased array antenna (PAA) with agile electronic beam steering capability has been successfully utilized in the communication, military application and weather observation [1]–[6], with respect to the mechanically scanning dish antenna. In addition, since the accurate multiparameter measurement ability of the polarimetry has matured for years, the polarimetric PAA (PPAA) with the dual-polarization capability will be indispensable in future targets or weather observation [7]. Currently, two main challenges that still trouble the PPAA technology are the requirements of wide angle scan and high accuracy for polarimetric measurements [8], [9]. Specifically, when the beam is scanned away from the antenna’s principal plane, the beam will broaden considerably, and the electric fields radiated from the horizontal (H) and vertical (V) polarization ports are not orthogonal to each other. The cross-polar components are produced, and this is where the polarization biases yield. Hence, the polarization isolation requirement should be met for alternate mode (where the polarization states are switched alternately on transmission) and hybrid mode (where both polarization states are transmitted and received simultaneously). And the polarization purity requirement is more stringent for the hybrid mode compared to the alternate mode of operation [10]. To solve the cross-polar problem, the projection matrix method was proposed in [3], and the method was examined with a pair of crossed dipoles. The biases can be removed through a multiplication operation to the measured scattering matrix.

Directional modulation (DM) as a physical layer security technique has been studied from many different aspects recently. Normally all existing designs based on antenna arrays lead to varying weight coefficients for different antennas and for different signal symbols, which poses a particular challenge for feed circuits design in analogue implementation. In this paper, to reduce the implementation complexity, a constant magnitude constraint is proposed for the first time, and the resultant non-convex constraint can be modified to a convex form so that the problem can be solved conveniently by existing convex optimisation toolboxes. Design examples are provided to show the effectiveness of the proposed design.

Introduction

With the fast development of the Fifth Generation (5G) network, communication has been more important than ever [1], [2]. As physical layer security technique to keep known constellation mappings in a desired direction or directions, while scrambling them for the remaining ones, directional modulation (DM) has been studied widely [3]. In [4], a four-element reconfigurable antenna array was designed, where the DM design was achieved by changing elements for each symbol. In [5], phased antenna array was employed for DM, with an individual tailor-made feed circuit (including phase shift and amplitude change) prepared for each antenna. Compared with a given antenna array design, to further reduce the number of antennas, DM design was extended to sparse antenna arrays [6]. To overcome the inherent limitation of DM, where eavesdroppers and the desired users share the same received signal when they are in the same spatial direction of the antenna array, two positional modulation (PD) designs were proposed, with one based on a reflecting surface [7], and the other employing multiple antenna arrays [8]. Both designs are based on the idea that if the received signal is a combination of signals from different paths with different directions, then signals at these locations can be distinguished. To increase the channel capacity of DM, two antenna array structures were proposed recently.

Generalized frequency division multiplexing (GFDM) is a block-based non-orthogonal multicarrier modulation scheme proposed for 5G PHY layer. In this paper, to efficiently coordinate the features of interference existing in the GFDM system, we propose a theoretical framework of the polar-coded GFDM (PC-GFDM) system, which allows jointly optimizing the combination of binary polar coding and GFDM modulation. The original GFDM channel is decomposed into multiple bit polarized channels by using a two-stage channel transform. The general modulation partition is performed in the first stage and the bit polarization transform is done in the second stage. Specifically, two schemes are considered for the first stage channel transform, multilevel coding (MLC) and bit-interleaved coded modulation (BICM). Based on the theorem we have proved in this paper that the capacities of channels corresponding to all transmitted symbols are identical, the MLC based PC-GFDM system is designed to optimize the combining of polar codes and GFDM system. Then the BICM based PC-GFDM system is designed to reduce the complexity and processing latency, which yields the suboptimal performance. Simulation results indicate that the proposed PC-GFDM systems significantly outperform the existing turbo-coded GFDM systems because of the joint design between the polar coding and the GFDM modulation.

