VIEWS: 55 PAGES: 133 CATEGORY: US History POSTED ON: 6/4/2010 Public Domain
March 2006 doc.: IEEE 802.22-06/0050r0 Huawei Proposal Clarifications IEEE P802.22 Wireless RANs Date: 2006-03-20 Authors: Name Company Address Phone email Linjun Lu Huawei Technologies Shenzhen, China 0086-755-28973119 lvlinjun@huawei.com Soo-Young Chang Huawei Technologies Davis, CA, U.S. 1-916 278 6568 sychang@ecs.csus.edu Jianwei Zhang Huawei Technologies Shanghai, China 86-21-68644808 zhangjianwei@huawei.com Lai Qian Huawei Technologies Shenzhen, China 86-755-28973118 qlai@huawei.com Jianhuan Wen Huawei Technologies Shenzhen, China 86-755-28973121 wenjh@huawei.com Vincent K. N. Lau HKUST Hong Kong, China 852-2358-7066 eeknlau@ee.ust.hk Roger S. Cheng HKUST Hong Kong, China 852-2358-7072 eecheng@ee.ust.hk Ross D. Murch HKUST Hong Kong, China 852-2358-7044 eermurch@ee.ust.hk Wai Ho Mow HKUST Hong Kong, China 852-2358-7070 eewhmow@ee.ust.hk Khaled Ben Letaief HKUST Hong Kong, China 852-2358-7064 eekhaled@ee.ust.hk Notice: This document has been prepared to assist IEEE 802.22. It is offered as a basis for discussion and is not binding on the contributing individual(s) or organization(s). The material in this document is subject to change in form and content after further study. The contributor(s) reserve(s) the right to add, amend or withdraw material contained herein. Release: The contributor grants a free, irrevocable license to the IEEE to incorporate material contained in this contribution, and any modifications thereof, in the creation of an IEEE Standards publication; to copyright in the IEEE’s name any IEEE Standards publication even though it may include portions of this contribution; and at the IEEE’s sole discretion to permit others to reproduce in whole or in part the resulting IEEE Standards publication. The contributor also acknowledges and accepts that this contribution may be made public by IEEE 802.22. Patent Policy and Procedures: The contributor is familiar with the IEEE 802 Patent Policy and Procedures http://standards.ieee.org/guides/bylaws/sb-bylaws.pdf including the statement "IEEE standards may include the known use of patent(s), including patent applications, provided the IEEE receives assurance from the patent holder or applicant with respect to patents essential for compliance with both mandatory and optional portions of the standard." Early disclosure to the Working Group of patent information that might be relevant to the standard is essential to reduce the possibility for delays in the development process and increase the likelihood that the draft publication will be approved for publication. Please notify the Chair Carl R. Stevenson as early as possible, in written or electronic form, if patented technology (or technology under patent application) might be incorporated into a draft standard being developed within the IEEE 802.22 Working Group. If you have questions, contact the IEEE Patent Committee Administrator at patcom@iee.org. Submission Slide 1 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Co-Authors: Name Company Address Phone email Edward K. S. Au HKUST Hong Kong, China 852-2358-7086 eeedward@ee.ust.hk Peter W. C. Chan HKUST Hong Kong, China 852-2358-7086 peter@ee.ust.hk Ernest S. Lo HKUST Hong Kong, China 852-2358-7086 eeern@ee.ust.hk Lingfan Weng HKUST Hong Kong, China 852-2358-7086 lingfan@ee.ust.hk Zhou Wu Huawei Technologies Shenzhen, China 86-755-28979499 wuzhou@huawei.com Jun Rong Huawei Technologies Shenzhen, China 86-755-28979499 rongjun@huawei.com Jian Jiao Huawei Technologies Beijing, China 86-10-82882751 jiao_jian@huawei.com Meiwei Jie Huawei Technologies Shenzhen, China 86-755-28972660 jiemingwei@hauwei.com Submission Slide 2 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Part 1: Channel Sensing Submission Slide 3 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Part 1: Channel Sensing Guard intervals for extra quiet period in TDD WRAN system Submission Slide 4 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Background • Synchronous Quiet Period a period in which all WRAN devices stop transmission in all channels available in the system used for sensing the signals in all channels of the system without interfering the system itself useful to enhance awareness to the surrounding radio environment • Can the sensing accuracy be further enhanced? Submission Slide 5 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Fully Utilize the Guard Intervals • Guard Intervals – When using OFDMA at the physical layer, guard intervals should be inserted at the switching points of transmission OFDM symbols of