PhD Thesis Defense Radio Resource Management in OFDMA-based Cellular Relay Networks

1. PhD Thesis Defense Radio Resource Management in OFDMA-based Cellular Relay Networks By Mohamed A. Rashad Salem, M.Sc. Supervisors: Prof. Halim Yanikomeroglu Prof. David Falconer Department of Systems and Computer Engineering, Carleton University, Ottawa, Canada January 10, 2011

2. Backbone Connection BS BS UT RS Background Chapter 1 • The main objective of next-generation wireless networks (NGNs) is to • accommodate the continuously increasing demand; • provide ubiquitous high-data-rate services comparable to those offered through the wired networks; • in a cost-efficient manner • Orthogonal frequency division multiple access (OFDMA) is the envisioned air interface • Multihop relaying is a promising cost-efficient radio-access-network (RAN) solution for NGNs such as LTE-A , IEEE 802.16j and 802.16m

3. Background Ch. 1 • A future network comprising various forms of dedicated relays is envisaged (e.g., fixed, nomadic, and mobile). • The synergy of relaying and OFDMA techniques offers a very rich set of opportunities yet renders a challenging environment • Intelligent radio resource management (RRM) schemes are required to • exploit the opportunities offered by such networks • operate the relays in a dynamic and opportunistic manner • combat the co-channel interference (CCI) associated with the inevitable aggressive resource reuse • It is not immediately clear how to perform RRM in such a complex environment  RRM could be the main obstacle confronting the deployment and operation of NGNs

4. Observations: Existing approaches in literature Ch.1 • Single-cell models  no inter-cell interference  results can not be extended to multi-cellular networks • Resource partitioning  requires planning; limited multi-user / frequency diversity • Static routing or relay selection  does not exploit the spatial diversity • Decoupling of routing and scheduling  suboptimal • Traffic- or queue-unawareness  does not exploit the statistical multiplexing; inefficient utilization • Fairness is often not considered  location and channel dependent service; defeating ubiquity • Usually difficult to accommodate multiple service classes (heterogeneous traffic) • High feedback overhead and/or complexity • Oversimplified channel models  do not exploit the inherent frequency diversity

5. BS FRS NRS FRS WT1 FRS NRS overhears NRS overhears WT2 Organization & Outline of Thesis Chapters 2 to 5 Multicellular/multi-user/multiple DF relays Network-level distributed RRM • Chapter 2 • Cell-level centralized RRA • Fixed relays (2-hop simple relaying) • Symmetric traffic • Throughput-optimal formulation • Dynamic joint routing & scheduling • Quasi full-duplex protocol • Fairness/service ubiquity • Cell-load balancing • Chapter 4 • Fixed relays (2-hop simple relaying) • Self-optimized WT-based routing • Decoupled routing & scheduling • Nomadic RSs (cooperative selective) • Self-optimized NRS MAC • Opportunistic intra-cell reuse • Tunable fairness/ubiquity of min.QoS High density of RSs Coexistence of FRSs & NRSs Self-organization/decentralized RRM • Chapter 3 • Half duplex relaying • Symmetric& asymmetric traffic • Generalized/separated formulations • Absolute & relative fairness/ubiquity • Cost assessment of queue-awareness • Chapter 5 • NRS joint power and subchannel alloc. • Subcahnnel packing approach • No-packing approach • Less CCI to BS’s transmissions • Green wireless/extended battery life

6. Chapter 2 • Centralized schemes are ‘optimal’ in principle and constitute performance benchmarks for their decentralized derivatives. • Distributed schemes are more attractive due to the complexity and signaling overhead concerns despite the potential performance losses • Therefore, this chapter addresses the following interesting questions: • Can we devise dynamic and efficient centralized RRM schemes with low complexity and limited feedback overhead? • With relay assistance, can the desired user fairness be attained through the channel-aware only fair schemes, e.g., proportional fair scheduling (PFS) ?