Introduction

Polar codes proposed by Arıkan [1] are the first constructive codes that have been proved to achieve the symmetric capacity for the binary-input memoryless channels (BMCs). This capacity-achieving code family is based on a technique named channel polarization, which recently has been employed in many scenarios for signal processing. In 2013, Seidl et al. have introduced the channel polarization idea into the 2m-ary modulation scheme [2] and proposed two different polar-coded modulation (PCM) schemes, where the bit-interleaved coded modulation (BICM) [3] and multilevel coding (MLC) approaches are considered respectively. Later, by extending the channel polarization idea to non-orthogonal multiple access (NOMA) scenario, whereby the original NOMA channel is split into a set of binary-input channels under a three-stage channel transform, Dai et al design a polar-coded NOMA system [4] to jointly optimize the combination of binary polar coding, signal modulation and NOMA transmission. Similarly, polar-coded multiple-input multiple-output (MIMO) schemes [5] also have been developed to improve the system performance.

A periodic fixed-frequency staggered line leaky wave antenna (LWA) with wide-range beam scanning capacity is proposed in this paper. The reconfigurable antenna is based on a periodic staggered line LWA. The dispersion and Bloch characteristics of the periodic antenna is analyzed by the Macro Cell Method (MCM). The open-stop band (OSB) of the antenna can be suppressed effectively using the relationship between widths of oblique feeding lines and stubs. The reconfigurable characteristic is achieved by switches and microstrip lines between the edge gaps of each unit cell. A ‘‘supercell’’ is established by combining several reconfigurable unit cells, and the state is controlled by connecting or disconnecting the edge gaps of each unit cell using switches. Fixed-frequency beam-scanning capacity is implemented due to different propagation constants in different supercell states. The prototyped reconfigurable antenna can scan the beam between 130◦ and 54◦ at 4.5 GHz.

Introduction

The microstrip leaky wave antenna (MLWA) has attracted significant research attention since being developed in 1979 [1] due to unique advantages, such as simple structure, low profile, and beam-scanning capacity in the H-plane [2], [3]. MLWA designs with backward-to-forward beam-scanning capacity using the periodic waveguide structure have been reported [4]. Various periodic antenna designs have been developed, such as the periodic halfwidth MLWA [5], periodic triangle truncated DSPSL-based antenna [6], and a new type of circularly polarized printed periodic LWA structure [7]. However, the present of an open-stop band (OSB) limits the radiation of periodic structures in the broadside direction, causing a degraded radiation pattern and high return loss [8], [9]. Several methods have been proposed to suppress the OSB in periodic structures. The balanced condition has been confirmed as a useful condition in designing the CRLH LWA to reduce the OSB [10]. An OSB elimination technique, equal phase shift condition, for a class of periodic LWA with two types of uniform transmission lines in alternatively cascading arrangement is proposed in [11]. The beam scanning capacity of the MLWA depends on variations in frequency [12]. Most communication systems operate in the predefined frequency; thus, scanning at a selected frequency is highly desirable [13], [14]. One method is to manipulate the phase constant of the leaky micro-slotline using reactive loading across the slotline [15].

In this paper a low profile circularly polarized annular ring slot antenna (CPARSA) with high gain is presented. An inverted U-shaped meandered slot is cut at the inner patch of annular ring to obtain the circularly polarized (CP) radiation. This annular slot loaded ground plane on a substrate thickness of 0.5 mm results in a very low profile (0.009λ0) antenna in term of free space wavelength at its center operating frequency. The largest side lengths of octagonal shaped antenna are 0.89λ0×0.89λ0 according to its center frequency. Measured impedance bandwidth of 23.54% from 5.4 GHz – 7.0 GHz (for |S11| ≤ −10 dB) and AR bandwidth (AR<3dB) of 5.15% from 5.67 GHz – 5.97 GHz are achieved by the antenna. Meanwhile, stable radiation pattern is observed throughout the entire bandwidth with peak gain of 7.6dBi at the center frequency of 5.8 GHz. A defected ground plane is employed in the design which results in reduction of surface waves leading towards higher gain of antenna. The proposed antenna operating at 5.8 GHz band is fabricated on low cost substrate with standard printed circuit board (PCB) process. The proposed design provides favorable characteristics for various lightweight applications including miniaturized drones and airborne applications. To validate the prototype, measured results are compared with simulation outcomes and good agreement is obtained.