different users can be synchronized at BS. CPE1 (d=0) 1 2 3 4 1 2 3 4 1 2 3 4 … Downlink sub-frame GI Uplink sub-frame Downlink sub-frame CPE2 (d=R) 1 2 3 4 1 2 3 4 1 2 3 4 … [R: cell radius] Downlink sub-frame Uplink sub-frame GI Downlink sub-frame – We can use these guard intervals as extra quiet periods for sensing! Submission Slide 6 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Related Work by I2R (Singapore) • All CPE should have a mandatory quiet period with fixed length at the switching point from downlink (DL) to uplink (UL). TTG2 TTG1 BS DL Subframe Sense UL 1 UL 2 TRS Tss CPE1 DL Subframe Sense UL 1 DL1 DS1 Tss SSRTG DL1 CPE2 DL Subframe Sense UL 2 DL2 DS2 Tss SSRTG DL2 Submission Slide 7 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Related Work by I2R (Singapore) • Disadvantages – Guard intervals from uplink to downlink have not been utilized. – Since a quiet period of fixed length is inserted to all CPEs (regardless of their distances to base station), for the CPEs at the edge of the cell in which guard intervals are usually not required, the uplink transmission of these CPEs will be deferred not effective CPE1 (d=0) 1 2 3 4 1 2 3 4 1 2 3 4 … Downlink sub-frame GI Uplink sub-frame Downlink sub-frame CPE2 (d=R) 1 2 3 4 1 2 3 4 1 2 3 4 … [R: cell radius] Downlink sub-frame Uplink sub-frame GI Downlink sub-frame Submission Slide 8 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Our Proposed Design • Assumptions – TDD (time division duplex) deployment – OFDMA (orthogonal frequency domain multiplexing access) is used in both uplink and downlink • Main Features – Adaptive Guard Interval Control – Asynchronous Quiet Period Submission Slide 9 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Our Proposed Design • Feature: Adaptive Guard Interval Control Conventionally, CPE1 should wait for CPE2 during the uplink transmission such that their first uplink symbols are synchronized at BS. We relax the above constraint: CPE1 (d=0) 1 2 3 4 1 2 3 4 1 2 3 4 … Downlink sub-frame GI Uplink sub-frame Downlink sub-frame CPE2 (d=R) 1 2 3 4 1 2 3 4 1 2 3 4 … Downlink sub-frame Uplink sub-frame GI Downlink sub-frame * CPE2’s first UL symbol is synchronized with CPE1’s second UL symbol Submission Slide 10 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Advantages of Adaptive GI Control For those CPEs being close to BS: they can start transmission in advance (1) Length of guard intervals from DL to UL can be shortened (2) More OFDM symbols can be transmitted For those CPEs being far away from BS (1) Uplink transmission will no longer be deferred (2) Number of transmitted OFDM symbols remains unchanged If considering some practical limitations such as the hardware limitation or the delay spread of the multi-path channel, a gap should be guaranteed between the DL and UL sub-frame when operating the adaptive GI control. 1 2 3 4 1 2 3 4 1 2 3 4 … Downlink sub-frame Uplink sub-frame Downlink sub-frame CPE3 (0<d<R) Without adaptive guard interval control 1 2 3 4 1 2 3 4 5 1 2 3 4 … Downlink sub-frame Uplink sub-frame Downlink sub-frame With adaptive guard interval control Submission Slide 11 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Our Proposed Design • Feature: Asynchronous Quiet Period Guard intervals from UL to DL can also be used as extra quiet period for channel sensing. Depending on the demand for sensing accuracy, some OFDM symbols can be replaced by the sensing period – Flexibility is ensured – BS notifies the assignment of such sensing periods to the CPEs by using the proposed Sensing Period Assignment (SPA) message. 1 2 3 4 1 2 3 4 5 1 2 3 4 … CPE3 (0<d<R) Downlink sub-frame Uplink sub-frame Downlink sub-frame With adaptive guard interval control Submission Slide 12 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 SPA Message Format Syntax Notes SPA_Message_Format() { Management Message Type Indicates the type of SPA message Connection ID Indicates the user to whom the message is sent Start Time Indicates the start time of the sensing period, in unit of OFDM symbols Duration Indicates the duration of the sensing period, in unit of OFDM symbols } Submission Slide 13 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Conclusion • Adaptive Guard Interval Control For CPEs being close to BS, more OFDM symbols can be transmitted Guard intervals from DL to UL can be shortened For CPEs being far away from BS, their uplink transmission will no longer be deferred Performance Gain: Assume cell size is 33km and frame length is 5ms, the round-trip delay is 0.22ms 4.4% of bandwidth can be used! • Asynchronous Quiet Period