7. Fairness-aware RRM in Downlink OFDMA Cellular Relay Networks Ch.2 Our cellular model Transmission protocol of multicommodity flows in the multihop mesh networks

8. Definition of the demand metric of RSm on subch. n links K+M max K K … … … • Definition of the demand metric of the BS on subch. n 1 m M … … … N n BILP Mathematical Formulation of the RRA at the BS Ch.2 To obtain a low-complexity solution, the 3-D assignment space is projected onto a 2-D space through the max function. Note: At any subchannel-node pair (n,m), the potential links have no interdependency. The problem is then solved iteratively as follows

9. The Low-complexity Iterative Algorithm Ch.2 Complexity:

10. The ‘Constrained Routing’ Mode of Operation Ch.2 • Routing is performed dynamically and jointly with the resource allocation using the maximum differential backlog • This generic mode of operation is ‘open routing' butRSs do not exchange the data  Two hops only  Initial accumulation of data may occur at some RSs with poor links to the UT  Differential backlog prevents further forwarding mistakes  Learning Ability • Another practical mode of operation for the dynamic routing strategy, named ‘constrained routing’, is examined to demonstrate its learning ability: • Constraints on BS-RS transmissions accounting for the geographical distribution of the RSs and UT locations • Faster routing convergence is expected due to fewer forwarding mistakes • The improvement comes along with substantial savings in feedback overhead due to the eliminated links

11. A PFS-based RRM Relay-aided Reference SchemeCh.2 • Partition the UTs into clusters around the serving nodes • Partition resources among nodes in proportion to the numbers of their connected UTs • Allow individual nodes to perform PFS as adopted in OFDMA-based systems At each RSm, subchannel n is assigned to an active user k* At the BS, subchannel n is assigned to j* ; either an active direct user or an active RS active ≡ has data buffered at the source node

12. Results: User throughput and ubiquity Ch.2 Time-average user throughput as function of user location and shadowing with 25 UTs/cell using 3 and 6 RSs. Fittings provide the distance-based conditional mean of user throughput.

13. Results: Throughput CDFs & Open routing vs. constrained Ch.2

14. Results: Average total cell throughput & Fairness Ch.2

15. Results: User fairness (long-term) & Load balancing Ch.2

16. Contributions Ch.2 • A novel joint routing and scheduling algorithm is proposed for OFDMA-based cellular relay networks • A ubiquitous location-independent service is demonstrated • The algorithm ensures short- as well as long-term fairness among users, including cell-edge users • The algorithm exploits the opportunities in multi-user, frequency, spatial, and traffic diversities • Simulation results prove the learning ability and the efficiency of the routing strategy • Cell-load balancing is achieved jointly with the RRA • Substantial savings in complexity and feedback overhead in contrast to traditional centralized schemes • Patent filings of this work have been made on different levels: Korea, US, and international

17. Chapter 3 • We have studied the performance of throughput-optimal scheduling as applied to the cellular setup under symmetric traffic and quasi full-duplex relaying. • Despite the significant performance returns the joint algorithm in Ch. 2 has achieved, we observe that it suffers from a performance limiting bottleneck as the traffic load increases • In this chapter, two variants of an efficient half-duplex relaying (HDR) scheme are developed to improve the capacity and practicality of the scheme in Ch. 2 • What about asymmetric traffic? Frame structure for Variants A & B In Variant-A, only RSs transmit in the second portion of the DL frame

18. Fair RRA Towards Ubiquitous Coverage with Asymmetric Traffic Ch.3 The Joint Routing and Scheduling for the BS Sub-frame

19. Formulation of the RS Sub-frame Ch.3 Variant-A Variant-B

20. Results - I: Cell-edge user throughput & total cell throughput Ch.3 Symm. Traffic Impact of the number of RSs and the associated spatial diversity on the lower tail and the cell-edge

