Research Center of 6G Mobile Communications, School of Cyber Science and Engineering, Huazhong University of Science and Technology, Wuhan 430074, China
Tao Jiang.
Constructing Air-Interface Links for Mobile Communications: From {0,1} to [0,1].
Engineering, 2025, 46(3): 17-23 DOI:10.1016/j.eng.2024.10.007
Mobile communications have catalyzed a new era of information technology revolution, significantly broadening and deepening human-to-human, human-to-machine, and machine-to-machine connections. With their incredible speed of development and wide-reaching impact, mobile communications serve as the cornerstone of the Internet of Everything, profoundly reshaping human cognitive abilities and ways of thinking. Furthermore, mobile communications are altering the patterns of production and life, driving leaps in productivity quality, and strongly promoting innovation within human civilization. Over the past decades, mobile communications in China have experienced rapid development, realizing the widespread commercialization of the fifth generation mobile communication technology (5G), enabling high-speed and reliable mobile communications to reach countless households, and accomplishing a remarkable transformation from weak to strong industry.
The history of mobile communications development can be succinctly summarized as the conflict between the determinate demand for information and the uncertainty of air-interface links. Mobile communications emerged in China in the 1980s [1]. At that time, the number of users in China was less than 18 000, and there was an abundance of spectrum resources. Thus, one of the primary manifestations of uncertainty in the air-interface links was signal distortion caused by uncertain channel fading, which results from path loss, obstacle shadowing, and multi-path effects. In the 1990s, to enhance the quality of communication signals, the second generation of mobile communications introduced a series of channel coding techniques, including block convolutional codes, cyclic redundancy check (CRC) codes, error correction cyclic codes, and parity codes [2]. These coding techniques were intended to establish a supervisory relationship for each codeword, enabling the receiver to identify the error location based on this relationship and subsequently correct it. Thus, the techniques led to a substantial improvement in the information error rates caused by uncertain fading.
As mobile communications underwent rapid expansion, particularly with the surge in the number of users, the scarcity of available spectrum resources and the complexity of the electromagnetic environment become increasingly significant, making uncertain interference another major challenge alongside uncertain fading. Interference and fading both result in signal distortion: Interference mainly originates from the electromagnetic disturbances of external communication systems, while fading stems from objective phenomena in electromagnetic wave propagation, including scattering, reflection, multi-path effects, and Doppler effects. Therefore, it is advantageous to employ more flexible and proactive strategies in response to uncertain interference. In the early 21st century, orthogonal frequency division multiplexing (OFDM) modulation technology was proposed for the fourth generation (4G) of mobile communications [3]. OFDM improves spectrum efficiency by maintaining the orthogonality among sub-carriers, greatly alleviating the interference issues caused by spectrum scarcity. In addition, the wireless spectrum allocation regime stipulated orthogonality between different communication systems, providing additional protection against uncertain interference. Since then, channel coding and signal modulation have been recognized as the two key technologies for addressing the uncertainty of air-interface links.
Throughout the developmental history of coding and modulation technologies, this development can be generalized as being rooted in the so-called “{0,1}” mindset, which is manifested in discrete decision-making at the technical level and stacked processing at the system level. At the technical level, classical channel coding and signal modulation technologies are based on fundamental assumptions, such as the Gaussian channel or binary symmetric channel (BSC). The key idea is to abstract the complex communication process into concise, intuitive, and analytically tractable mathematical and physical models, and to design corresponding technical paradigms. For example, in the design of low-density parity-check (LDPC) codes, there are two discrete states: the “1” state (where all information bits are protected by check bits) and the “0” state (where none of the information bits are protected). Similarly, quadrature amplitude modulation (QAM) technology is based on the assumption that constellation points are uniformly distributed and Euclidean distances transition abruptly between multiple discrete states. The “0” state indicates that the Euclidean distances between all constellation points are equal and remain constant, while the “1” state indicates that all Euclidean distances transition to the next state. At the system–component level, classical digital communication systems are modularly designed with channel coding, signal modulation, and other components. These components are relatively independent and perform their respective functions, and the communication processes are realized through simple module stacking (i.e., a simple combination of various components). As a result, improving the performance of an individual component comes at the expense of a loss of flexibility for other components. For example, in QAM systems, to achieve higher transmission rates using higher modulation orders, more check bits must be used to protect information bits, thereby sacrificing the flexibility to adjust check-bit relationships. This {0,1} design mindset simplified the complexity of the air-interface link construction in the early stages of mobile communications technology development and accelerated the research progress of channel coding and signal modulation, driving the rapid advancement of mobile communication technology. The {0,1} design mindset also efficiently combatted link uncertainty in communication scenarios with known and generic channel environments.
