Pattern Division Multiple AccessâA Novel Nonorthogonal Multiple

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 66, NO. 4, APRIL 2017

3185

Pattern Division Multiple Access—A Novel Nonorthogonal Multiple Access for Fifth-Generation Radio Networks Shanzhi Chen, Senior Member, IEEE, Bin Ren, Qiubin Gao, Shaoli Kang, Shaohui Sun, and Kai Niu, Member, IEEE

Abstract—In this paper, pattern division multiple access (PDMA), which is a novel nonorthogonal multiple access scheme, is proposed for fifth-generation (5G) radio networks. The PDMA pattern defines the mapping of transmitted data to a resource group that can consist of time, frequency, and spatial resources or any combination of these resources. The pattern is introduced to differentiate signals of users sharing the same resources, and the pattern is designed with disparate diversity order and sparsity so that PDMA can take the advantage of the joint design of transmitter and receiver to improve system performance while maintaining detection complexity to a reasonable level. System level simulation results show that PDMA can support six times simultaneous connections than that of conventional and at least 30% improvement in spectrum efficiency over orthogonal frequency division multiple access. Index Terms—Fifth generation (5G), nonorthogonal multiple access (NOMA), pattern division multiple access (PDMA), successive interference cancellation (SIC).

I. INTRODUCTION ULTIPLE access schemes have been regarded as the landmark of each generation of mobile communication systems. Frequency division multiple access (FDMA) was used in the first generation (1G). Time division multiple access and code division multiple access were introduced in the second generation and the third generation (3G), respectively. Orthogonal frequency division multiple access (OFDMA) is the key component of the fourth generation (4G). As already known,

M

Manuscript received February 5, 2016; revised May 27, 2016; accepted July 4, 2016. Date of publication July 29, 2016; date of current version April 25, 2017. This work was supported in part by the National High Science & Technology Plan (863 Plan, No. 2015AA01A709) and the National Science Fund for Distinguished Young Scholars under Grant 61425012 in China. The review of this paper was coordinated by Dr. C. Xing. S. Chen and S. Sun are with the State Key Laboratory of Wireless Mobile Communications, China Academy of Telecommunications Technology, and Datang Telecom Technology and Industry Group, Beijing 100191, China (e-mail: [email protected]; [email protected]). B. Ren is with the School of Information and Communication Engineering, Beijing University of Posts and Telecommunications, Beijing, China (e-mail: [email protected]). Q. Gao and S. Kang are with the State Key Laboratory of Wireless Mobile Communications, China Academy of Telecommunications Technology, Beijing 100191, China (e-mail: [email protected]; [email protected]). K. Niu is with the Key Laboratory of Universal Wireless Communications, Ministry of Education, Beijing University of Posts and Telecommunications, Beijing, China (e-mail: [email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TVT.2016.2596438

above-mentioned multiple access schemes separate users in orthogonal resources such as frequency, time, or code domain resources. These orthogonal multiple access (OMA) schemes make it possible to build the system with low complexity; however, they are unable to achieve the capacity boundary of a multiuser channel [1]. The fifth generation (5G) needs to support much higher capacity, as well as provide much larger number of connected users [2]. These requirements are rather challenging and it is difficult to satisfy the requirements by using OMA schemes. Nonorthogonal multiple access (NOMA) has thus been considered as a promising candidate to meet the requirements in connection numbers and system capacity of 5G. Signals of multiple users are superposed and advanced detection algorithm are employed to separate the superposed signal. Theoretically, NOMA is optimal in terms of achieving the capacity boundary [1]. Actually, NOMA has been used in previous wireless systems, such as a 3G wideband code division multiple access (WCDMA) system. For example, in the uplink of WCDMA, data symbols are spread by long spreading codes, and multiple users transmit their spread signals on the same frequency and time resources. Since long spreading codes are used, only linear detection algorithm is possible at the receiver due to the complexity of a nonlinear detection algorithm. As a result, nonorthogonal transmission of this type is demonstrated to be inefficient in terms of spectrum efficiency (SE) [3]. In recent years, a number of NOMA techniques have been studied in both industry and academics, such as interleave division multiple access (IDMA) [4], bit division multiplexing (BDM) [5], sparse code multiple access (SCMA) [6], and multiuser sharing access (MUSA) [7]. With IDMA, different users are distinguished by user-specific interleavers. BDM was first proposed to provide differentiated services for a broadcasting system. It extends the multiplexing from the symbol level to the bit level. Both SCMA and MUSA are based on code domain superposition [8]. To separate users being multiplexed on the same resource, nonlinear detection algorithms such as successive interference cancellation (SIC), maximum a posteriori (MAP), or maximum likelihood (ML) are suggested for receiving. MAP and ML are so complex that they are very difficult to implement. An SIC receiver reaches a good tradeoff between performance and complexity. However, SIC suffers from the error propagation problem and degrades the performance of NOMA transmission. Specifically, if a user happens to be not

0018-9545 © 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications standards/publications/rights/index.html for more information.

3186


correctly detected, the following user will be likely to be erroneously detected, i.e., the performance of SIC relies heavily on the correctness of precedent detected users. This paper proposes a novel NOMA scheme based on code pattern, called pattern division multiple access (PDMA). Joint optimization of transmitting and receiving is considered with SIC amenable pattern design at the transmitter side and SIC based detection at the receiver side. PDMA patterns are designed to offer different order of transmission diversity, so that the disparate diversity order between multiple users could be introduced to alleviate the error propagation problem of the SIC receiver. A PDMA pattern is also required to be sparse to facilitate the advanced detection algorithm, such as belief propagation (BP). Iterations between BP and channel decoding could further boost system performance. A PDMA pattern is also extended to include power scaling and phase shifting to harvest additional constellation shaping gain. It is demonstrated by simulation results that PDMA could provide significant gain over OFDMA at both link level and system level. The main contribution of this paper is a complete reporting of PDMA, including the principle, framework, key technologies, application analysis, and performance evaluation. The rest of this paper is organized as following: In Section II, the principle of PDMA as well as system models for both uplink and downlink is introduced. In Section III, a PDMA pattern design at transmitter is discussed, including the principle of PDMA pattern design and extension of PDMA pattern matrix. In Section IV, detection algorithms at the receiver are discussed and analyzed. PDMA application scenarios and system design aspects are described in details in Section V. Performance evaluation and results are provided for PDMA and OFDMA in Section VI. Finally, in Section VII, concluding remarks are given. II. SYSTEM MODEL A. Principle of PDMA According to theoretical results of a multiuser channel [9], superposition coding at a transmitter and SIC at a receiver are able to achieve capacity boundary of multiple access channels or degraded broadcast channels when Transmitter and Receiver are working together. From theoretical perspective, it is rational to use SIC to achieve channel capacity, since the packet error rate tends to be zero with the increased code length as long as a user’s transmission rate is below the channel capacity. However, in a real system, detection error is inevitable due to various nonideal conditions, such as limited code length, channel fading, and glitches. For an SIC receiver, if a former user’s packet is detected erroneously, it is very unlikely that the following users’ packet could be detected correctly. This is the so-called error propagation problem. Since multiple users are detected one by one in a serial order, the detection order of all users is usually arranged according to their signal strength. That is, the signal of the first detected user is the strongest, the signal of the second detected user is weaker, and so on. For the first detected user, it is recovered directly from the original receiving. While for the following detected users, they are recovered respectively from related cancellation receiving, which should cancel those former detected