21. Results - I: User fairness (time-average & long-term) Ch.3 Symm. Traffic

22. Results -II: User throughput and service ubiquity Ch.3 Asymm. Traffic

23. Results - II: CDF of class-based time-average throughput Ch.3 Asymm. Traffic

24. Results - II: Absolute & Relative fairness (time-average) Ch.3 Asymm. Traffic

25. Contributions Ch.3 • A novel generalized formulation of the joint policy is proposed for the more HDR relaying operation (Asym. Traffic/Constrained Routing/Effect of ARQ/with or w/o LB ) • The network capacity (users/traffic load) has been increased significantly as compared to the reference quasi-FDR scheme. • A more ubiquitous and location-independent fair service is demonstrated under higher loads • Traffic diversity and queue-awareness are better exploited • At low to moderate loads, the quasi-FDR is more adequate for delay-insensitive applications as it incurs less complexity • Without empirical priority weights, our efficient implementation of throughput-optimal scheduling achieves a ubiquitous and fair service within each class of users (with symmetric traffic) and across classes of asymmetric traffic in a relative sense, on the time-average and long-term time scales. • Load-balancing could be maintained among the active RSs. • Patent filings of this work have been made on different levels: Korea, US, and international

26. Chapter 4 • A future network comprising a plethora of dedicated wireless relays is envisaged • RRM schemes should • Exploit the characteristics of different types of wireless relays • Facilitate their coexistence and different functions • Enable aggressive and opportunistic spatial reuse rather than the traditional static reuse patterns • Decentralized RRM have to be considered • Self-Organizing Networks (SONs): A self-organizing network entity possesses one function or more of: Self-configuration, Self-optimization Self-healing

27. Integrating Self-Organizing NRSs into the Fixed-Relay Network Ch.4 Wireless Terminal Relay Station BS sub-frame FRS sub-frame BS BS sub-frame reuse Traffic1 Traffic2 FRS NRS FRS WT1 FRS NRS overhears NRS overhears WT2

28. The WT-based Self-Optimized Dynamic Routing (link Selection)Ch.4 It is assumed that the quality of any subchannel of a BS-FRS feeder is higher than or equal to that of a second hop from the FRS to its connected WT Sort N achievable rates (descend) Mean value of chosen percentile

29. FRS-NRS Cooperation Ch.4 The NRS assists its associated WT on the data segments it received reliably while the WT received in error. Here, we assume that the data segment is a subcarrier. However, the idea is quite generic and applicable to larger resource units, e.g., subchannel.

30. The Semi-Centralized Scheme: Second hops are reported via the FRS feeder Ch.4

31. The Distributed SchemeCh.4 Each FRS masks its connected WTs and appear as a large WT with collective minimum rate requirements. A column with satisfied minimum rate is disabled in the BS’s RRA matrix.

32. Sample Results: User throughputCh.4

33. Sample Results: Normalized histograms for network reuse factorCh.4 Higher values of reuse factor are realized as the number WTs and/or FRSs increases

34. Contributions Ch. 4 • We devise a self-optimizing user-based routing strategy that significantly reduces the feedback overheadand achieves interference avoidance • We introduce a self-optimizing opportunistic medium access (MAC) technique by which NRSs autonomously acquire resources • Through that MAC technique, smart intra-cell reuse is achieved in contrast to the static reuse patterns • We devise a protocol to enable the cooperation between an NRS and an FRS to assist troubled WTs using selective relaying • We introduce novel limited-feedback heuristic decentralized RRA algorithms with tunable fairness of the minimum QoS • Thus, we establish the concept of nomadic relay-augmented fixed relay networks. • A patent filing of this work has been made in Korea

35. Chapter 5 • Power control (PC) is an important interference combating mechanism thereby constituting a means for improving a network's performance metric such as throughput. • In addition, PC could also be employed to ensure that a wireless transmitter uses just enough power to facilitate its operation without sacrificing system performance. • In this chapter we further investigate the schemes proposed in Chapter 4 by devising a novel joint power and subchannel allocation algorithm at the NRS. • Interestingly, the adaptive PC mechanism, unlike the majority in literature, is realized in an open-loop form as no feedback is required by the NRS to function. • We address the following questions: • How much power savings can be obtained through NRS power control? • How much is the throughput gain? • Is there any throughput-power savings trade-off?