Guard intervals from UL to DL can also be used for channel sensing Flexibility: some OFDM symbols can be replaced by sensing period Sensing Period Assignment (SPA) message: one kind of MAC management message Submission Slide 14 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Part 1: Channel Sensing Region-based Bayesian method for RF sensing in WRAN system Submission Slide 15 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Background • The WRAN system needs to detect the presence of incumbent systems and avoid the interference to the incumbent system • Detection of Incumbents – Presence of the incumbents – Locations of the incumbents Submission Slide 16 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Detection of Incumbents • Detect the presence of the incumbents – The subband needs to be vacated in the whole cell/sector – Lower spatial efficiency • Detect the locations of the incumbents – When the operation range of incumbent is small, the subband may be used without interfering to the incumbent. – Higher spatial efficiency – Complexity grows exponentially with the number of targets – Many previous work requires: knowledge of number of targets, knowledge of signatures, and detection of time of arrivals, etc. Submission Slide 17 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Region-based RF Sensing Algorithm • Partition the cell/sector into a number of disjoint regions • For each region, decide whether some incumbents exist Higher spatial efficiency The number of targets need not be known a priori Complexity does not exponentially grow with the number of targets • Control overhead Each CPE feedbacks incumbent types and the corresponding subband Submission Slide 18 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 The Bayesian Method (1) Define d = The vector {di} of all sensors. and Cij = Cost of deciding Hi(), given Hj() is true The decision rule of the Bayesian method is 1 i ( ) arg min Ci ( ) arg min P(d, H j ( ))Cij * i0,1 i0,1 j 0 where is a subset of PIT region iff i*() = 1. Cost matrix example: C00 C01 0 10 C 1 1 C10 C11 Submission Slide 19 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 The Bayesian Method (2) For simplicity, assume conditional independence, P(d, H 0 ( )) P( H 0 ( )) P(d i 0 | H 0 ( )) P(d i 1 | H 0 ( )) iS , 0 iS ,1 P(d, H1 ( )) P( H1 ( )) P(d i 0 | H1 ( )) P(d i 1 | H1 ( )) iS , 0 iS ,1 P(di 0 | H 0 ( )) P(di 0, H 0 ( Ai ) | H 0 ( )) P(di 0, H1 ( Ai ) | H 0 ( )) 1 PF ,i e ( Ai ) 1 PD,i ( Ai ) 1 e ( Ai ) P(di 1 | H 0 ( )) P(di 1, H 0 ( Ai ) | H 0 ( )) P(di 1, H1 ( Ai ) | H 0 ( )) PF ,i e ( Ai ) PD ,i ( Ai ) 1 e ( Ai ) P(d i 0 | H1 ( )) 1 PD ,i ( ) P(d i 1 | H1 ( )) PD ,i ( ) Submission Slide 20 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 The Bayesian Method (3) We assume the detection process of a sensor is modeled by Bernoulli trials. Each IT within its detection region is an i.i.d. trial. The probability of detecting a particular IT is independent of its position. Flag di = 1, iff at least one IT is detected. Submission Slide 21 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 The Bayesian Method (4) PD1i) P(di 1 | Exact one incumbent transmitte r is within Ai ) ( , PD,i ( ) P(di 1 | H1 ( )), Ai P(d i 1, H1 ( )) P(Some IT in is detected) P(Some IT exists in but none is detected) P(Some IT is detected in Ai ) 1 e ( ) PD1,i) ( e ( ) PD1,i) ( e ( ) 1 e ( Ai ) PD1,i) ( ( Ai ) PD1,i) ( ( Ai ) PD1,i) ( 1 e ( ) e ( ) e e ( Ai ) PD1,i) ( ( ) ( Ai ) PD ,i (1) e e e PD ,i ( ) 1 1 e ( ) Submission Slide 22 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Algorithm Flow System Initialization Compute PIT (potential Compute incumbent transmitter) region Protection Region Submission Slide 23 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 PIT Region Distribution • 400 CPEs, 4 ITs (incumbent transmitters) • Detection radius = 10 grids (Grid space = 50m) • 5km by 5km square region CPE Number = 400, IT Number = 4 CPE Number = 400, IT Number = 4 100 100 90 90 80 80 70 70 y (grid point) y (grid point) 60 60 Grid Index Y Grid Index Y 50 50 40 40 30 30 20 20 10 10 0 0 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Grid Index X Grid Index X x (grid point) x (grid point) Union Algorithm Region-based Algorithm Submission Slide 24 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Region-based vs. Union Algorithms • PF,i = 0.01 (per CPE per subband) 0.4 0.7 Union Alg Union Alg Ratio of areas of PIT region 0.35 Region Based Alg 0.6 Region Based Alg Probabilitymiss miss 0.3 0.5 Ratio of PIT Region of 0.25 0.4 Probability of 0.2 0.3 0.15 0.2 0.1 0.1 0.05 0 