However, given the explosive growth in communication demands, the technologies for constructing air-interface links must be based on deep exploration and exploitation of the spatial, temporal, and spectral dimensions of communication resources to satisfy such demands. As a result, the uncertainty of links presents increasingly complex multi-domain coupling characteristics, such that the coding and modulation technologies based on the {0,1} mindset can hardly sustain the development of mobile communications. In the technical aspect, classical channel coding is based on common channel assumptions, employing uniform checks on information bits, which makes it difficult to address the disparity in error probabilities caused by channel variations over time. Moreover, the Gaussian assumption of independent interference across spatial, temporal, and spectral dimensions conflicts with the coupled characteristics of real-world channels, posing difficulties for QAM to adaptively deal with complex and uncertain interference. In the system–component aspect, there is an inherent contradiction between channel coding and signal modulation. More specifically, different distributions of check bits in traditional channel coding lead to varying symbol distortion probabilities in signal modulation, which causes some of the symbols to be error-prone. Similarly, the mapping methods and constellation order choices in traditional signal modulation result in different error probabilities for encoded information bits, making some of the bits error-prone. In summary, the classical {0,1} mindset does not sufficiently consider the continuity of bits and symbols, thus exhibiting apparent limitations. Therefore, flexibly adjusting the internal structure and parameters of channel coding and signal modulation is essential in order to establish a complementary technical framework, paving the way for new perspectives in the evolution of mobile communication technology.
In this paper, we introduce a novel mindset called “[0,1],” which is defined as a flexible configuration of key components in a communication system, covering all possible states in the continuous interval from “0” to “1” in coding and modulation. The aim is to provide guidelines for the design of more effective mechanisms for coding and modulation technologies. In contrast to the traditional {0,1} mindset, which confines the structure and parameters to a few discrete states, the proposed [0,1] mindset emphasizes a more flexible configuration for coding and modulation. We have devised mechanisms for continuous variation in codeword and symbol distributions to grant the air-interface links unprecedented dynamic adaptability, thereby breaking the previous isolated and independent operation pattern among the various components of air-interface link construction. More specifically, we propose finite-length channel coding and asymmetric signal modulation to achieve a flexible transition from {0,1} to [0,1]. The proposed technologies overcome the limitations caused by the rigid mechanisms accumulated from previous evolutions of technologies, enabling elastic collaboration among different components of air-interface link construction. This collaborative effort holds promise for more efficiently addressing issues arising from the uncertainty of links, such as fading and interference, particularly in scenarios with medium-to-short code lengths, thus offering a potential breakthrough for the future development of mobile communication technologies.
2. Future requirements for mobile communications
For the future of mobile communications, it is far more meaningful to reconsider the current source of demands than to rely solely on technology-driven approaches. This becomes especially apparent as we approach the sixth generation (6G) era, where novel services such as immersive cloud extended reality (XR), holographic communications, and sensory interconnectivity continue to emerge, characterized by their immersive and interactive nature. Moreover, emerging scenarios such as deep-sea, deep-space, and deep-earth communications present increasingly diverse and complex technical challenges. Meeting these requirements may not be achieved by simply extending and combining classical technologies. The key challenges lie in the technological complexities and uncertainty associated with achieving low-latency interaction, immersion, and comprehensive service coverage. In contrast, the development from 1G to 5G benefited from the commonality of scenarios inherent in the cellular network architecture. More specifically, the coding and modulation techniques for air-interface links have continuously evolved along with the development of mobile communication technologies. These advancements include the evolution of coding techniques from convolutional codes to turbo codes, LDPC codes, and polar codes; constellation modulation approaches from 4-QAM to 64-QAM; and transmission techniques from single-antenna to multi-antenna. It can be found that classical Shannon information theory is based on the basic premise that the characteristics of the channel environment are known. However, it is essential to deal with the challenge of the globalized uncertainty in 6G brought by the various emerging scenarios. Thus, future mobile communications should be demand-oriented.
