Ultra-Dense Networks (UDNs) for 5G

By Jialing Liu, Huawei US Wireless Research and Standards, Weimin Xiao, Huawei US Wireless Research and Standards, Chih-Lin I, China Mobile Research Institute, Chenyang Yang, Beijing University of Aeronautics and Astronautics, Anthony Soong, Huawei US Wireless Research and Standards

IEEE 5G Tech Focus: Volume 1, Number 1, March 2017


This article gives an overview of the current state of UDN development and issues related to 5G RAN.

1. Emerging ultra-dense networks (UDNs)

Since the beginning of mobile industry, cell splitting and densification has been one of the most effective means to deliver ever-increasing capacity and improving user experience. In recent years, UDN has emerged as a prominent solution to meet the challenges of fulfilling IMT-2020 (5G) extremely high capacity density requirements of up to 10 Mbps/m2. Qualitatively, UDN is a network with much higher density of radio resources than that in current networks, i.e., much denser small cell network in terms of either relative density or absolute density of the BSs. Quantitatively, the definition of UDN varies among the literature. In [1-4], UDN is defined as a network where the BS (or AP) density potentially reaches or even exceeds the user density, which is appropriate to characterize the scenario when the traffic per user increases while the number of users does not. In [5], an UDN is characterized as a network where the inter-site distance is only a few meters. In [6], UDN is identified as a network reaching the point where its capacity grows sub-linearly, due to the growing impact of interference, as the BS density increases.

2. Challenges of UDNs

Interference in an UDN becomes more severe, with higher volatility, and there may be a large number of strong interferers but none dominant. This leads to interference statistics different from those of an existing network with one (or a small number of) dominant interferers [6]. Under the assumption of heavy and uniform traffic load, all the BSs are always active in conventional cellular networks. In these networks with sparsely deployed BSs, the density of users often exceeds the density of BSs, at least during peak time. For such sparse networks, universal frequency reuse has long been believed as optimal to maximize the capacity [7], and the assumption that every BS has at least one user to serve (and hence all BSs should be activated) is reasonable. In a universal frequency reuse sparse network with time division multiple access, the average SE of the network increase with BS density linearly. When the network becomes dense where some BSs have no user to serve but are still activated, the SE first increases slowly and then decreases with BS density [8], and hence the density of BSs can be optimized [8-9]. The assumption of a constant path loss exponent in these papers might mask the UDN effect [6], nonetheless they showed that interference should be handled differently in an UDN. More interesting behaviors of the network can be found in [6, 17] where networks with variable path loss exponents are studied.

Due to the traffic load fluctuation, turning off the BSs in the cells with low or no traffic load is an essential way for UDNs in improving EE as well as reducing interference. In practice, the network traffic fluctuates over different times and locations due to user behavior and mobility, which is especially true for UDNs and naturally calls for BS sleeping. In a universal frequency reuse sparse network with BS sleeping where BS density is less than user density, the average SE still increases linearly with BS density [10] as in the network without BS sleeping. In a universal frequency reuse UDN with BS sleeping where BS density is larger than user density, the SE only logarithmically increases with BS density [1].

 Utilizing the massive amount of radio resources optimally in an UDN becomes increasingly complex. Misallocation of increased radio resources can cause higher interference, unbalanced load distributions, and higher power consumption. Furthermore, due to interference, local radio resource allocation may have a global impact to a UDN. In other words, “locality” does not really exist in the UDN, and radio resource allocation has to be done based on a bigger picture of the UDN by taking into account of the tight coupling across the network [11].

Sufficient bandwidth over wired connectivity to directly backhaul each and every BS in an UDN may be practically infeasible. Wireless self-backhauling has been proposed, which consumes valuable radio resources, generates additional interference, and leads to extra latency.

