Throughput-Guaranteed Resource-Allocation Algorithms for Relay-Aided Cellular OFDMA System

1. Throughput-Guaranteed Resource-Allocation Algorithms for Relay-Aided Cellular OFDMA System 1Megumi Kaneko, 2Petar Popovski, and 1Kazunori Hayashi 1 Graduate School of Informatics, Department of Systems Science, Kyoto University (京都大學), Japan 2 Department of Electronic Systems, Aalborg University (奧爾堡大學), Denmark < IEEE Transactions on Vehicular Technology, vol. 58, no. 4, MAY 2009 >

2. Outline • Introduction • System model • Goal • Proposed Resource-Allocation Algorithms • Single-Relay Case: FTD and ATD • Multiple-Relay Case: MRPA and MRAA • Performance • Conclusion

3. Relayed link Relayed link Introduction • Installing relay stations in strategic positions in a cell • higher data rates can be provided in remote or shadowed areas of the cell • low-cost devices that can easily be deployed Direct link BS MS RS

4. Introduction • This paper investigates the problem of resource allocation for a relay-aided cellular system based on OFDMA. • This paper focuses on the downlink (DL) transmission from a BS to mobile stations (MSs) or RS in a single cell. Freq. BS-Subframe (TBS) BS–MSorBS–RS RS-Subframe (TRS) RS–MS Time TF

5. System model • Relay station • stores the received packets • decodes the received packets • re-modulates the received packets • Assume that packets sent to the RS in a frame cannot immediately be forwarded due to hardware limitations • The data for relayed users takes two frames to be delivered. • The data for direct users takes one frame.

6. System model • MSs feed back to the BS their CSIon every subchannel. • CSI = Channel-State Information • Assumption that BS knows the achievable rate rk,n • for each user k on subchannel n • Path Selection • A user is linked to the RS only if rkRS-MS 2 rkBS-MS ; • Otherwise, it is linked to the BS.

7. Goal • The BS-subframe is shared between the • direct users and BS-RS links • If the BS forwarded all the packets for the relayed users as they arrive in the BS queue, there will be less resource that is available for the direct users. direct users Freq. relayed users BS-Subframe (TBS) BS–MSANDBS–RS RS-Subframe (TRS) RS–MS Time

8. Goal • The target of this work is to design an algorithms with good throughput and outage performance. • Single-Relay Case • Multiple-Relay Case r1 MS1 BS RS r2 MS2 r1 > r2 Outage !

9. Proposed Resource-Allocation Algorithms • RS makes its own initial allocation to minimize the outage. • BS optimizes the final allocation to improve the overall throughput. • Single-Relay Case: FTD and ATD • Multiple-Relay Case: MRPA and MRAA BS-Subframe (TBS) BS–MSorBS–RS RS-Subframe (TRS) RS–MS TF/2 TF/2

10. Single-Relay Case: FTDFixed Time-Division Algorithm • RS makes its own initial allocation to minimize the outage. • rk,n : achievable rate on subchannel n for user k • βk (t–1) : previous average rate for user k • R : the minimum data rate requirement for user k BS-Subframe (TBS) BS–MSorBS–RS RS-Subframe (TRS) RS–MS Average rates are higher than their required rates TF/2 TF/2

11. Single-Relay Case: FTDFixed Time-Division Algorithm • RS makes its own initial allocation to minimize the outage. • rk,n : achievable rate on subchannel n for user k • βk (t–1) : previous average rate for user k • R : the minimum data rate requirement for user k BS-Subframe (TBS) BS–MSorBS–RS RS-Subframe (TRS) RS–MS Average rates are lower than their required rates TF/2 TF/2

12. Single-Relay Case: FTDFixed Time-Division Algorithm • If user k has the higher φand its packets are queued at the RS, the user kis serviced first by RS. • If user k has the higher φ than the allocated one butits packets are not queued at the RS, RS sends a Request Message to the BS. • User k set UReq • φmax : The maximum value of φ for user in UReq BS-Subframe (TBS) BS–MSorBS–RS RS-Subframe (TRS) RS–MS TF/2 TF/2