In the following discussion, we provide some examples to further elaborate on the source of demands for 6G and the new technologies that need to be explored. With the development of services such as cloud XR, terminal devices are expected to become sleeker and smarter and to deliver more immersive user experiences. The arrival of the complete immersion era makes XR a promising technology in the fields of cultural entertainment, healthcare, education, and social production. Furthermore, with advancements in terminal display devices, future holographic information transmission will achieve naturalistic visual replication, enabling dynamic three-dimensional (3D) interaction among individuals, objects, and their environment. The integration of cloud-based XR and holography will liberate individuals from the constraints of time and space, allowing them to seamlessly transition between virtual and real environments. This integration erases the boundary between reality and virtuality, facilitating a novel immersive experience. However, achieving immersive interaction demands ultra-low latency transmission, necessitating extremely short code lengths. Existing lengths of coded groups with clear performance boundaries are generally above 200. However, when coding lengths drop to the magnitude of 100, traditional coding methods—which are guided by Shannon’s theorem and rely on adding redundancy to mitigate the effects of channel non-flatness and noise—not only increase the complexity but also considerably diminish the information-transmission efficiency. Obviously, classical coding theory is no longer suitable for the realm of short codes. Therefore, devising precise error-correction solutions for data with varying importance and achieving differentiated efficient transmission are critical aspects to be investigated in future mobile communications.
In addition, as user demands evolve, communication networks will encounter increasingly diverse and complex challenges in various environments, such as high-altitude, marine, and remote unmanned areas—especially extreme environments. In the region of deep space, the characteristics of the interstellar medium (e.g., abundance and isotope ratio) and the distribution of high-energy particles under the influence of the solar wind and interstellar magnetic field are ambiguous, leading to a substantial increase in the uncertainty of communication links. In the region of the deep ocean, due to the complex interaction between the deep ocean’s physical properties and its dynamics, the conditions (e.g., water temperature, salinity, and ocean currents) become more complex and dynamic. In the field of deep earth, the harsh environment of high earth stress, high osmotic pressure, high earth temperature, and mining disturbances, as well as the complex deformation nature of the geological structure, rock mechanical behavior, and other complex deformation properties, all lead to highly complex fading and interference in communication channels with the earth as the transmission medium. It is evident that extreme environments exhibit significant differences in fading levels, electromagnetic interference, radiation levels, medium parameters, and so forth, resulting in distinct channel characteristics. Conventional modulation designs, which are passive, patch-based, and incremental, are limited to offering modulation choices that suit singular physical fields and medium properties, adhering to the typical features of the discrete {0,1} mindset. Thus, they are unable to continuously adapt to signals from multiple physical fields and the varying propagation characteristics (e.g., wave velocity and attenuation) across different media, unlike designs based on the [0,1] mindset as demanded by new mobile communications. Hence, conventional designs encounter difficulties in addressing the increased complexity of link uncertainty in extreme environments.
In conclusion, to address the increasingly complex and diverse demands of future mobile communications, it is essential to flexibly design and configure coding and modulation. Therefore, leveraging the [0,1] mindset to extend the performance boundaries of coding and modulation is crucial for establishing an advanced air-interface link system.