3. Rethink universal frequency reuse

Universal frequency reuse has been made popular since the 3G era. It is also the common practice considered in UDN. Both the SE and EE are impacted by interference in such scenario. Complicated interference coordination in an UDN is undesirable due to the network scale and expensive backhaul. To manage the interference with less information sharing among the BSs, various semi-dynamic interference avoidance methods such as soft frequency reuse have been proposed. As the network becomes denser, i.e., as the ratio of BS density to user density increases, BS sleeping very effectively reduces the interference in the network. To further improve the average SE and EE of an UDN with BS sleeping, partial frequency reuse, i.e., with reuse factor greater than 1, was investigated [4]. When BS sleeping is allowed for the cells without active users, the frequency reuse factor that maximizes the SE or EE upper bound of the network with given ratio of BS density to user density was found, and the SE and EE gains of universal frequency reuse over partial frequency reuse in UDNs were quantified. It’s found that universal frequency reuse is SE-optimal for the networks with arbitrary BS/user density ratios, but is EE-optimal only when the ratio exceeds a threshold. This threshold is highly dependent on the total bandwidth of the network and the number of antennas at each BS. Both the normalized SE and EE gains of universal frequency reuse increase with the BS/user density ratio and slowly approach a constant that is dependent on the reuse factor.

4. Integrated resource allocation, interference management, and traffic steering

There are complex interactions among resource allocation, interference management, and traffic steering in an UDN. Joint considerations of all three to reshape interference and steer traffic load as desired bring forth techniques such as variable and flexible resource reuse patterns, load aggregation/balancing, enhanced UE-BS association, carrier selection and BS on/off, etc., in semi-static time scales (e.g., hundreds of milliseconds or longer). Optimal resource allocation taking into account of traffic distribution, interference, and performance requirements such as total latency have shown considerable improvement in network resource utilization efficiency, which in turns causes less interference and leads to higher SE [12]. Non-localized impact of interference in an UDN requires a large-scale optimization problem to be efficiently solved. Scalable algorithms have been pursued [11-13], including transforming a non-convex programming into a sequence of convex programming, and distributed decision making with network-wide iterations. Moreover, (sub-) optimal solution for an UDN may be pursued via optimal solutions for clusters of BSs by ignoring faraway interferers and considering cluster boundary constraints. Such solution for an UDN cluster of about 100 BSs and 1000 UEs can be obtained within seconds on a regular PC, which is applicable in a practical network.

5. Standards consideration

The above approach also leads to a fast-adapting network, in terms of its BS and carrier on/off status and traffic load redistribution. The cellular standards may consider supporting fast accessible carriers, including fast carrier on/off, fast carrier selection and switching, reduction or removal of always-on signals, streamlined measurement procedures, simplified connection establishment mechanisms, etc. [14] These features facilitate fast load balancing/shifting across BSs and carriers, as well as fast interference coordination and avoidance across BSs and carriers.

To address wireless in-band backhauling, an optimization study targeting the best end-to-end (i.e., multi-hop) performance and accounting for the split between backhaul links and access links was carried out in [15]. Performance benefits demonstrated by numerical results were encouraging, while extending the in-band backhauling without radio resource split or multi-hop will be more attractive.

Finally, in practice, an UDN is likely to be just part of a non-homogeneous network, or part of a hierarchical network. Since the RAN architecture in 5G is more revolutionary having introduced in the CU-DU structure [16], the realization of UDN may benefit from similar approaches of C-RAN.

6. Conclusions

Even without a uniquely agreed quantitative definition, UDN is expected to be an essential element of 5G networks for various deployment scenarios. In addition to implementation related technologies, standards related designs are necessary to realize its full potential. Comparing with the fast development of mmWave or mMIMO technologies, the progress of UDN specific work requires more attention and effort.