users from the original receiving by user reconstruction. If a user is not correctly detected, its reconstruction is impossible to be correct. In addition, the accuracy of reconstruction also impacts on the performance of following users. For example, based on distorted channel estimation, the reconstructed signal will also be distorted. Even though the user’s packet is detected correctly, it still has adverse effect on the following users’ detection. Error propagation is a crushing blow for multiuser detection, and it will deteriorate the performance of an SIC-based multiuser system. In general, two approaches can be considered to alleviate the error propagation problem. The first is to enhance the reliability of those early-decoded users, either by selecting users with good channel condition or by designing transmission parameters such that the early-decoded users have higher reliability and better channel condition. Analytical results of multiple-input multiple-output (MIMO) detection from [10] and [11] show that the ith detected layers of SIC receiver could achieve diversity order Ndiv (i) = NR − NT + i

(1)

where NR is the receiving antenna number, and NT is number of data layers. The diversity order increases with the detection order. The PDMA design is inspired by above-mentioned result [12]–[14]. A multiuser channel can be viewed as a virtual MIMO channel and the above-mentioned result could be generalized to multiuser nonorthogonal transmission. For nonorthogonal transmission employing an SIC receiver, diversity order of each user varies with the order of detection. The first detected user has the lowest diversity order, and the last detected user has the highest diversity order. In a fading channel, diversity order affects transmission reliability significantly. Increasing the diversity order typically leads to more reliable transmission. With the SIC receiver, the first detected user actually determines the overall detection performance, but unfortunately its diversity order is the lowest. To optimize system performance, it is desirable to have identical prodetection diversity order for each user. Diversity could be obtained from transmission or reception, or from both. Assuming that transmission diversity order of the ith detected user is DT (i), the diversity order after the SIC receiver can be expressed as Ndiv (i) = DT (i) − K + i

(2)

where K is the number of users. By joint design, from transmitter and receiver, PDMA deliberately selects DT (i) so that the diversity order after the SIC receiver is as close as possible. The definition of transmission diversity means that multiple copies of a signal are transmitted from independent resources to avoid transmission error due to deep fading on one resource. The resources could be time, frequency, or spatial resource. PDMA maps transmitted data onto a group of resources according to PDMA pattern to realize disparate transmission diversity order. A PDMA pattern defines the mapping from transmitted data to a resource group. A resource group can consist of time resource, frequency resource, spatial resource, or any combination of these resources. The number of mapped resources in a group determines the order of transmission diversity. Data of multiple users can be multiplexed onto the same resource

CHEN et al.: NOVEL NONORTHOGONAL MULTIPLE ACCESS FOR FIFTH-GENERATION RADIO NETWORKS

3187

where g k is an N × 1 binary vector with elements “0” or “1.” “1” means that the user’s data are mapped to the corresponding RE and, otherwise, not. The K users’ PDMA patterns on N REs [N ,K ] construct a PDMA pattern matrix GPDM A with the dimensions of N × K: [N ,K ]

GPDM A = [g 1 , g 2 , . . . , g K ] . Fig. 1.

PDMA pattern for six users on four REs.

The received signal y at the base station (BS) can be expressed

group with a different PDMA pattern. In this way, nonorthogonal transmission is realized. By assigning the PDMA pattern with different diversity order, disparate transmission diversity order among users could be achieved. The second approach to alleviate the error propagation problem is to adopt a more advanced and complex detection algorithm, such as ML or MAP. It is anticipated that PDMA with an advanced detection algorithm can alleviate the error propagation effect to a substantial degree. However, an ML or MAP algorithm incurs tremendous detection complexity, and it is hard to implement. Fortunately, the detection complexity could be reduced significantly by making the PDMA pattern sparse. That is, data are only mapped to a small part of the resources in the resource group. This draws on the idea of sparse coding in low-density parity check coding. Sparsity makes it possible to use the lowcomplexity BP algorithm to approach the MAP detection. In addition, convergence of the BP algorithm could be sped up by disparate transmission diversity of PDMA. In summary, PDMA uses a PDMA pattern to define sparse mapping from data to a group of resources. The PDMA pattern could be represented by a binary vector. The dimension of the vector equals the number of resource in a group. Each element in the vector corresponds to a resource in a resource group. A “1” means that data shall be mapped to the corresponding resource. Actually, the number of “1” in the PDMA pattern is defined as its transmission diversity order. Fig. 1 shows an example of resource mapping according to the PDMA pattern. Six users are multiplexed on four resource elements (REs). A PDMA pattern is assigned to a user. User1’s data are mapped to all four REs in the group, and user2’s data are mapped to the first three REs, etc. The order of transmission diversity of the six users is 4, 3, 2, 2, 1, and 1, respectively. B. System Model of Uplink Without loss of generality, we assume that both transmitter and receiver have a single antenna, and an example of an uplink PDMA system is shown in Fig. 2(a). At the transmitter, a PDMA encoder maps the modulation symbols xk onto resources and generates a PDMA modulation vector vk . At the receiver, a PDMA multiuser detector is used to detect data of multiplexed users. The uplink PDMA system consists of K users and their data are mapped onto N REs by using distinguished PDMA patterns. The PDMA modulation vector v k of user k is obtained by spreading the user’s modulation symbol xk according to the PDMA pattern g k v k = g k xk , 1 ≤ k ≤ K

(4)

(3)

as y=

K

diag(hk )v k + n

(5)

k =1

where n represents noise and interference at the receiver; hk is the uplink channel response of the kth user; y, n, and hk are vectors with length N . diag(hk ) represents a diagonal matrix with elements from hk . Considering (3) and (4), (5) can be written in a compact form: y = Hx + n

(6)

T

where x = [x1 x2 , . . . , xK ] , H is the PDMA equivalent channel response matrix of K users multiplexed on N REs and is given by [N ,K ]

H = H CH • GPDM A H CH = [h1 , h2 , . . . , hK ]