3. General design framework of the [0,1] mindset
The [0,1] mindset enables a more flexible configuration for coding and modulation. In contrast to the traditional {0,1} mindset, which restricts structures and parameters to a few discrete states, coding and modulation techniques with the [0,1] mindset have all possible states of their key parameters, as if the parameters can take values in continuous intervals. Fig. 1 presents the general design framework of the [0,1] mindset. Based on the [0,1] mindset, we design a parity-check-concatenated (PCC) polar code with [0,1] flexible checksum relations, rather than the {0,1} full-check in the coding module, to provide differentiated protection for various information bits by portraying the uncertainty affecting each coded bit. Using the [0,1] mindset, we also design a generalized mapping-assisted golden angle modulation (GAM) in the modulation module whose modulation order can be any positive integer. This results in an adjustable Euclidean distance—that is, the [0,1] distance—between individual constellation points according to the interference strength of different symbols, instead of using fixed equal-interval Euclidean distances—that is, the {0,1} distance. The [0,1] mindset provides unprecedented flexibility in coding and modulation modules. Subsequent sections provide detailed applications of the [0,1] mindset in coding and modulation techniques.
4. Finite-length channel coding
Channel coding, as an effective error-control method, plays an indispensable role in mitigating uncertain fading in communication links. It is implemented by introducing redundant check bits into transmitted information for error correction. In 1967, Claude Shannon, the pioneer of information theory, identified two primary objectives for channel coding: achieving the Shannon’s bound for infinite-length coding and approaching this bound for finite-length coding. Over decades of development, channel coding has evolved from early techniques such as Hamming codes, convolutional codes, Bose–Chaudhuri–Hocquenghem (BCH) codes, and Reed–Solomon (RS) codes to more recent innovations such as turbo codes and LDPC codes, gradually enhancing the performance of infinite-length coding. Notably, turbo codes and LDPC codes are capable of nearing the Shannon bound, garnering significant attention from academia and industry. Turbo codes were even adopted as the technical standard for channel coding in third-/fourth-generation mobile communication systems. In 2009, Arikan’s polar code [4] was a groundbreaking advancement in channel coding, as it is the only structured coding method capable of theoretically reaching the Shannon bound with infinite code length. This shift in research focus toward achieving a coding performance close to the bound under practical finite code lengths marked a significant milestone in the field of channel coding.
For decades, achieving optimal performance in finite-length channel coding has remained a challenge, with a persistent 1 dB gap from the Shannon bound [5]. Overcoming this gap has been a global concern. In his seminal work on the mathematical theory of communication, Shannon [6] emphasized the necessity of appropriately introducing redundancy to combat the particular noise structure involved. In other words, channel coding needs to be designed with check bits in conjunction with channel characteristics. However, conventional channel coding primarily relies on Shannon’s theory, assuming bit channels to be independently and identically distributed, which is an idealization that overlooks the varying impact of uncertainty in air-interface links on bit channels.
Classical channel coding methods, such as the CRC polar code [7], the polarization-adjusted convolutional (PAC) code [8], and the classical polar code [4], typically adopt {0,1} binary check relationships, offering either full protection for all information bits or none at all, resulting in inflexible structures. While effective for infinite code lengths, such approaches struggle to approximate the desired error–correction performance under finite lengths due to varying probabilities of information errors. Hence, designing flexible check relationships in finite-length channel coding is essential to address the differential impact of link uncertainty.
While prior works have explored different combinations of coding-channel characteristics [9], they primarily focused on optimizing information mapping, without addressing the limitations imposed by a rigid check relationship. In response, our research began in 2005 with the proposal of the complementary block coding (CBC) model [10], [11]. CBC introduced the concept of providing different levels of protection to information bits by incorporating complementary bits into code groups and designing check relationships of varying strengths for them. This approach not only enhanced the error–correction performance but also reduced the peak-to-average power ratio in multicarrier modulated OFDM systems, hence mitigating the devices’ nonlinearity effects. Building upon this concept, we developed invertible subset LDPC coding [12], [13], [14], which differentiated the bit distribution within all symbols in modulation, leveraging LDPC’s superior error–correction capabilities.