  1. Park, S. Kim, and J. Zander, “Asymptotic behavior of ultra-dense cellular networks and its economic impact,” IEEE GLOBECOM, 2014.
  2. Thurfjell, M. Ericsson, and P. de Bruin, “Network densification impact on system capacity” IEEE VTC Spring, 2015.
  3. Stefanatos and A. Alexiou, “Access point density and bandwidth partitioning in ultra dense wireless networks,” IEEE Trans. on Communications, Sept 2014.
  4. Su,C. Yang and C-L. I, "Energy and Spectral Efficient Frequency Reuse of Ultra Dense Networks." IEEE Trans. on Wireless Communications, Aug. 2016.
  5. Baldemair, T. Irnich, K. Balachandran, E. Dahlman, G. Mildh, Y. Selen, S. Parkvall, M. Meyer, and A. Osseiran, “Ultra-dense networks in millimeter-wave frequencies,” IEEE Communications Magazine, January 2015.
  6. Liu, W. Xiao, and A. C. K. Soong, “Dense networks of small cells,” in Design and Deployment of Small Cell Networks, A. Anpalagan, M. Bennis, and R. Vannithamby, Eds. Cambridge University Press, 2016.
  7. Ng, J. Evans, and S. Hanly, “On the capacity of cellular networks with global LMMSE receiver,” IEEE ICC, 2007.
  8. C. Li, G. Wu, and R. Q. Hu, “Analytical study on network spectrum efficiency of ultra dense networks,” IEEE PIMRC 2013.
  9. Li, J. Zhang, and K. Letaief, “Throughput and energy efficiency analysis of small cell networks with multi-antenna base stations,” IEEE Trans. Wireless Communications, May 2014.
  10. Yang and K. Sung, “Tradeoff between spectrum and densification for achieving target user throughput,” IEEE VTC’spring, 2015.
  11. Zhuang, D. Guo, and M. L. Honig, “Traffic-driven spectrum allocation in heterogeneous networks,” IEEE J. Sel. Areas Commun., 33(10), 2015.
  12. Liu and W. Xiao, “Optimal resource allocation in ultra-dense networks with many carriers,” Asilomar Conf. Signals, Systems, and Computers, 2015.
  13. Liu and W. Xiao, “Advanced Carrier Aggregation Techniques for Multi-Carrier Ultra-Dense Networks”, IEEE Commun. Mag., July 2016.
  14. 3GPP TR38.802 “NR Access Tech Physical Layer Aspects”, Mar. 2017
  15. Ge, Z. Zhou, J. Liu, and W. Xiao, “Spectrum allocation in cellular networks with wireless in-band backhaul”, submitted, 2016.
  16. 3GPP TR38.801 “Radio Access Architecture and Interfaces”, Mar. 2017
  17. G. Andrews et al., “Are we approaching the fundamental limits of wireless network densification”, IEEE Commun. Mag., Oct. 2016.


 Jialing Liu (This email address is being protected from spambots. You need JavaScript enabled to view it.) received the B.S. degree in mechanical engineering and the B.S. degree in automatic control, both from Tsinghua University, Beijing, China, in 1998, and the M.S. and Ph.D. degrees in electrical engineering from Iowa State University, Ames, Iowa, in 2002 and 2006, respectively. He joined Garmin International in 2005, Motorola Inc. in 2007, and Huawei Technologies in 2010. He is interested in communications, information theory, dynamical systems and control, and the interconnections among them. 



Weimin Xiao (This email address is being protected from spambots. You need JavaScript enabled to view it.) received the B.S. degree from Huazhong University of Science and Technology, Wuhan, China in 1992, the M.S degree from Tsinghua University, Beijing, China, in 1994, and the Ph.D. degree from Northwestern University, Evanston, Illinois, in 2001, all in electrical engineering. He joined Motorola Inc. in 2001, and Huawei Technologies in 2009. He is interested in communications and information theory, signal processing, and network optimizations.



Chih-Lin I (This email address is being protected from spambots. You need JavaScript enabled to view it.) received her Ph.D. degree in electrical engineering from Stanford University. She has been working at multiple world-class companies and research institutes leading the R&D, including AT&T Bell Labs; Director of AT&T HQ, Director of ITRI Taiwan, and VPGD of ASTRI Hong Kong. She received the IEEE Trans. COM Stephen Rice Best Paper Award, is a winner of the CCCP National 1000 Talent Program, and has won the 2015 Industrial Innovation Award of IEEE Communication Society for Leadership and Innovation in Next-Generation Cellular Wireless Networks.