(7) (8)

where the (n, k) element of H CH is the channel response from the kth user to the BS on the nth RE, and • indicates elementwise product of two matrices. Overload factor is defined as the ratio between the number of users and the number of REs in a resource group. It reflects the multiplexing times of PDMA relative to orthogonal scheme. Taking N = 3, K = 6 as an example, the overload factor is then α = K/N = 200%, which means that PDMA supports two times user compared with OMA. In an example, the PDMA pattern matrix is expressed as ⎡ ⎤ 1 1 0 1 0 0 [3,6] (9) GPDM A = ⎣ 1 0 1 0 1 0 ⎦ . 0 1 1 0 0 1 The received signal is then given by

⎡

⎤ x1 ⎡ ⎤ ⎡ ⎤⎢ x2 ⎥ ⎡ ⎤ ⎥ n1 y1 h1,1 h1,2 0 h1,4 0 0 ⎢ ⎢ ⎥ ⎣ y2 ⎦ = ⎣ h2,1 0 h2,3 0 h2,5 0 ⎦⎢ x3 ⎥ + ⎣ n2 ⎦ . ⎢ x4 ⎥ ⎥ y3 n3 0 h3,2 h3,3 0 0 h3,6 ⎢

⎣ x5 ⎦ H x6 (10) As a further extension, the constellation mapping and PDMA encoding can be realized by mapping the user’s data bits directly to a PDMA modulation vector. This is a joint operation between PDMA encoding and constellation mapping—joint PDMA modulation. Specifically, the candidate set of PDMA modulation vector for user k, called codebook, shall be designed according to the PDMA pattern of user k. The codebook is designed offline and stored in a receiver and transmitter pair. Once the user’s PDMA pattern is decided, the associated codebook is therefore determined. Joint PDMA modulation is to choose

3188

Fig. 2.


PDMA system model. (a) Uplink. (b) Downlink.

a PDMA modulation vector from the codebook based on the user’s data bits. The vectors in the codebook and the PDMA pattern associated with it have the same sparsity property, i.e., the zero elements appear at the same position. Actually, nonzero elements in the vector is a multidimensional constellation in high-dimensional space. Joint PDMA modulation completes the PDMA encoding and the constellation mapping simultaneously in high-dimensional space. C. System Model of Downlink In a downlink system shown in Fig. 2(b), we assume that each user is assigned a PDMA pattern. After PDMA encoding, multiple data streams are superimposed at the BS and transmitted simultaneously. The received signal y k at user k can be expressed as y k = diag(hk )

K

g i xi + nk

i=1 [N ,K ]

= (diag(hk ) GPDM A )x + nk = H k x + nk

(11) [N ,K ]

H k = diag(hk ) GPDM A

(12)

where nk represents noise and interference at the receiver; hk is the downlink channel response of the kth user; and y k , nk , and hk are vectors with length N . H k is the PDMA equivalent channel response matrix of user k on N REs. x = [x1 x2 , . . . , xK ]T , where xk is the kth user’s modulation symbol. III. PATTERN DIVISION MULTIPLE ACCESS PATTERN DESIGN A. Pattern Matrix Design Users multiplexed PDMA patterns, on the same resource group, can construct a PDMA pattern matrix. Each column of the PDMA pattern matrix represents a PDMA pattern. Properties of the PDMA pattern matrix, such as dimension and level of sparsity, contribute both complexity and system performance.

Given a certain overload factor, there are a number of pattern matrices available, as long as K and N are selected properly. For example, overload factor of 150% could be achieved by a 2 × 3 pattern matrix, i.e., three users are multiplexed on two REs. The pattern matrix is

1 1 0 [2,3] GPDM A = 1 0 1 and another design for 150% overload is 4 × 6 pattern matrix: ⎡ ⎤ 1 0 1 1 1 0 ⎢1 1 0 1 0 1⎥ [4,6] ⎥ GPDM A = ⎢ ⎣1 1 1 0 1 0⎦. 0 1 1 0 0 1 Though both pattern matrices have the same overload fac[4,6] tor, GPDM A can achieve better performance, while the cost of [2,3] detection complexity is higher, compared to that of GPDM A . PDMA pattern matrices with different dimensions are able to achieve a given overload factor. With a higher dimension, detection complexity is also higher, and better performance is expected. Given the overload factor, the dimension of a pattern matrix shall be selected to reach a tradeoff between complexity and performance. If N is the size of a resource group (row number of PDMA pattern matrix), there are 2N − 1 possible binary vectors for a pattern matrix. Assuming that K is the column number determined based on overload factor, we can thus choose K patterns out from 2N − 1 candidates to construct the PDMA pattern matrix. Selection of patterns also makes an impact on performance and complexity. 1) A pattern with heavier weight (number of “1” elements in the pattern) provides a higher diversity order. More reliable data transmission can be anticipated, and detection complexity is also increased. If the system can conduct complex computation, patterns with heavy weight will be preferable; otherwise, lightweight patterns have to be selected, aiming at the sparse PDMA pattern matrix.


2) According to the design principle of PDMA, it is desirable to have different diversity orders in the pattern matrix to alleviate the error propagation problem of the SIC receiver or fasten convergence of the BP receiver. Thus, the selected patterns shall have as many different diversity orders as possible. 3) For patterns with identical diversity order, smaller inner product between the patterns leads to less interference against each other. Small inner product means that the two patterns have less “1” elements in common positions. That is, the number of REs shared by the two patterns is low. Data of two users are multiplexed on only few REs. For example, if two patterns have inner product of 0, the two patterns actually maps data onto a different set of REs; hence, there is no interference between the two patterns. For a given diversity order, the selected patterns shall minimize the maximum inner product between any two patterns. Of course, this rule is also applied to patterns with different diversity order. The design of a pattern matrix shall take overload factor, diversity order, and detection complexity into account. A good pattern matrix can reach good tradeoff among these aspects. The criteria of maximum constellation-constrained capacity (CC-Capacity) [15] can be used to design a PDMA pattern matrix especially for uplink application. That is, with the input information on matrix dimension and its row weight, all the candidate sets of a PDMA pattern matrix are calculated by CC-capacity, then the PDMA pattern matrix with maximum CC-capacity is selected [16]: [N ,K ]

Gopt

= arg max{C(N, K, Ω)G[N ,K ] ⊂ G[N ,M ] }

st.G[N ,K ] (:, k)2 = 1(k = 1, 2, . . . , K), x ∈ ΩK ×1 (13) where C(N, K, Ω) denotes the CC-capacity for the parameters N , K; Ω, Ω denotes the constellation set M = 2N − 1; and x is defined in (6). As an example, assuming that the matrix dimension is 4 × 6, the calculation of C(N, K, Ω) is as follows: C(4, 6, ΩQPSK ) = I(v 1 ; y) + I(v 2 ; y|v 1 ) + I(v 3 ; y|v 1 , v 2 ) + I(v 4 ; y|v 1 , v 2 , v 3 ) + I(v 5 ; y|v 1 , v 2 , v 3 , v 4 )