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 2 2 CPE density (#CPE/km ) CPE density (#CPE/km ) CPE density (#CPE/km2) CPE density (#CPE/km2) Submission Slide 25 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Region-based vs. Union Algorithms • PF,i = 0.1 (per CPE per subband) 0.8 0.7 Union Alg Union Alg Region Based Alg Region Based Alg Ratio Ratio of PIT Region region 0.7 0.6 0.6 Probability missmiss 0.5 of areas of PIT 0.5 Probability of of 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 CPE density (#CPE/km 2) CPE density (#CPE/km 2) CPE density (#CPE/km2) CPE density (#CPE/km2) Submission Slide 26 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Sensitivity to Estimate of CPE density (1) • PF,i = 0.1 (per CPE per subband) • Actual average number of IT per km2 = 0.16 IT/km2 Ratio of PIT region -- lambda (Actual lambda = 0.16) Probability of miss -- lambda (Actual lambda = 0.16) 0.7 0.35 Union Algorithm Union Algorithm Ratio of areasRegion region Region Based Alg Region Based Alg 0.6 0.3 Probabilitymiss miss Ratio of PIT of PIT 0.5 0.25 Probability of of 0.4 0.2 0.3 0.15 0.2 0.1 0.1 0.05 0 0 -4 -3 -2 -1 0 1 2 -4 -3 -2 -1 0 1 2 10 10 10 10 10 10 10 10 10 10 10 10 10 10 2 lambda used in the algorithm (#estimated IT/km 2) lambda used Expected in the algorithm (#estimated IT/km number of IT per km2 ) Expected number of IT per km2 Submission Slide 27 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Sensitivity to Estimate of CPE density (2) 20 Ln( P(H0()) / P(H1())) Ln( P(di=0|H0()) / P(di=0|H1())) 10 Black curve 0 = Sum of three curves P(di 0 | H 0 ) i P(di 1 | H 0 ) P( H ) iS S ,1 ln 0 ,0 P( H1 ) P(d i 0 | H1 ) P(d i 1 | H1 ) -10 Decision Curve: <0 corresponds to PIT region iS , 0 iS ,1 Ln( P(di=1|H0()) / P(di=1|H1())) -20 -30 -40 -4 -3 -2 -1 0 1 2 3 10 10 10 10 10 10 10 10 Estimated Estimate of Submission Slide 28 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Region-based Algorithm: Transceivable Region • Downlink System – Ideal antenna with 120-degree beam-width and front-to-back ratio GFB of 13dB. – Uniform gain within main beam and constant attenuation of 13dB outside. – Cell radius is 33km; path loss exponent in a cell, pl = 3. – 10 circular clusters of CPEs, with radius of 3km, center uniformly distributed – For every cluster, 100 CPEs are uniformly distributed within it. – Pth = Maximum WRAN signal power allowed in the protection region – PRmin = Minimum required receiving power of a CPE: Pth + 3dB – Drr = The radius of the receivable region – Dpro = The minimum distance between BS and protection region, Dpro. 1 Pth pl P D pro Drr R min Submission Slide 29 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Region-based Algorithm: Transceivable Region Region based Algorithm 40 Sector of a WRAN cell 35 A cluster 30 Cyan point: CPE that reports d=0 Red point: CPE that reports d=1 25 20 y (km) km 15 Yellow region: PIT region 10 Blue star: IT 5 0 CPE Transceivable Region -5 -20 -10 0 10 20 30 40 km x (km) Submission Slide 30 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Comparison of Usable Subbands • Average number of subbands available to the system: FDD – Region-based Algorithm: 32 subbands / base station – Union Algorithm: 12 subbands / base station TDD – Region-based Algorithm: 31 subbands / base station – Union Algorithm: 12 subbands / base station Submission Slide 31 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Complexity Comparison 14 Complexity Ratio of Region-based to Union Algorithm Complexity of Region Based Alg over Union Alg 12 At a reasonable CPE density, the complexity of the region- 10 based algorithm is about 10 times of the union algorithm 8 and its complexity will converge to less than 14 times 6 of the union algorithm. 4 2 0 2 4 6 8 10 12 14 16 2 CPE density (#CPE/km2 CPE density (#CPE/km ) ) Submission Slide 32 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Conclusion • More efficient use of space for frequency reuse – Smaller PIT and protection regions larger transceivable region – CPEs inside the receivable region can use the channel – Noticeable gain in the number of usable channels per cell • Compared with union algorithm: – Large reduction in PIT region – Small increase in probability of miss – The tradeoff can be controlled by the cost matrix • Moderate Computation Complexity • Low overhead for sensing report Submission Slide 33 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Part 1: Channel Sensing The MAC management message for channel sensing Submission Slide 34 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Proposal • Design of MAC Management Messages for channel