In 2011, she joined China Mobile as its Chief Scientist of wireless technologies, established the Green Communications Research Center, and launched the 5G Key Technologies R&D. She is spearheading major initiatives including 5G, C-RAN, high energy efficiency system architectures, technologies and devices; and green energy. She was an Area Editor of IEEE/ACM Trans. NET, an elected Board Member of IEEE ComSoc, Chair of the ComSoc Meetings and Conferences Board, and Founding Chair of the IEEE WCNC Steering Committee.

She was a Professor at NCTU, an Adjunct Professor at NTU, and currently an Adjunct Professor at BUPT. She is the Chair of FuTURE 5G SIG, an Executive Board Member of GreenTouch, a Network Operator Council Founding Member of ETSI NFV, a Steering Board Member of WWRF, the ComSoc Rep of IEEE 5G Initiative, a member of IEEE ComSoc SDB, SPC, and CSCN-SC, and a Scientific Advisory Board Member of Singapore NRF. Her current research interests center around “Green, Soft, and Open”.

Chenyang Yang (This email address is being protected from spambots. You need JavaScript enabled to view it.) received her Ph.D. degrees from Beihang University, China, in 1997. She has been a full professor with the university since 1999. She has published over 200 papers and filed over 80 patents in the fields of energy efficient transmission, CoMP, interference management, cognitive radio, and relay, etc. She was supported by the 1st Teaching and Research Award Program for Outstanding Young Teachers of Higher Education Institutions by Ministry of Education of China during 1999-2004. Her recent research interests lie in green radio, ultra-dense networks, URLLC and wireless caching.



Anthony C. K. Soong (S’88-M’91-SM’02-F”14, This email address is being protected from spambots. You need JavaScript enabled to view it.) received the Ph.D. degree in electrical and computer engineering from the University of Alberta. . He is currently the Chief Scientist for Wireless Research and Standards at Huawei Technologies Co. Ltd, in the US and Huawei global head of open source networking. His research group is actively engaged in the research, development and standardization of the next generation cellular system. He currently serves on the board of OPNFV and the Engineering College Industrial Advisory Board for the University of North Texas. He had served as the chair for 3GPP2 TSG-C NTAH (the next generation radio access network technology development group) from 2007-2009 and vice chair for 3GPP2 TSG-C WG3 (the physical layer development group for CDMA 2000) from 2006-2011. Prior to joining Huawei, he was with the systems group for Ericsson Inc and Qualcomm Inc. His research interests are in statistical signal processing, robust statistics, wireless communications, spread spectrum techniques, multicarrier signaling, multiple antenna techniques, network virtualization, SDN and physiological signal processing. 

Dr. Soong is a Fellow of the IEEE. He has published numerous scientific papers and has more than 50 patents granted with another 30+ patents pending. He received the 2013 IEEE Signal Processing Society Best Paper Award and the 2005 award of merit for his contribution to 3GPP2 and cdma2000 development. He was on the advisory broad of 2014 IEEE Communication Theory Workshop and has served on the technical program committee, as well as, chaired at numerous major conferences in the area of communications engineering. He has acted as guest editor for the IEEE Communications Magazine and IEEE Journal on Selected Areas in Communications.

Editor: Geoffrey Ye Li 

Download the article (PDF, 154 KB)  


Subscribe to Tech Focus

Join our free IEEE 5G Technical Community and receive IEEE 5G Tech Focus.

Subscribe Now


Article Contributions Welcomed

If you wish to have an article considered for publication, please contact Amine Maaref, Managing Editor, at amine.maaref@gmail.com or Siming Zhang, Associate Managing Editor at simingzhang89@gmail.com

Author guidelines can be found here

IEEE 5G Tech Focus Editorial Board

Geoffrey Li, Editor-in-Chief
Amine Maaref, Managing Editor
Siming Zhang, Assoc. Managing Editor
Imran Shafique Ansari
Anwer Al-Dulaimi
Stefano Buzzi
Yunlong Cai
Rose Qingyang Hu
Chih-Lin I
James Irvine
Meng Lu
Paul Nikolich
Shugong Xu
Haijun Zhang
Anding Zhu