(14)

+ I(v 6 ; y|v 1 , v 2 , v 3 , v 4 , v 5 ) where I(·) represents the mutual information between the input QPSK symbol and the output of a Gaussian channel, and I(v 2 ; y|v 1 ) denotes the conditional mutual information between v 2 and y with the given value of v 1 , where v k and y are defined in (3) and (5), respectively. I(.) and I(v 2 ; y|v 1 ) in (14) can be calculated by adopting the Monte Carlo integration method. Considering different row weights, the selected PDMA pattern

3189

matrices are expressed as follows:

row weight 2:

[4,6]

GPDM A

⎡

1 ⎢1 = ⎢ ⎣0 0 ⎡

row weight 3:

[4,6]

GPDM A

1 ⎢1 = ⎢ ⎣0 0

0 0 1 1

1 0 0 0

0 1 0 0

0 0 1 0

⎤ 0 0⎥ ⎥ 0⎦ 1

1 0 1 0

1 0 0 1

0 1 1 0

0 1 0 1

⎤ 0 0⎥ ⎥. 1⎦ 1

As another example, in case of overload factor 200%, assuming selecting the row weight 4, the finally selected PDMA pattern matrix with dimension 3 × 6 is expressed as ⎡ ⎤ 1 1 1 0 1 0 [3,6] GPDM A = ⎣ 1 1 0 1 0 1 ⎦ . 1 0 1 1 0 0 The finally selected PDMA pattern matrix with dimension 4 × 8 is expressed as ⎡ ⎤ 1 1 1 1 0 0 0 0 ⎢1 1 0 0 1 1 0 0⎥ [4,8] ⎥ GPDM A = ⎢ ⎣1 0 1 0 1 0 1 0⎦. 0 0 0 1 0 1 1 1 B. Extension of Pattern Matrix [2,3]

Taking the PDMA pattern matrix GPDM A as an example, data of three users are mapped onto two REs. The transmission signal on these REs can be expressed as ⎡ ⎤

x v1 1 1 0 ⎣ 1⎦ x2 (15) = v2 1 0 1 x3 where vj is the transmission signal on the jth RE, and xk is the modulation symbol of the kth user. Unlike orthogonal transmission, transmission signal on each RE is linear combination of multiple modulation symbols: v1 = x1 + x2 v2 = x1 + x3 .

(16)

This combination may alter the characteristics of the transmission signal on each RE. For example, if all three users adopt BPSK modulation, the modulation symbol of user 1, user 2, and user 3 is either +1 or −1. The combined transmission signal takes value from {−2, 0, +2}. Assuming a noiseless channel, if the receiver receive −2, or +2 on an RE, then the receiver can infer that the transmitted symbols on the RE is [−1, −1] or [+1, +1], but if 0 is received, it is impossible for the receiver to recover the transmitted symbols, as both [+1, −1] and [−1, +1] resulting in the same output. Furthermore, if each user adopts QPSK or 16QAM modulation, the combined constellation consists of nine or 49 constellation points. From the above discussions, we can see that the combined constellation has nonuniform distribution and it is no longer a one-to-one map between a constellation point and an input user data, i.e., the combination leads to ambiguity.

3190


Fig. 3. Combined constellation from two users by phase shifting ϕ = π/4. (a) QPSK. (b) 16QAM. Fig. 5.

SIC receiver.

given as

[2,3] GE −PDMA =

α11 e−j ϕ 11 α21 e−j ϕ 21 α12 e−j ϕ 12

0

0

α32 e−j ϕ 32

(18)

where αkj and ϕkj are the power scaling and phase shifting factors of the kth user on the jth RE. The optimal value of power scaling and phase shifting depend on the number of users and the shape of input constellation. Fig. 4. Combined constellation from two users by power scaling and phase shifting β = 0.8, ϕ = π/4. (a) QPSK. (b) 16QAM.

To resolve the ambiguity, power scaling and phase shifting can be introduced in the PDMA pattern matrix. Specifically, before two users’ symbols are mixed, a power scaling factor and a phase shifting factor shall be applied: v=

βx1 ej ϕ +

1 − βx2

(17)

where β is power scaling factor and ϕ is phase shifting factor. As an example, by setting ϕ = π/4 and β = 0.5, i.e., only phase shifting difference is introduced between users, the combined constellation is shown in Fig. 3. It can be observed that by adding a phase shifting factor, the ambiguity is resolved. Moreover, the distribution of combined constellation is closer to Gaussian distribution. It is known that Gaussian distribution is the capacity maximizing input distribution for the additive white Gaussian noise channel. That is, by introducing a phase shifting factor, channel capacity gain is reaped. The gain is also called the shaping gain. When both power scaling and phase shifting are introduced, the effect is exemplified in Fig. 4. As expected, the shape of constellation is changed, and it also approaches a Gaussian distribution. For PDMA, the power scaling and phase shifting can naturally be incorporated into the PDMA pattern matrix. That is, the value “1” in the PDMA pattern matrix is substituted with a complex value reflecting both power scaling and phase shifting, forming an extended PDMA pattern matrix. For the PDMA [2,3] pattern matrix GPDM A , the extended PDMA pattern matrix is

IV. DETECTION ALGORITHM AT RECEIVER Detection algorithm at receiver is the key to reap performance gain of PDMA in uplink and downlink. This section is dedicated to the details of the three algorithms suitable for PDMA: SIC, BP, and BP-iterative detection and decoding (IDD). By arranging diversity order of PDMA patterns, error propagation problem of the SIC receiver could be alleviated to a certain extent. Sparsity of the PDMA pattern greatly reduces the complexity of the BP and BP-IDD algorithms, making them suitable for the PDMA system. In addition, the PDMA pattern could be designed to speed up convergence of BP and BP-IDD. A. SIC As shown in Fig. 5, the basic idea of the SIC receiver is to reconstruct a user’s signal and then subtract it from the received signal. The construction could be carried out either at the symbol level or codeword level. For the symbol-level SIC (SL-SIC), the construction is made from the demodulated symbols. Instead, the codeword-level SIC (CW-SIC) is based on signal construction from decoded data bits. As channel decoding is able to correct most errors, CW-SIC is expected to perform better than SL-SIC. B. Belief Propagation BP algorithm has been demonstrated to be able to approach MAP detection asymptotically [17], [18]. Furthermore, the sparsity of the PDMA pattern reduces the complexity of the BP algorithm, making it suitable for the PDMA system, and disparate transmission diversity of PDMA can speed up the convergence of BP.