sensing of the CPE’s • Our proposed RF sensing algorithm suggests the following information is sufficient for satisfactory performance in sensing report of CPE’s – Incumbent type – Channel occupied by the incumbent • Design Criteria: Reduce control overhead Submission Slide 35 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Measurement Request (MS-REQ) Syntax Size Notes Management Message Type 8 bits The type of MAC management message Transaction ID 16 bits The transaction ID of the corresponding channel measurement request for which a report is required by the BS. System Type 8 bits The type of incumbent system to be measured. See Table 2. If this field is 0, the CPE should sense all incumbent systems. Start frame 8 bits Indicate the frame from which the channel measurement starts. Duration 8 bits The actual duration of the measurement (units in frame). Full_Report 1 bit 0: Does not request full report. 1: Request full report. CINR_Flag 1 bit 0: Does not request CPE to report CINR (Carrier Interference-to-Noise Ratio). 1: Request CPE to report CINR for each channel of each incumbent type. Table 1. Measurement Request (to be cont’d) Jianwei Zhang, Huawei Submission Slide 36 March 2006 doc.: IEEE 802.22-06/0050r0 Measurement Request (MS-REQ) Channel_List_Format 1 bit 0: Channel-basis 1: Interval-basis If (Channel_List_Format=0){ Channel-basis Channel List Variable See Table 3. } else{ Interval-basis Channel List Variable See Table 4. } Table 1. (Cont’d) Measurement Request Submission Slide 37 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 System Types System Type Description 0 All types 1 802.22 2 ATSC 3 NTSC 4 Part 74 5 DVB 6-255 Reserved Table 2. System Types Submission Slide 38 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Channel-/Interval- basis Channel Lists Channel-basis Channel List Size Notes Number of Channels, c 8 bits Total number of channels the CPE should measure. for i = 1:c { Channel Index 8 bits } Table 3. Channel-basis Channel List Interval-basis Channel List Size Notes Number of Intervals, d 8 bits Total number of channel intervals the CPE should measure. for i = 1:d { Starting Channel Index 8 bits The index of the starting channel Number of Channels 8 bits The number of channels in the current interval } Table 4. Interval-basis Channel List Submission Slide 39 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Measurement Report (MS-REP) Syntax Size Notes Management Message Type 8 bits Transaction ID 16 bits Report_Detail 1 bit 0: Incremental Measurement Report 1: Full Measurement Report If(Report_Detail=0){ Incremental Measurement Variable See Table 6. Report } else{ Full Measurement Report Variable See Table 7. } Table 5. Measurement Report Submission Slide 40 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Incremental Measurement Report Syntax Size Notes Number of System Types, n0 8 bits Number of system types that has been detected. for i = 1:n0{ System Types 8 bits See Table 2. Number of Channels, mi 8 bits Number of channels occupied by current incumbent system. for j = 1:mi{ Start frame 8 bits Indicate the frame from which the channel measurement starts. Duration 8 bits The actual duration of the measurement (units in frame). Channel Index 8 bits Leave or Arrive 1 bit 1: the channel is occupied by the incumbent system. 0: the channel is released from the incumbent system. if(CINR_Flag=1){ CINR 8 bits } } } Table 6. Incremental Measurement Report Submission Slide 41 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Full Measurement Report Syntax Size Notes Number of System Types, n1 8 bits Number of system types that has been detected. for i = 1:n1{ System Types 8 bits See Table 2. Number of Channels, mi 8 bits Number of channels occupied by current incumbent system. for j = 1:mi{ Start Frame 8 bits Indicate the frame from which the channel measurement starts. Duration 8 bits The actual duration of the measurement (units in frame). Channel Index 8 bits if(CINR_Flag=1){ CINR 8 bits } } } Table 7. Full Measurement Report Submission Slide 42 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Example Submission Slide 43 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Conclusion • Compatible with IEEE802.16 MAC Management Messages – Flexible – Support various schemes of control channel assignment • Design Criteria: Reduce control overhead – Interval-basis Channel List – Incremental Measurement Report Submission Slide 44 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Part 1: Channel Sensing Pilot design for channel estimation and interference detection in WRAN system Submission Slide 