Given that their performance at infinite code length can reach the Shannon bound, polar codes are a promising class of channel coding schemes. In our work, we address the impact of link uncertainty in polar code design, aiming to push the engineering implementation of polar codes closer to the Shannon bound. Based on the concept of flexible check relationships, we have proposed a PCC polar code [15]. The parity bits of the PCC polar code can fully utilize the under-polarized bit channels and can be flexibly distributed in the bit channels with high reliability to protect the information bits differentially and unevenly, thereby achieving a [0,1] design of the coding structure and enabling finite-length polar codes to approach the capacity limit. As shown in Fig. 1, by accounting for the impact of uncertainty on each coded bit (where K is the length of information bits and M is the length of parity bits), the PCC polar code offers varying levels of protection for different information bits . We initialize the check relationships based on modulation modes and encode the coded bits of PCC polar codes (where N is the codeword length and L is the number of codewords) accordingly. Receivers calculate the frame error rate (FER) and provide feedback to adjust check relationships adaptively. With these processes, optimal check relationships are obtained.
A comparison of the performance of our proposed PCC polar code with those of other schemes—that is, the CRC polar code [7], the PAC code [8], and Arikan’s classical polar code [4]—is shown in Fig. 2. All schemes are validated in a binary input additive white Gaussian noise (BI-AWGN) channel. A successive cancellation list (SCL) is adopted by all schemes, where the list size of the SCL decoder is set as ζ = 8. The number of parity bits of the PCC polar code is MPCC = 20, the number of CRC bits of the CRC polar code is MCRC = 8, the codeword length is N = 512, and the bitrate is R = 0.5. The comparative performance analysis demonstrates an improvement in the error–correction performance of our proposed PCC polar code of more than 0.5 dB compared with the classical polar code at FER = 5 × 10–3[16]. In particular, each check equation of the PCC polar code only involves a few simple XOR operations. Therefore, our proposed scheme increases the negligible encoding/decoding complexity compared with the classical polar code, where only M check equations are added to the computation.
It is evident that finite-length channel coding schemes based on the [0,1] mindset offer an effective means to achieve highly reliable information transmission. At present, our proposed PCC polar code has demonstrated the capability to tailor the coding structure to specific communication scenarios, resulting in significant improvements in error–correction performance. However, the complexity remains high, posing challenges for engineering implementation. With the anticipated proliferation of mobile communication scenarios in the future, diverse higher-order and higher-dimension modulation schemes with increasingly intricate constellation distributions and mapping relationships are expected, resulting in an increased demand for stringent coding performance. Consequently, the design and application of high-performance finite-length channel coding schemes are of great importance in practical systems.
5. Asymmetric signal modulation
Signal modulation entails converting one or more coded bits into symbols (where r is the number of symbols), which are then systematically mapped onto the points of a modulation constellation diagram. These symbols, characterized by their phase and amplitude, serve to distinguish various bit combinations and mitigate the impact of interference. The effectiveness of signal modulation in combating interference is visualized in the constellation diagram through the Euclidean distance between neighboring constellation points. A larger Euclidean distance indicates greater disparity in phase and amplitude, which is expected to enhance the system’s resilience against interference. However, the amplitude range of constellation points is finite, and a larger Euclidean distance between neighboring points implies fewer points within the constellation diagram, thereby reducing the number of representable symbols and consequently lowering the transmission rate. Thus, the design of an appropriate signal modulation scheme is crucial in ensuring the anti-interference capability of mobile communication systems while maintaining the transmission rate performance.
The constellation diagrams of conventional signal modulation schemes are typically designed based on the average interference strength experienced by all symbols over a given period. Consequently, the Euclidean distances between neighboring constellation points and modulation orders can only vary discretely in steps between multiple discrete states. In 1962, Bell Labs introduced QAM [17], which evenly distributes 2B (B denotes the bit length in a symbol) constellation points in a square format on the constellation diagram. Building upon this, in 1974, Foschini et al. [18] proposed equilateral triangular modulation to further enhance the anti-interference capability by adjusting the distribution of constellation points, employing equilateral triangles for all neighboring constellation points. Subsequently, in 1984, Forney and Wei [19] introduced cross modulation, which maximizes the Euclidean distance between neighboring constellation points within a limited range of constellation diagrams, thereby improving the anti-interference capability.