Fig. 6.

3191

[3, 6]

Factor graph of PDMA with GP D M A .

Fig. 7.

Structure of the BP-IDD receiver for PDMA uplink system.

Fig. 8.

Factor graph of PDMA with GP D M A and Turbo decoder.

Given the received signal vector y and the PDMA equivalent channel response matrix H in (7), the optimal detection of x is a joint MAP detection ˆ = arg max p(x|y, H) x x∈ℵK

(19)

where ℵK represents the constellation alphabet of K users. Equation (19) can be approximated by a local MAP solution based on Bayesian formula xˆk = arg max P (x) p(yn |x) (20) s∈ℵ

x∈ℵK ,x k =s

n ∈N v (k )

where ℵ and Nv (k) represent constellation alphabet and the RE index set corresponding to the PDMA pattern of the kth user. According to [17] and [18], the problem can be solved by applying the BP algorithm on the underlying factor graph. [N ,K ] A PDMA system with a pattern matrix GPDM A can be represented by a factor graph consisting of channel observation nodes (CND) and user nodes (UND). A factor graph with a PDMA [3,6] pattern matrix GPDM A is shown in Fig. 6. The kth UND corresponds to the kth user’s data symbol, and the jth CND represents the received signal on the jth RE yj (1 ≤ j ≤ N ). If there is an edge between the kth UND and the jth CND (i.e., [N ,K ] GPDM A (j, k) = 0), the received signal on the jth RE includes contributions from the kth user. For a detailed procedure about the BP algorithm, see [17] and [18]. C. Belief Propagation-Iterative Detection and Decoding The basic principle of the BP-based iterative detection and decoding (BP-IDD) algorithm is that the decoded bit loglikelihood ratio (LLR) is fed back from a Turbo decoder and converted to a symbol-LLR as the a priori information of the BP detector. There are two iterative processes in the BP-IDD receiver: 1) inner iteration processing of the BP detector and 2) outer iteration processing between the BP detector and Turbo decoder. As shown in Fig. 7, besides a traditional BP multiuser detector, the BP-IDD receiver includes multiple parallel iterative processes, each of which is composed of modules of deinterleaver, Turbo decoder, and interleaver. Here, soft information is transferred between the multiuser detector and the Turbo decoder, in the form of LLR. A factor graph of the BP-IDD algorithm based on the [3,6] PDMA pattern matrix GPDM A is shown in Fig. 8. VND denotes the variable node. Let xk (k = 1, . . . , K) be data symbols

[3, 6]

TABLE I COMPUTATION COMPLEXITY PER MODULATION SYMBOL Algorithm SIC BP BP-IDD

Number of multiplications

Number of additions

O (K N 3 ) O (d f N M d f ) O (d f N M d f )

O (K N 3 ) O (T i n d f N Q m M d f ) O (T o u t T i n d f N Q m M d f )

corresponding to the UND of the kth user and associated to the VND ck ,i (i = 1, . . . , m), where m represents the modulation order of the kth user. The connection between UND and VND shall satisfy a certain condition imposed by channel encoding. According to Fig. 8, the iterative process between UND and CND, as described in Section IV-B, is called inner iteration, and the iterative process between UND and VND is called outer iteration. For a detailed procedure about the BP-IDD algorithm, see [18] and [19]. D. Comparison of SIC, BP, and BP-IDD Among the detection algorithms described above, it is expected that BP-IDD is the best of all in terms of performance, and BP would be better than SIC. Table I summarizes computation complexity of the abovementioned three detection algorithms [20], [21]. In Table I, M denotes the size of modulation constellation, Qm = log2 (M ), Tin , Tout , and df represent BP-IDD inner iteration number,

3192


outer iteration number, and maximum row weight of the PDMA pattern matrix. The number of additions of BP-IDD receiver is about Tout times of the BP receiver, yet they have the same number of multiplications as the value of |yj − f j x|2 in [21, (17)] is calculated only once and reused for the following inner and outer iterations in BP-IDD. It should also be noted that the computation of the Turbo decoding is not accounted for in Table I. As the processing ability of a BS is usually more powerful compared with that of a user terminal, BP-IDD and BP could be used in the PDMA uplink. Detection at user terminal in the PDMA downlink can choose either BP or SIC, depending on its processing ability. V. PATTERN DIVISION MULTIPLE ACCESS IN FIFTH-GENERATION IN 5G SYSTEM A. Application Scenarios PDMA can be applied in typical scenarios of 5G, for example, in enhanced mobile broadband (eMBB) or massive machine type communication (MTC). In an eMBB scenario, the main challenge is to enhance the transmission rate. Massive MIMO, ultradense networking, and high-frequency communication are candidate technologies. PDMA can be used in conjunction with these technologies. In a massive MTC scenario, a BS needs to provide connection to a huge number of low-cost terminals. The main challenge for a massive MTC scenario is how to effectively deal with massive connection with power constraint. In a 4G Long-Term Evolution (LTE) system, to transmit data, a user shall first issue scheduling request (SR) on periodically occurred resources, which is configured by BS. The base station then makes scheduling decisions and sends uplink grant to the user indicating the resources on which the user can transmit data. Generally, the procedure may take 10 ms or more. For some massive MTC applications, such a long latency is unacceptable. Moreover, the uplink grant is carried by downlink control signaling, and with a massive number of connections the downlink control channel may become a bottleneck. In such a situation, grant-free transmission is a viable option. By means of grant-free transmission, a user autonomously selects resource for transmission without an SR and scheduling of BS. To avoid interfering with other traffic scheduled by the BS, resource for grant-free transmission shall be confined within a certain set of resources. The resource set is called the resource pool. For orthogonal transmission, the resource pool consists of resource in time and frequency domains. A user selects a resource from the pool for transmission. As there is no coordination between users sharing the same resource pool, it is likely that two users select the same resource. When a collision happens, it may lead to failure in detection. The probability of collision is proportional to the number of users sharing the resource pool, and is inversely proportional to the number of resources in the pool. That is, enlarging the resource pool could reduce the collision probability. As a nonorthogonal transmission scheme, PDMA could naturally be incorporated into grant-free transmission to reduce the