45 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Proposed Downlink Pilot Design ● ● ● ● ● ● ● ● ● … Block 0 ● ● ● ● ● ● ● ● ● … Block 1 ● ● ● ● ● ● ● ● ● … Block 2 ● ● ● ● ● ● ● ● ● … Block 3 … Block 4 … Block 5 … Block 6 … Block 7 … Block 8 ………………………………………. ●: Pilot subcarrier : Data subcarrier Major consideration The channel is slow varying The subcarrier spacing is about several KHz To facilitate the interference detection Submission Slide 46 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Proposed Uplink Pilot Design ●●●● Block 0 Block 1 Block 2 ……………… ●: Pilot subcarrier : Data subcarrier Major consideration The channel is slow varying The subcarrier spacing is about several KHz Subband-based OFDMA Submission Slide 47 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Interference Detection Channel Channel + Interference • Left graphs stands for the constellation of pilots on the same subcarriers of different OFDM blocks • Right graphs stands for the constellation of corresponding received signals • Interference symmetric structure of the constellation will be destroyed • No matter the interference varies or not • No matter what constellation size used Jianwei Zhang, Huawei Submission Slide 48 March 2006 doc.: IEEE 802.22-06/0050r0 Interference Detection Algorithm • System model: Y = P*H + n one subcarrier • Pk,i: Pilot on the k-th subcarrier of i-th OFDM block. • Pk,i = - Pk,i+1 • Hypothesis test: H0: |Yk,i + Yk,i+1|2 = |Pk,i*Hk + nk,i + Pk,i+1*Hk + nk,i+1|2 = |nk,i + nk,i+1|2 H1: |Yk,i + Yk,i+1|2 = |Pk,i*Hk + Ik,i + nk,i + Pk,i+1*Hk + Ik,i+1 + nk,i+1|2 = |Ik,i + Ik,i+1 + nk,i + nk,i+1|2 P(|Yk,i+Yk,i+1|2 > threshold | H0) = Palarm • |Yk,i + Yk,i+1|2 given H0 χ2 distribution. Submission Slide 49 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Simulation Model and Parameters Noise signal + Remove CP FFT • Interference generated in Interference time domain more close to the real situation AWGN Filter • Interference on one 0 subcarrier of different -10 -20 OFDM blocks varies Frequency response Power profile of filter, 20 log |H| • False alarm probability is 10 -30 of the filter. -40 -50 set to 0.01 -60 -70 • Noise power is known a -80 -90 prior -100 100 200 300 400 500 600 700 800 900 1000 k th subcarrier Submission Slide 50 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Simulation Results --- 2 Pilots 1 1 0.9 0.9 0.8 0.8 Detection Probability Detection Probability 0.7 0.7 INR = 0dB INR = 10dB 0.6 0.6 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 0 100 200 300 400 500 600 700 800 900 1000 0 100 200 300 400 500 600 700 800 900 1000 Subcarrier Index Subcarrier Index 1 0.9 0.8 Detection Probability 0.7 INR = 20dB 0.6 0.5 0.4 0.3 0.2 0.1 0 0 100 200 300 400 500 600 700 800 900 1000 Subcarrier Index Jianwei Zhang, Huawei Submission Slide 51 March 2006 doc.: IEEE 802.22-06/0050r0 Simulation Results --- 5 Pilots 1 1 0.9 0.9 0.8 0.8 Detection Probability Detection Probability 0.7 0.7 INR = 0dB INR = 10dB 0.6 0.6 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 0 100 200 300 400 500 600 700 800 900 1000 0 100 200 300 400 500 600 700 800 900 1000 Subcarrier Index Subcarrier Index 1 • 0.9 0.8 Threshold does not change Detection Probability • 0.7 0.6 INR = 20dB But use N 1 Y 0.5 2 0.4 k ,i Yk ,i 1 0.3 i 1 0.2 N 1 0.1 0 0 100 200 300 400 500 600 700 800 900 1000 to do the hypothesis test Subcarrier Index Submission Slide 52 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Conclusion • Pilot design for both downlink and uplink • Interference detection – Do not require extra overhead – No matter the interference is varying or not – No matter the constellation size used – Performance only depends on interference to noise ratio Submission Slide 53 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Joint Interference Detection & Decoding • Existence of Narrowband Interference in WRAN • Avoids Transmission in Interference Jammed Subcarriers – Transmitter may not know the existence of interference due to bursty nature of interference • Receiver Detect Interference – Pilot based approaches – Data based approaches • Based on estimated data • Based on correlation of channel fading in frequency and time domain • Existing Decoders Require Interference Knowledge – Performance determined by the accuracy of the interference detection Submission Slide 54 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 System Model Parallel Channel Par al l el Channel Enc IL Mod Dem DI L Dec For Each Codeword Enc IL Mod Dem DIL Dec Fading AWGN Interference Submission Slide 55 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Existing