Since these signal modulation schemes adopt a regular symmetric distribution of constellation points, the Euclidean distances between neighboring constellation points are uniform, which restricts the variation of Euclidean distances to two discrete states in {0,1}—that is, maintaining equal and constant distances between neighboring points, or uniformly transitioning to another Euclidean distance corresponding to a different modulation order. This discretization design method based on the {0,1} mindset endows all symbols with uniform anti-interference capability, making it widely applicable in communication scenarios due to its simplicity and ease of implementation. However, in actual communication systems, due to device nonlinearities, frequency effects, and other factors, different modulation symbols are subject to varying degrees of interference and distortion. Consequently, modulation schemes based on the {0,1} mindset struggle to effectively address the disparate interference levels experienced by individual symbols.
Hence, we have proposed an asymmetric modulation scheme based on the [0,1] mindset that can flexibly adjust the Euclidean distance between each constellation point according to the interference intensity at different symbols. In 2011, we presented reshaped quadrature amplitude modulation (R-QAM) [20], which reshaped orthogonal amplitude modulation constellations by rotating the angles of specific constellation points in QAM. This involved retaining the constellation points with negative imaginary parts and distorting those with positive imaginary parts according to certain rules. Building upon this, we have further developed generalized mapping-assisted GAM [21] to achieve flexible modulation, where the Euclidean distance can be continuously varied, as is needed for [0,1]. In the constellation design, the mth constellation point (sm) is defined as follows:
where , j is the imaginary part, and is the complex amplitude of the mth constellation point. In order to adapt to the characteristics of the actual channel, we find the optimal mapping by minimizing the average bit error probability as follows:
where denotes the pairwise error probability (PEP), and denotes the Hamming distance between the transmitted symbol and the estimated symbol . Specifically, the bit-to-symbol mapping is accomplished by designing a preset threshold for the PEP, then generating the PEP matrix and giving higher priority to symbols with larger PEP during the mapping process, thereby reducing the Hamming distance of such symbols. It is important to highlight that the selection of the preset PEP threshold is a pivotal aspect of this process. In order to select as many larger PEP values as possible, the preset PEP threshold is designed to be a value within a certain range interval below the maximum PEP value, with the range of the interval depending on the specific form of constellation modulation. As the PEP is computed based on the statistical information of the channel, the preset PEP threshold does not generally require frequent adjustment to follow channel variations. Fig. 3 illustrates the discrepancy in the distribution of the constellation diagram and the mapping methods of 16-QAM and 16-GAM. The GAM’s constellation diagram demonstrates an irregular shape, presenting various Euclidean distances between neighboring constellation points, which realizes constellation modulation with symbols of different anti-interference capabilities.
Although some non-uniform modulation methods, such as amplitude phase-shift keying (APSK), use geometric shaping for constellation design, in which the Euclidean distances of neighboring constellation points are not exactly the same, these methods are usually designed under specific channel conditions such as additive white Gaussian noise (AWGN). Thus, they are still categorized as being part of the [0,1] mindset. However, the channel condition experienced by the modulated symbols is unknown in advance (e.g., there may be carrier frequency differences), and symbols with different amplitudes undergo the same fading with different degrees of distortion, which will result in a different probability of error. Since the bit error rate (BER) changes with the channel conditions, the constellation mapping method needs to be adjusted accordingly. Figs. 3(b) and (c) show the mapping of the proposed method under an AWGN channel and a Rayleigh channel, respectively, which demonstrates the flexibility of the proposed method in adapting to different channel conditions. In addition, the proposed generalized mapping-assisted GAM can realize constellation modulation with any positive integer order, which provides a higher degree of freedom. Fig. 3(d) demonstrates the application of the proposed method in orthogonal frequency division multiplexing with index modulation (OFDM-IM) [22]. It can be observed that the three subcarriers of OFDM can be transmitted with six and five GAM constellation points, respectively, which enables rate-adaptive modulation.