collision probability. That is, PDMA provides another dimension for resource sharing—the PDMA pattern. A traditional resource pool could be extended to include the PDMA pattern. Specifically, each resource group in the pool is associated with a PDMA pattern matrix. An UE selects a time–frequency resource as well as a PDMA pattern from the pattern matrix for transmission. Even though two users may select the same time–frequency resource, as long as their PDMA patterns are different, the receiver is able to decode the two users’ data successfully. The resource pool is α − 1 times larger than a traditional resource pool, where α is the overload factor of the PDMA pattern matrix. B. System Design Aspects To enable PDMA in practice, a number of aspects in a system design shall be considered. 1) Air Interface and Process Design: PDMA enables a large number of users to transmit on the same resource, especially when PDMA is used jointly with massive MIMO. The need for a reference signal will be multiplied accordingly. A reference signal shall be designed carefully to meet the requirements of detecting a PDMA signal and keeping the overhead incurred by the reference signal to a reasonable level. Multiuser transmission of PDMA also leads to demanding requirements on control channel, as each users data shall be accompanied by a control channel to provide necessary information for detection. Techniques such as multisubframe scheduling, group control signal design, and the control signaling content design could serve as a starting point to cope with the problem. Link adaptation for downlink PDMA is based on user reporting of the channel quality information (CQI). However, the multiuser pairing nature of PDMA makes it difficult for a user to predict CQI without knowing its pairing users and their PDMA patterns. Power domain optimization may further complicate the problem, as the user will not know what the transmission power is going to be before making the scheduling decision. A flexible CQI calculating and reporting mechanism is needed. As discussed above, the grant-free transmission is able to reduce data latency and to control overhead. Although, by the introduction of PDMA, a resource pool could be extended, it is still possible that two users collide on the same resource and PDMA pattern especially when the system is heavily loaded. To facilitate grant-free transmission, a resource selection method and a mechanism to resolve conflicts play a fundamental role. 2) Radio Resource Management: Traditionally, the radio resource management deals with the allocations of time, frequency, and spatial resources to make full use of the wireless resources. PDMA introduces another dimension—the PDMA pattern. The optimization problem becomes really challenging, as more optimization variables are involved. A low-complexity radio resource management algorithm, which could achieve near-optimal performance, is worth seeking.


3193

TABLE II PDMA LINK-LEVEL SIMULATION ASSUMPTIONS Parameter

Value

Carrier System bandwidth Channel model Modulation coding rate Antenna configuration Channel estimation HARQ Uplink overload factor Uplink average signal to noise ratio (SNR)

2 GHz 10 MHz UMA-NLOS [22] QPSK 1/2; LTE Turbo 1 Tx 2 Rx Perfect No 150% ; 200% ; 300% The same of all users

Fig. 10.

PDMA downlink performance.

TABLE III PDMA SYSTEM-LEVEL SIMULATION ASSUMPTIONS Parameters Topology Number of usersper cell Carrier Bandwidth ISD Channel model Power control

Fig. 9.

PDMA uplink performance.

VI. PERFORMANCE EVALUATION A. Comparison of PDMA and OMA Taking an LTE orthogonal system as a reference, uplink and downlink performances of PDMA are evaluated and results are provided in this section. The assumptions of the link-level simulations are shown in Table II, and the results are given in Figs. 9 and 10. The system-level simulation assumptions are shown in Table III, and the simulation results are provided in Figs. 11 and 12. Fig. 9 shows the total SE of uplink PDMA under the overload factors of 150%, 200%, and 300% as well as SE of OFDMA. For fair comparison, same number of source bits is assumed for PDMA and OFDMA. For the given overload factors of 150%, 200%, and 300%, SE gains of 50%, 100%, and 200% can be achieved by PDMA over OFDMA when SNR is high enough. In a downlink system, the SNR differences between users have tremendous influence on the performance of PDMA. The performance gain of PDMA over OFDMA gets more remarkable when the difference gets larger. As shown in Fig. 10, when the SNR difference of two users is 12 dB, the SE gain is 14% when the SNR is −4 dB; reaches the maximum of 50% when SNR is 0 dB; and vanishes when the SNR is higher than 10 dB. The uplink grant-free OFDMA and PDMA transmissions are evaluated under traffic with small burst packet and the latency is

The number of antenna Antenna configuration

Channel estimation Scheduler MCS Maximum HARQ transmission times Traffic model Receiver

Value Hexagonal homogeneous network;19 sites/57 sectors 10, 20, or 30 2GHz Uplink: 5 MHz Downlink: 10 MHz 500 m ITU UMa [22] Uplink: Open-loop power control, alpha = 1, P0 = −95 dBm Uplink: 1T times 2R times Downlink: 2T times 2R times Uplink: User vertical polarization; BS ±45◦ cross polarization Downlink: User ±45◦ cross polarization; BS ±45◦ cross polarization Perfect Uplink: Grant-free Downlink: PF schedule Uplink: 160 bits @ 1PRB Downlink: Adaptive (Based on LTE downlink MCS definition) Uplink: 0 Downlink: 3 Uplink: Bursty traffic with small packet Downlink: Full buffer traffic Linear MMSE receiver for OFDMA, BP-IDDfor PDMA

Fig. 11. PDMA uplink system performance (supported packet arrival rate when packet drop rate = 1%).

3194


TABLE IV RECEIVER COMPLEXITY OF DIFFERENT PDMA PATTERN MATRIX Pattern matrix 4 × 6 row weight d f 4 × 6 row weight d f 3 × 6 row weight d f 4 × 8 row weight d f

Fig. 12.

= = = =

2 3 4 4

N m u l ≈ 8 × 42 N m u l ≈ 12 × 43 N m u l ≈ 12 × 44 N m u l ≈ 16 × 44

PDMA downlink system performance.

Fig. 14.

Fig. 13.

Computation complexity

Performance of different PDMA pattern matrices.

required to be not more than 1 ms. A gain of 500% in terms of the number of supported users under the given system packet drop rate of 1% is observed in Fig. 11. The gain comes from two facts. First, PDMA provides a larger resource pool than OFDMA does, so that the collision probability of PDMA is lower than that of OFDMA. Second, the BP-IDD receiver employed by PDMA is more capable of dealing with interference when collision occurs. From the results of downlink PDMA shown in Fig. 12, PDMA can get about 30% gain compared with OFDMA in terms of both SE at cell edge and cell average SE. The gain increases with the user number in a cell increasing, because with more users it is easier to find suitable users for pairing in PDMA. B. Comparison of Different Pattern Matrices Taking overload factor 150% and 200% as examples, uplink performances of PDMA with different pattern matrices are provided in this section. For the given pattern matrices 4 × 6, 3 × 6, and 4 × 8 described in Section III-A, Fig. 13 shows their performance comparison. As shown in Fig. 13, higher row weight (df ) can get better performance. For PDMA pattern matrix 4 × 6, row weight df =

Performance of uplink PDMA with BP or BP-IDD at receiver.