Introduction

Circularly polarized (CP) and high gain antennas with their favorable characteristics such as low profile, less massy and planner structure that offer ease in integration with monolithic IC’s are demanded recently. CP antennas are also favorable in situations where polarization mismatch may degrade the performance of communication system [1], [2]. Advantages of circular polarization over linear polarization are generally because of high efficiency in mobility, alleviate in multi-path distortion and polarization mismatch losses [3]–[5]. Moreover, CP antennas have influence to overcome multipath fading and mitigate Farady rotation effects [6] that can help to reduce the loss. These characteristics make them favorable candidate for wireless power transfer applications. Circular polarization requires generation of two equal but orthogonal radiating modes. This can be achieved either by employing multiple feeds or a single feed. Several CPARSA designs producing circular polarization have been reported in literature [5]–[13]. A circularly polarized antenna based on fractal geometry architect was reported for UHF RFID applications. The measured gain of the antenna is 1.35 dBi [5]. Another design based on metallic ground reflector was presented in [6] for dedicated short range communication (DSRC) applications. Although the antenna retains peak gain at 4.95 dBic but its 6.926 mm dense profile limits the use of antenna for various lightweight applications.

Unmanned aerial vehicles (UAVs) can be used as low-altitude flight base stations to satisfy the coverage requirements of wireless users in various scenarios. In practical applications, since the transmitted power and energy resources of the UAVs are limited and the propagation environments are complicated and time-variant, it is challenging to control a group of UAVs to ensure coverage performance while preserving the connectivity and safety of the UAV networks. To this end, a two-step environment-learningbased method is proposed for the intelligent deployment of the UAVs. First, a machine learning algorithm is used to establish an accurate prediction model of the link qualities from the UAVs to the users under a specific scenario for the next step. Then, a modified deep deterministic policy gradient (DDPG) algorithm is employed to control the movements of the UAVs according to the predicted link qualities and to maximize the proportion of covered users. The prioritized experience replay mechanism is introduced to the standard DDPG algorithm to accelerate the deployment procedure. The coverage performance is analyzed in both the interference-free situation and the situation with co-channel interference. Simulation results have shown that the proposed method has a higher convergence speed than the standard DDPG method. Additionally, the proposed deployment method can achieve higher coverage performance and better adaptability to the dynamic environment than three commonly used methods, the random method, the K-means-based method, and the statistical-channel-model-based method.

Introduction

In recent years, unmanned aerial vehicles (UAVs) have attracted great attention due to small size, low price, and high flexibility. The UAVs with variable positions can establish line-of-sight (LoS) communication links to the users. Therefore, the UAVs are suitable to be used as low-altitude flight base stations (BSs) to reduce the signal attenuation and improve the coverage performance [1], [2]. For example, in the case of a terrestrial BS failure, the UAV-BSs can be rapidly deployed to satisfy temporary coverage demands for wireless services [3]–[5]. The cellular networks can also be assisted by the UAV-BSs in the temporary hotspot area [6]–[9]. The deployment problem of the UAVs is more complicated than that of the terrestrial BSs. First, in practical application scenarios, due to the limited transmitted power and energy resources, multiple UAVs are often required to be deployed together to ensure the large coverage. Second, since the propagation environment is complex and changeable, the UAVs are expected to have certain adaptability to the environment to rapidly satisfy the coverage requirements of the users. In addition, multiple UAVs should be set with a certain distance limitation to maintain both the connectivity to ensure the robustness of the network and the security to prevent collisions caused by unexpected situations. Therefore, how to effectively deploy the positions of the UAVs to improve the coverage performance is a challenging problem.