Decoding Schemes • Optimal Maximum Likelihood Decoding – Decoding metric is optimized differently for noise and interference – Require noise and interference statistics (position and power) • Conventional Decoding – Only require interference position; not require interference power – Ignore (erase) interference jammed symbols – Decoding metric is Euclidean distance (Optimal metric for AWGN) – Undetected interference corrupts decoder because of metric mismatch all require interference detector Submission Slide 56 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Joint Erasure Decoding • Given the number of erasures, search all possible codewords x with all possible erasure positions e min min x y e x 2 e • Determine the number of erasures – Apply sufficiency criteria • Achievable performance – Maximum Likelihood decoding with the exact knowledge of the noise and interference statistics Submission Slide 57 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Implementation – Product Trellis • Erasure Indicator Trellis ε1 ε2 ei 1i 0 ei 0 1 ei 1 • Bit Trellis a a c d e f b b • Product Trellis [a,0] [a,0] [a,1] Submission Slide 58 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Sufficiency Criteria • Error Checking Code Based – Output the first candidate codeword that passes error checking and terminate decoding • Path Metric Difference Based – Calculate path metric difference of consecutive candidate codewords • Metric difference is decreasing • Metric difference is small after all interference are erased – If the metric difference is less than a threshold Dec , then output the candidate codeword & terminate decoding Submission Slide 59 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Complexity Reduction • Find the most likely path sequentially • Demodulator marks symbol erasures – Erase the symbol if any of the corresponding bit is marked as an erasure by decoder – Erase the symbol based on the channel output Undetectable left to decoder detectable Submission Slide 60 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Simulations – Fixed SIR or Jams • Rate-½ 64-state convolutional code • 16QAM with Gray mapping • 864 subcarriers • Profile A multipath fading channel; constant over each packet • Fixed SIR or number of jammed subcarriers • Sufficiency criterion: path metric difference based • Demodulator does not mark erasure based on channel output Submission Slide 61 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Simulations – Fixed SIR or Jams SIR=0dB, SNR=20dB 5 Jams, SNR=20dB -2 -1 10 10 Conventional Conventional Proposed Proposed Maximum Likelihood Maximum Likelihood -2 10 -3 BER BER 10 -3 10 -4 10 -4 0 2 4 6 8 10 12 10 -10 -5 0 5 Number of Jammed Subchannels SIR(dB) The proposed decoder (1) almost achieves the performance of the optimal decoder (2) reduces sensitivity to the number of jammed subcarriers (3) is insensitive to interference power Submission Slide 62 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Simulations – Fixed SIR or Jams SIR=0dB, SNR=20dB 5 Jams, SNR=20dB -4 -4 x 10 x 10 7.4 11 SIR=-10dB 11 SIR=-5dB 7.2 SIR=0dB 10 SIR=5dB 9 7 9 7 6.8 8 BER BER 7 6.6 5 6 6.4 5 3 6.2 4 1 3 6 2 0 10 20 30 40 50 60 5.8 2 0 10 20 30 40 50 60 Dec / 2 Dec / Optimal threshold of the path metric difference based sufficiency criterion is (1) almost independent of number of jammed subcarriers (2) almost independent of interference power Threshold can be determined offline Submission Slide 63 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Simulation – W/O Interference Detector • 864 subcarriers • 20 OFDM symbols per packet • One codeword per OFDM symbol • 2 OFDM pilot symbols for – Interference detection – Frequency domain LS channel estimation • 32 jammed subcarriers (wireless microphone) • SIR uniformly distributed in [-20dB,10dB] • Sufficiency criterion: path metric difference based Submission Slide 64 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 0 Simulation – W/O Interference Detector 10 0 10 -1 10 Packet Error Rate -2 BER 10 -1 10 -3 10 Conventional Conventional Proposed Proposed Conventional (ID) Conventional (ID) -4 Proposed (ID) Proposed (ID) 10 16 17 18 19 20 21 22 16 17 18 19 20 21 22 SNR(dB) SNR(dB) (1) Without interference detector (red) Great gain over conventional decoder for BER and PER Complexity increase by 1.5 times for PER=0.1 relative to conventional (2) With interference detector (blue) Smaller gain for BER but significant gain for PER Complexity increase by 15% for PER=0.1 relative to conventional with or without interference Jianwei Zhang, Huawei (3) Proposed decoder performs similarly 65 