3 has 1 dB gain compared with row weight df = 2. In addition, the PDMA pattern matrix with higher dimension can get better performance. For the same overload 200%, pattern matrix 4 × 8 has 0.8 dB gain compared with pattern matrix 3 × 6. However, a higher row weight or a higher dimension in the construction PDMA pattern matrix means a higher complexity in detection. Table IV gives a comparison of the computation complexity about these pattern matrices. For the same dimension 4 × 6, the complexity of the pattern matrix with row weight df = 3 is six times of the pattern matrix with row weight df = 2. For the same overload 200%, the complexity of pattern matrix 4 × 8 is 1.33 times of pattern matrix 3 × 6. Therefore, a tradeoff between performance and complexity should be considered in the pattern matrix design. C. Comparison of Different Receivers For those receivers described in Section IV, uplink performances of PDMA with BP or BP-IDD receivers are compared, and the results are shown in Fig. 14. Also, downlink performances of PDMA with BP or SIC are compared, and the results are shown in Fig. 15. Table V gives comparisons of the computation complexity. As shown in Fig. 14, the BP-IDD receiver has 0.8-1.6 dB gain compared with the BP receiver. Especially, the gain is higher with overload increasing, e.g., the gain is 0.8 dB for overload 150% with pattern matrix 2 × 3, and the gain is 1.6 dB for overload 300% with pattern matrix 4 × 12. This is because that


3195

System-level simulation results show that PDMA can support up to six times connected users and at least 30% improvement in SE over OFDMA. Further research directions of PDMA include combination of PDMA encoding and modulation, lowcomplexity detection algorithm, and combining with multiple antennas, etc. At present, with the support of the 863 project (Item No.2015AA01A709) of the Chinese government, the PDMA testbed is being developed including multiple transmitters and multiple receivers. The PDMA field tests in both the eMBB scenario and the massive MTC scenario are expected in 2016.

ACKNOWLDGMENT

Fig. 15.

Performance of downlink PDMA with BP or SIC at the receiver.

TABLE V COMPLEXITY OF DIFFERENT RECEIVERS Link type

Receiver

Computation complexity

Uplink [2, 3]

BP-IDD BP BP SIC

N m u l ≈ 64, N a d d ≈ 1152 N m u l ≈ 64, N a d d ≈ 384 N m u l ≈ 64, N a d d ≈ 384 N m u l ≈ 24, N a d d ≈ 24

Downlink [2, 3]

Note: N m u l denotes the number of multiplications, N a d d denotes the number of additions, and ≈ means approximately equal to.

the performance of the BP algorithm converges more slowly for an overload factor of 300% compared with 150% and the frequency diversity degree is higher for 300% than for 150%. Whereas, from computation complexity, addition in BP-IDD receiver is three times that in BP receiver. As shown in Fig. 15, the BP receiver has a gain of 2 dB compared with the SIC receiver in downlink where the multicircuit substation (MCS) of the near user and far user is QPSK 1/2, and the power factor is 0.2 and 0.8. Whereas the computation complexity of the BP receiver is about 1.67 and 15 times higher than the SIC receiver in multiplication and addition, respectively VII. CONCLUDING REMARKS PDMA has been incorporated into ITU-R Report “Future technology trends of terrestrial IMT systems” [23] by ITU organization since 2014. It is a novel NOMA scheme based on SIC amenable multiple access [24], [25]. To achieve an integral optimization for a multiuser communication system, PDMA considers the joint design of transmitting and receiving, with the SIC amenable pattern design at the transmitter- and SIC-based multiuser detection at the receiver. This paper provides a whole picture of PDMA, including system model, PDMA pattern design, PDMA detection algorithm, application in a 5G system, and performance evaluation.

The authors would like to thank Dr. Y. Wang from China Academy of Telecommunications Technology and Prof. X. Dai from Beijing Science and Technology University for their valuable comments.

REFERENCES [1] D. Tse and P. Viswanath, Fundamentals of Wireless Communication. Cambridge, U.K.: Cambridge Univ. Press, 2005. [2] IMT-2020(5G) Promotion Group, Whitepapers (2004). [Online]. Available: http://www.imt-2020.org.cn/zh [3] Fifth generation non-orthogonal waveforms for asynchronous signalling (2013). [Online]. Available: http://www.5gnow.eu [4] P. Li, L. Liu, K. Wu, and W. Leung, “Interleave division multiple access,” IEEE Trans. Wireless Commun., vol. 5, no. 4, pp. 938–947, Apr. 2006. [5] H. Jin, K. Peng, and J. Song, “Bit division multiplexing for broadcasting,” IEEE Trans. Broadcast., vol. 59, no. 3, pp. 539–547, Apr. 2013. [6] L. Lu et al., “Prototype for 5G new air interface technology SCMA and performance evaluation,” China Commun., vol. 12, pp. 38–48, Dec. 2015. [7] Z. Yuan et al., “Multi-user shared access for Internet of Things,” in Proc. IEEE 83th Conf. Veh. Tech., 2016, pp. 1–5. [8] L. Dai et al., “Non-orthogonal multiple access for 5G: Solutions, challenges, opportunities and future research trends,” IEEE Commun. Mag., vol. 53, no. 9, pp. 74–81, Sep. 2015. [9] M. C. Thomas and A. T. Joy, Elements of Information Theory, 2nd ed. Hoboken, NJ, USA: Wiley, 2006. [10] S. Loyka and F. Gagnon. “Performance analysis of the V-BLAST algorithm: An analytical approach,” IEEE Trans. Wireless Commun., vol. 3, no. 4, pp. 1326–1337, Feb. 2002. [11] D. Truhachev, “Universal multiple access via spatially coupling data transmission,” in Proc. Int. Symp. Wireless Pers. Inf. Theory, Jul. 2013, pp. 1884–1888. [12] X. Dai, S. Sun, and Y. Wang, “Successive interference cancellation amenable space-time codes with good multiplexing-diversity tradeoffs,” Wireless Pers. Commun., vol. 55, no. 4, pp. 645–654, Dec. 2010. [13] X. M. Dai, S. H. Sun, and Y. M. Wang, “Reducing complexity of quasimaximum-likelihood detectors through companding for coded MIMO systems,” IEEE Trans. Veh. Technol., vol. 61, no. 3, pp. 1109–1123, Mar. 2012. [14] S. Sun, “Pattern division multiple access (PDMA),” in Proc. FutureTaiwan 5G Workshop (Nov. 2014). [Online]. Available: http://futureforum.org (Invited speech). [15] R. Hoshyar, F. Wathan, and R. Tafazolli, “Novel low-density signature for synchronous CDMA systems over AWGN channel,” IEEE Trans. Signal Process., vol. 56, no. 4, pp. 1616–1626, Jan. 2008. [16] B. Ren, et al., “Pattern matrix design based on PDMA for 5G UL application,” China Commun., to be published. [17] X. Dai, S. Sun, and Y. Wang, “Reduced-complexity performance-lossless (Quasi-)maximum-likelihood detectors for S-QAM modulated MIMO systems,” Elect. Lett., vol. 49, no. 11, pp. 724–725, May 2013. [18] X. Dai, Z. Zhang, K. Long, S. Sun, and Y. Wang, “Unequal error correcting capability aware iterative receiver for (parallel) Turbo coded communications,” IEEE Trans. Veh. Technol., vol. 63, no. 7, pp. 3446–3451, Sep. 2014.