As shown in Fig. 4, at , the proposed generalized mapping-assisted GAM achieves performance gains of 0.6 and 0.9 dB compared with GAM using classical Gray mapping for modulation orders of 16 and 256, respectively. We also compare the proposed method with APSK. Fig. 4 shows the BER performance at modulation orders and . It can be seen that the proposed generalized mapping-assisted GAM achieves a performance gain of 6.8 and 9.8 dB compared with APSK for modulation orders of 16 and 256, respectively, at .
The computational complexity of the proposed mapping method is determined by three components: calculating the PEP, determining the mapped symbol index, and selecting the mapped bit index. Thus, the complexity is for a constellation diagram with M symbols, where is the number of possible candidates for the mth step in the mapping process. This complexity is much smaller than the complexity of all mapping combinations.
It can be clearly seen that an asymmetric modulation scheme based on the [0,1] mindset is effective in ensuring the modulation performance against channel interference. With the development of mobile communications, the demands for a high data rate are constantly increasing, necessitating modulation schemes with higher modulation orders and more effective anti-interference capabilities in the presence of link uncertainty. Currently, the optimal distribution of higher-order constellations, which does not rely on assumptions of the channel model and can cope with rapidly changing interference, remains to be investigated. With a powerful nonlinear fitting ability, machine learning algorithms can learn constellation distributions and mapping relations from training sequences, without requiring prior information on the channel interference distribution, allowing them to obtain the optimal solutions for different interference distributions. Thus, machine-learning-based approaches are expected to provide a more effective means of dealing with the challenges in asymmetric modulation.
6. Coding and modulation co-evolution
Considering the limitations of the technologies for air-interface link construction caused by stacked development, a new perspective on the relationship between coding and modulation is necessary. For example, for the design of polar codes in the context of 5G, if a certain bit is transmitted over a highly reliable polar subchannel but is assigned to a less-reliable modulation subchannel with higher-order modulation, the polarization effect of the polar code is weakened, resulting in degraded system reliability. Obviously, the mutual impact between coding and modulation in practical application scenarios cannot be overlooked. Therefore, the [0,1] mindset should be holistically considered from the perspectives of both channel coding and signal modulation. By considering the Euclidean space characteristics between codewords, the coding, and the modulation, the codeword construction and the constellation diagram design can be comprehensively optimized, making it possible to efficiently intertwine the codewords and the modulated symbols to enhance the overall performance of the communication systems.
More specifically, during signal modulation, the bits obtained after channel coding are amalgamated into symbols and then mapped onto the modulation constellation diagram. By considering the correlation of the interference experienced by the bits in each symbol, the necessary degree of protection for each bit can be predicted. This facilitates the design of a more appropriate error-checking relationship tailored to the channel environment. Moreover, signal modulation must factor in the impact of channel coding. Since the distribution of check bits in the coding phase varies, the distortion probabilities for different symbols are distinct. In addition, as the peripheral constellation points are more prone to nonlinear effects and distortion characteristics, it becomes feasible to devise a constellation distribution and symbol mapping methodology that aligns symbol error probabilities with the distortion levels of constellation points. This facilitates a more nuanced accommodation of diverse encoding schemes, thereby amplifying the transmission rates.
In addition to issues related to channel coding and signal modulation, the current air-interface link architecture was developed with a stacked pattern, encompassing elements such as source coding, large-scale multiple antennas, and other components. The most advanced manifestation and the ultimate pursuit of this evolutionary progression, termed the [0,1] evolutionary mindset, involves efficiently synergizing all stacked components of an air-interface link into a cohesive whole. However, synergizing algorithms and modules that are specifically tailored for each specific component results in significant complexity. On the other hand, the mere interweaving of modules in a simplistic design fails to achieve efficient synergy across all components. One prospective approach is leveraging the abstract representation capability of artificial intelligence technology to conceptualize all components within a traditional system as a deep neural network model, which would enable concise and efficient communications. Such efficient synergy among all components is anticipated to further elevate the overall performance of air-interface link construction, paving the way for future advancements in mobile communications.