36. NRS Joint Power and Subchannel Allocation Algorithm Ch. 5 1) The No-packing Approach • Low complexity • Potentially, more acquired subchannels The ‘no-packing’ approach: Potentially consumes more reuse subchannels

37. NRS Joint Power and Subchannel Allocation Algorithm Ch. 5 1) The Subchannel Packing Approach • More complexity • Least number of acquired subchannels The subcarrier ‘packing’ approach: Least number of reuse subchannels yet some subcarriers enjoy more than necessary transmit power

38. NRS Joint Power and Subchannel Allocation Algorithm Ch. 5

39. NRS Total Power Saving Gain: A throughput-power saving tradeoff as fn(#FRSs) Ch.5 15 WTs & 15 NRSs Median Saving Gain: With 6 FRSs: 6.44 dB With 3 FRSs: 10.98 dB

40. Instantaneous Direct User Throughput Ch. 5 15 WTs & 15 NRSs with 6 FRSs per cell

41. Contributions Ch. 5 • We present a novel NRS-directed joint power-subchannel allocation algorithm for the OFDMA-based nomadic-augmented fixed relay networks. • The algorithm performs adaptive power control (APC) within the autonomous opportunistic NRS medium access and channel reuse, using two different approaches • The APC mechanism is realized in an open-loop manner requiring no feedback from the WT • We identify a throughput-power saving tradeoff in terms of the number of deployed FRSs. • While no other work in the literature has addressed such systems and architecture, we further sustain the concept of nomadic relay-augmented fixed relay networks • A patent filing of this work has been made in Korea

42. Publications & Patent Filings Refereed Journal Papers Chapter 1: • Mohamed Salem, Abdulkareem Adinoyi, Mahmudur Rahman, Halim Yanikomeroglu, David Falconer, Young-Doo Kim, Eungsun Kim, and Yoon-Chae Cheong, “An overview of radio resource management in relay-enhanced OFDMA-based networks,” IEEE Communications Surveys and Tutorials, vol.12, no.3, pp.422-438, Third Quarter 2010. • Mohamed Salem, Abdulkareem Adinoyi, Halim Yanikomeroglu, and David Falconer, "Opportunities and Challenges in OFDMA-Based Cellular Relay Networks: A Radio Resource Management Perspective," IEEE Transactions on Vehicular Technology, vol.59, no.5, pp.2496-2510, June 2010. • Mohamed Salem, Abdulkareem Adinoyi, Mahmudur Rahman, Halim Yanikomeroglu, David Falconer, and Young-Doo Kim, "Fairness-aware radio resource management in downlink OFDMA cellular relay networks," IEEE Transactions onWireless Communications, vol.9, no.5, pp.1628-1639, May 2010. • Mohamed Salem, Abdulkareem Adinoyi, Halim Yanikomeroglu, and David Falconer, “Fair resource allocation towards ubiquitous coverage in OFDMA-based cellular relay networks with asymmetric traffic,"accepted to IEEE Transactions on Vehicular Technology, January 2011. Chapter 2: Chapter 3:

43. Publications & Patent Filings • Mohamed Salem, Abdulkareem Adinoyi, and Halim Yanikomeroglu, “Integrating self-organizing nomadic relays into OFDMA fixed-relay cellular networks", submitted to IEEE Transactions on Mobile Computing, submission no. TMC-2010-10-0487, 19 October 2010. • Mohamed Salem, Abdulkareem Adinoyi, and Halim Yanikomeroglu, “Joint power and subchannel allocation for the self-organizing nomadic relays in OFDMA-based cellular fixed-relay networks," (interim title), to be submitted to IEEE Transactions onWireless Communications, January 2011. • Refereed Conference Papers: • Mohamed Salem, Abdulkareem Adinoyi, Halim Yanikomeroglu, and Young-Doo Kim, “Nomadic relay-directed joint power and subchannel allocation in OFDMA-based cellular fixed relay networks”, IEEE Vehicular Technology Conference(VTC2010-Spring), 16 – 19 May 2010, Taipei, Taiwan. • Mohamed Salem, Abdulkareem Adinoyi, Halim Yanikomeroglu, and Young-Doo Kim, “Radio resource management in OFDMA-based cellular networks enhanced with fixed and nomadic relays”, IEEE WCNC 2010, 18 – 21 April 2010, Sydney, Australia. • Mohamed Salem, Abdulkareem Adinoyi, Halim Yanikomeroglu, David Falconer, and Young-Doo Kim, “A fair radio resource allocation scheme for ubiquitous high-data-rate coverage in OFDMA-based cellular relay networks,” IEEE Globecom 2009, 30 November – 4 December 2009, Hawaii, USA. • Mohamed Salem, Abdulkareem Adinoyi, Mahmudur Rahman, Halim Yanikomeroglu, David Falconer, Young-Doo Kim, and Eungsun Kim, “Fairness-aware joint routing and scheduling in OFDMA-based multi-cellular fixed relay networks”, IEEE ICC 2009, 14 – 18 June 2009, Dresden, Germany. Chapter 4: Chapter 5:

44. Publications & Patent Filings Patent Filings: 1- Method for Relaying Data in Wireless Network and Personal Relay of Enabling the Method, and Mobile Device for Communicating with the Personal Relay. Mohamed Salem, Abdulkareem Adinoyi, Halim Yanikomeroglu, and Young-Doo Kim; filed by Samsung, Korea; Korea patent application no: P2009-0084026 (application date: 07 September 2009). 2- Method for Performing Fair Resource Allocation in OFDMA-based Relay Networks. Mohamed Salem, Abdulkareem Adinoyi, Halim Yanikomeroglu, David Falconer, and Young-Doo Kim; - Filed on 16 March 2009 in Korea (filing number: P2009-0022132); - Filed in USA on 27 September 2009 (application number is 12/567,776); - International application is under process. 3- Apparatus and Method for Allocating Subchannels and Controlling Interference in OFDMA Systems. Mohamed Salem, Abdulkareem Adinoyi, Mahmudur Rahman, Halim Yanikomeroglu, David Falconer, and Young-Doo Kim; - Filed on 11 June 2008 in Korea (filing number: P2008-0054726); - Filed on 22 December 2008 in USA (United States Patent Application No. 12/341,933); - Filed internationally on 23 April 2009 (international application No. PCT/KR2009/002119).

45. Summary of Thesis Contributions • The opportunities and challenges in the next-generation relay networks have been identified and discussed from an RRM perspective in Ch.1. A comprehensive literature review has been also provided. • A study of the performance of throughput-optimal policies in the OFDMA cellular relay network has been conducted in Ch.2, while retaining the traditional frame structure, through a novel formulation achieving joint routing and scheduling, service ubiquity and load balancing. Unlike existing centralized schemes, substantial savings in complexity and feedback overhead have been achieved. • A generalized throughput-optimal formulation has been proposed in Ch.3 for the practical HDR protocol capturing asymmetric traffic, constrained routing and effect of ARQ with and without load balancing. We show that service differentiation is attained through the queue length weightsrather than preset weights. Cost assessment of queue-awareness has been provided. • Our pioneer decentralized schemes presented and evaluated in Ch.4 have considered modern architecture comprising high density of relays of different characteristics and functionalities. Novel self-optimized terminal-based routing and NRS medium access have been devised achieving substantial savings in feedback and opportunistic intra-cell reuse, respectively. • We present a novel NRS-directed joint power-subchannel allocation algorithm in Ch.5 performing adaptive power control within the autonomous opportunistic NRS medium access and channel reuse, using two different open-loop approaches (requiring no feedback from the WT). A throughput-power saving tradeoff has been demonstrated in terms of the number of deployed FRSs.

46. Chapter 6 Possible Extensions • Extensions to the Centralized Schemes in Chapter 2 and Chapter 3 • Intra-cell Resource Reuse • Decentralization of the Proposed Schemes • Connection Admission Control for Load Balancing via Inter-cell Routing • Partial Feedback: How Much and How Often is Enough? • Extensions to the Distributed RRM in Chapters 4 and 5 • The WT-based Self-optimized Dynamic Routing: What Are the Proper Averaging Percentile and Time Window? • Characterization of the FRS Feeder Link Support to the Allocated Rate on the WT's Access link for Self-optimized Routing • Performance of the PC Mechanism with Subcarrier Packing • Development of the NRS Medium Access and PC Techniques in Chapters 4 and 5 Towards Femtocells • Terminal-to-terminal Relaying and Ad-hoc Networks

47. Thank you !