3196


[19] J. Xu, X. Dai, W. Ma, and Y. Wang, “A component-level soft interference cancellation based iterative detection algorithm for coded MIMO systems,” in Proc. IEEE 80th Veh. Technol. Conf., Sep. 2014, pp. 1–5. [20] D. N. Liu and M. P. Fitz, “Low complexity affine MMSE detector for iterative detection-decoding MIMO-OFDM systems,” IEEE Trans. Commun., vol. 56, no. 1, pp. 150–158, Jan. 2008. [21] B. Ren et al., “Advanced IDD receiver for PDMA uplink,” in Proc. IEEE/CIC Int. Conf. Commun. China, Chengdu, China, Jul. 2016. [22] Rep. ITU-R M.2135-1, “Guidelines for evaluation of radio interface technologies for IMT-advanced,” Dec. 2009. [Online]. Available: http://www.itu.int/dms_pub/itu-r/opb/rep/R-REP-M.2135-1-2009PDF-E.pdf [23] Report ITU-R M.2320-0, Future technology trends of terrestrial IMT systems, Nov. 2014. [Online]. Available: http://www.itu.int/ITU-R/ [24] X. Dai et al., “Successive interference cancelation amenable multiple access (SAMA) for future wireless communications,” in Proc. IEEE Int. Conf. Commun. Syst., Nov. 2014, pp. 222–226. [25] S. Kang, X. Dai, and B. Ren, “Pattern division multiple access for 5G,” in Proc. Telecommun. Netw. Technol., May 2015, pp. 43–47.

Shanzhi Chen (SM’04) received the Ph.D. degree in Communication and Information Systems from Beijing University of Posts and Telecommunications, Beijing, China, in 1997. He joined Datang Telecom Technology & Industry Group in 1994, where the has been serving as Chief Technology Officer since 2008. His current research interests include network architecture, fifthgeneration mobile communication, and Internet of Things. Dr. Chen received the 2001 and 2012 National Awards for Science and Technology Progress, China; the 2015 National Award for Technological Invention, China; and the 2014 Distinguished Young Scholar Award of National Natural Science Foundation, China. He was a member of the steering expert group on information technology of the 863 Hi-Tech Research and Development Plan of China from 1999 to 2011. He is the Director of State Key Laboratory of Wireless Mobile Communications and is a Board Member of Semiconductor Manufacturing International Corporation. He has devoted his works to the research and development of TD-SCDMA third-generation industrialization and TD-LTE-advanced fourth-generation standardization.

Bin Ren received the B.E. and M.S. degrees in information and communication engineering in 2006, and 2009, respectively, from Beijing University of Posts and Telecommunications, Beijing, China, where he is currently working toward the Ph.D. degree. He joined the Key Laboratory of Wireless Mobile Communications, China Academy of Telecommunication Technology, Beijing, in 2009. His current research interests include wireless communications theory and wireless communications systems.

Qiubin Gao received the B.S. and Ph.D. degrees in 2002, and 2008, respectively, in control science and engineering from Tsinghua University, Beijing, China. He is currently a Senior Research Engineer with the Datang Wireless Mobile Innovation Center, China Academy of Telecommunications Technology. His current research interests include physical layer design for mobile communication, multiple-antenna technology, Coordination of Multiple Points, deviceto-device communication, and system performance evaluation. He is the inventor/co-inventor of more than 300 patents in wireless communications and the author/co-author of a number of journal and conference papers. Dr. Gao received the International Union of Radio Science (URSI) award for Young Scientists in 2014.

Shaoli Kang received the Ph.D. degree in signal and information processing from Beijing Jiaotong University (BJTU), Beijing, China, in 2002. She is the Head Expert of fifth-generation (5G) standardization with the Wireless Innovation Centre, China Academy of Telecommunication Technology (CATT), Beijing, China. From November 2000 to May 2005, she was a Project Manager with CATT, focusing on R&D of TD-SCDMA. Then she was with the Communication Centre of System Research, University of Surrey, U.K., as a Research Fellow, doing research on projects from EPSRC and OFCOM and leading the Antenna and Propagation Club. In September 2007, she returned to CATT and acted as the Vice Chief Engineer of the TDD R&D product line, focusing on speeding up the standard and industrial progress of TDD technology. Since early 2011, she has been with the Wireless Innovation Centre and acting as the Head Expert, leading 5G research with CATT. She has applied for more than 50 patents and published more than 20 papers.

Shaohui Sun received the Ph.D. degree in Communication and Information Systems from Xidian University, Xi’an, China, in 2003. From March 2003 to June 2006, he was a Postdoctoral Research Fellow with the Datang Telecom Technology and Industry Group, Beijing, China. From June 2006 to December 2010, he was with the Datang Mobile Communications Equipment Company Ltd., Beijing, where he had been deeply involved in the development and standardization of the Third-Generation Partnership Project Long-Term Evolution. Since January 2011, he has been the Chief Technical Officer of Datang Wireless Mobile Innovation Center, Datang Telecom Technology and Industry Group. His current research areas of interest include multiple antenna technology, heterogeneous wireless networks, and relays.

Kai Niu (S’00–M’04) received the B.S. degree in information engineering and the Ph.D. degree in signal and information processing from Beijing University of Posts and Telecommunications (BUPT), Beijing, China, in 1998 in 2003, respectively. He is currently a Professor with the School of Information and Communication Engineering, BUPT. His research interests include coding theory and its applications, space-time codes, and broadband wireless communication. Dr. Nu has been serving as a Senior Member of the Chinese Institute of Electronics (CIE) and a Committee Member of the Information Theory Chapter of CIE since 2008.

Pattern Division Multiple AccessâA Novel Nonorthogonal Multiple

Pattern Division Multiple AccessâA Novel Nonorthogonal Multiple

Pattern Division Multiple AccessâA Novel Nonorthogonal Multiple

Pattern Division Multiple AccessâA Novel Nonorthogonal Multiple