13. Single-Relay Case: FTDFixed Time-Division Algorithm BS DL Queue: MS3, MS7 RS DL Queue: MS1, MS2 φmax φ3 MS1 MS2 MS3 MS4 … φmin UReq φ1= 500 φ2= 600 φ3= 900 φ4= 400

14. Single-Relay Case: FTDFixed Time-Division Algorithm • BS optimizes the final allocation to improve the overall throughput. • BS calculates the number of sub-channels nBR that are required to send all the packets that are queued at the BS of the users in UReq. • The φ metric of the direct user is compared with φmax, and the sub-channel n is allocated to the link with the highest value. UReq (φmax) (400) φ direct user BS-Subframe (TBS) BS–MSorBS–RS RS-Subframe (TRS) RS–MS φmax BS RS (950) TF/2 TF/2

15. Single-Relay Case: FTDFixed Time-Division Algorithm • Channel utilization: Number of allocated packets for user k BM / RM Number of allocated time slots for user k BS-Subframe (TBS) BS–MS (BM)orBS–RS RS-Subframe (TRS) RS–MS (RM) TF

16. Single-Relay Case: FTDFixed Time-Division Algorithm • Throughput: 1 or 0 Number of allocated time slots for user k

17. Single-Relay Case: ATDAdaptive Time-Division Algorithm • Starting from the allocation by the FTD algorithm for TBS = TRS = TF /2, time division can be adapted to increase the overall throughput. BS-Subframe (TBS) BS–MSorBS–RS RS-Subframe (TRS) RS–MS TF/2 TF/2

18. Direct link Relayed link Multiple-Relay Case: MRPAMultiple-RS Parallel with Activation Algorithm • Path Selection (user k is on direct or relayed link) • User id linked to the RS only if ;otherwise, it is linked to the BS. Due to • The data for relayed users takes at least two frames to be delivered. • The data for direct users takes only one frame. Direct link BS BS-Subframe (TBS) BS–MSorBS–RS RS-Subframe (TRS) RS–MS MS Relayed link RS TF/2 TF/2

19. Direct link Relayed link Multiple-Relay Case: MRPAMultiple-RS Parallel with Activation Algorithm • If the number I of relays is an even number, there are I/2 relay pairs by regrouping the diametrically opposed relays. • Frequency reuse because of minimized interference. • FTD-based resource allocation. RS6 RS1 RS5 BS RS2 RS4 RS3

20. Direct link Relayed link Multiple-Relay Case: MRPAMultiple-RS Parallel with Activation Algorithm • If the number I of relays is an even number, there are I/2 relay pairs by regrouping the diametrically opposed relays. • Frequency reuse because of minimized interference. • FTD-based resource allocation. RS6 RS1 RS5 BS RS2 RS4 RS3

21. Multiple-Relay Case: MRAAMultiple-RS Adaptive Activation Algorithm • Without assuming frequency reuse. • RSj with the worst throughput is removed, and the RSj-subframe is reallocated to the BS-subframe. • For higher throughput performance • ATD-like resource allocation RS6 BS RS2 RS4 RS3

22. Performance–實驗參數 • BS Cell : 1000m radius • Relay placed in 800m away from BS • BS/RS Power: 20/5 W • Subchannels : 12 • Frame duration : 12 ms • Packet arrive at BS : Poisson process

23. All Fwd: Relays selected in random w/o UReq PFS: Proportional Fair Scheduling w/o considering R and w/o Relays Performance – Single-Relay Case Upper Bound Upper / Lower Bound Packets from BS through RS to MS is in the same frame

24. All Fwd: Relays selected in random w/o UReq PFS: Proportional Fair Scheduling w/o considering R and w/o Relays Performance – Single-Relay Case Upper Bound (throughput) Lower Bound

25. PFS: Proportional Fair Scheduling w/o considering R and w/o Relays Performance – Multiple-Relay Case Upper Bound

26. PFS: Proportional Fair Scheduling w/o considering R and w/o Relays Performance – Multiple-Relay Case Lower Bound

27. Conclusion • This paper investigated the problem of resource allocation for a relay-aided cellular system based on OFDMA. • Design FTD, ATD, MRPA and MRAA algorithms for good throughput and outage performance.

28. The End THANK YOU