Acknowledgments
The author expresses his gratitude to Prof. Lixia Xiao and Prof. Guanghua Liu for their invaluable support towards this paper. This work was supported by the National Key Research and Development Program of China (2019YFB1803400).
IskanderMF, YunZ.Propagation prediction models for wireless communication systems.IEEE Trans Microw Theory Tech2002; 50(3):662-673.
[2]
AbdulwahabWK, KadhimAA.Comparative study of channel coding schemes for 5G.In: Proceedings of the International Conference on Advanced Science and Engineering (ICOASE); 2018 Oct 9–11; Duhok, Iraq; 2018.
[3]
WeinsteinSB.The history of orthogonal frequency-division multiplexing [history of communications].IEEE Commun Mag2009; 47(11):26-35.
[4]
ArikanE.Channel polarization: a method for constructing capacity-achieving codes for symmetric binary-input memoryless channels.IEEE Trans Inf Theory2009; 55(7):3051-3057.
[5]
CostelloDJ, ForneyGD.Channel coding: the road to channel capacity.Proc IEEE2007; 95(6):1150-1177.
[6]
ShannonCE.A mathematical theory of communication.Bell Syst Tech J1948; 27(3):379-423.
[7]
TalI, VardyA.List decoding of polar codes.IEEE Trans Inf Theory2015; 61(5):2213-2226.
[8]
RowshanM, BurgA, ViterboE.Polarization-adjusted convolutional (PAC) codes: sequential decoding vs list decoding.IEEE Trans Vehicular Technol2021; 70(2):1434-1447.
[9]
SeidlM, SchenkA, StierstorferC, HuberJB.Polar-coded modulation.IEEE Trans Commun2013; 61(10):4108-4119.
[10]
JiangT, ZhuG.Complement block coding for reduction in peak-to-average power ratio of OFDM signals.IEEE Commun Mag2005; 43(9):S17-S22.
[11]
JiangT, ZhuG, ZhengJ.Block coding scheme for reducing PAPR in OFDM systems with large number of subcarriers.J Electron2004; 21(6):482-489.
[12]
QuD, LiL, JiangT.Invertible subset LDPC code for PAPR reduction in OFDM systems with low complexity.IEEE Trans Wirel Commun2014; 13(4):2204-2213.
[13]
ShuS, QuD, LiL, JiangT.Invertible subset QC-LDPC codes for PAPR reduction of OFDM signals.IEEE Trans Broadcast2015; 61(2):290-298.
[14]
XiaY, KongD, XinY, XiaoL, JiangT.Codes over GF(q) with alterable subset for PAPR reduction in OFDM systems.IEEE Commun Lett2022; 26(10):2262-2266.
[15]
WangT, QuD, JiangT.Polar codes with repeating bits and the construction by cluster pairwise error probability.IEEE Access2019; 7:71627-71635.
[16]
JiangT, LiuY, XiaoL, LiuW, LiuG.PCC polar codes for future wireless communications: potential applications and design guidelines.IEEE Wirel Commun2024; 31(3):414-420.
[17]
ForneyG, GallagerR, LangG, LongstaffF, QureshiS.Efficient modulation for band-limited channels.IEEE J Sel Areas Comm1984; 2(5):632-647.
[18]
FoschiniG, GitlinR, WeinsteinS.Optimization of two-dimensional signal constellations in the presence of Gaussian noise.IEEE Trans Commun1974; 22(1):28-38.
[19]
ForneyG, WeiL.Multidimensional constellations-part I: introduction, figures of merit, and generalized cross constellations.IEEE J Sel Areas Comm1989; 7(6):877-892.
[20]
LiC, JiangT, ZhouY, LiH.A novel constellation reshaping method for PAPR reduction of OFDM signals.IEEE Trans Signal Process2011; 59(6):2710-2719.
ZhaiX, XiaoL, LiuW, LiuG, XiaoP, JiangT.Golden angle modulation aided OFDM-IM for an arbitrary number of subcarriers.IEEE Commun Lett2023; 27(12):3380-3384.
PDF
(1374KB)