Submission Slide detector March 2006 doc.: IEEE 802.22-06/0050r0 Simulation – W/O Channel Estimation Error • Random interference for each carrier with probability 0.04 • SIR uniformly distributed in [-20dB,10dB] • 2 OFDM pilot symbols for frequency domain LS channel estimation • Each codeword is transmitted through 200 carriers and 10 OFDM symbols • Each convolutional codeword is encoded by CRC • Demodulator marks erasures • Sufficiency criterion: CRC and path metric difference based – CRC generator polynomial is 435(octal ) Submission Slide 66 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 0 Simulation – W/O Channel Estimation Error 10 0 10 -1 10 -2 10 Word Error Rate -1 10 -3 10 BER Gain of joint over separate -4 10 -2 Conventional 10 Conventional 10 -5 Proposed1 (separate) Proposed1 (separate) Proposed2 (joint) Proposed2 (joint) Maximum Likelihood Maximum Likelihood -6 Conventional (CE error) Conventional (CE error) 10 Proposed1 (CE error) Proposed1 (CE error) Proposed2 (CE error) -3 Proposed2 (CE error) 10 10 11 12 13 14 15 16 17 18 10 11 12 13 14 15 16 17 18 SNR(dB) SNR(dB) (1) Without channel estimation error (solid) Joint erasure marking and decoding Performs closely to optimal decoder Complexity increase by 50% for WER=0.01 relative to conventional decoder (2) With channel estimation error (dashed) Joint erasure marking and decoding is less sensitive to channel estimation error than separate erasure marking and decoding using demodulator only Complexity increases by twice for WER=0.01 relative to conventional decoder Submission Slide 67 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Conclusion • The proposed decoding scheme almost achieves the optimal decoder performance without knowing the interference statistics • Threshold of sufficiency criterion does not depend on interference characteristics and can be determined offline • Complexity increase is reasonably small especially for high SNR or with an interference detector • Performance loss due to channel estimation error is much smaller than that of conventional decoding scheme • Therefore, it is robust and effective to combat unknown interference in practical situations Submission Slide 68 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Part 2: Radio Resource Allocation Effective and flexible structure for CPE CSIT collection at base station for TDD/FDD OFDMA architecture Submission Slide 69 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Motivation Radio resource is very scarce Design good resource allocation algorithm to fully utilize the resource CSIT is a crucial input Submission Slide 70 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 CSIT Acquisition • Using the reciprocity of the uplink and downlink channel CSIT of the excited subchannels of those currently uplink-active CPEs of TDD system • Using feedback CSIT of the un-excited subchannels of those currently uplink-active CPEs of a TDD system CSIT of the currently uplink-inactive CPEs of TDD system CSIT of all the CPEs of FDD system Very important to design a good CSIT collection mechanism Submission Slide 71 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Features of Downlink WRAN System • BS knows the QoS requirements and queueing states of all the CPEs BS can determine which CPEs have higher priority and are more urgent • Maximum Doppler frequency is very small The CSIT can be updated rather infrequently • Variation of Doppler frequency among CPEs is limited The CSIT update frequencies of CPEs are similar Polling-based CSIT feedback mechanism Submission Slide 72 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Main Features of Our Proposed Structure • Centralized polling at the BS BS decides which CPEs to poll based on QoS requirements, queueing states, etc. BS decides for each selected CPE which subband to estimate based on power mask, history, etc. BS decides for each selected CPE through which subchannels to convey CSIT • Placement of the polling information For currently active CPEs, the polling information is contained in the UL-MAP For currently inactive CPEs, the polling information is contained in some broadcast channel Submission Slide 73 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 CSIT Collection Request Message (1) Syntax Size (bits) Remarks CSIT_Collection_Request() { N_DL_RCID is the number of selected downlink-active- N_DL_RCID 8 only CPEs and both-downlink-and-uplink-active CPEs that are in this subband for i = 1: N_DL_RCID { Downlink RCID 8 Feedback_Control( ) variable } 0: no selected CPE is uplink-active-only UL_RCID_flag 1 1: there are selected CPEs that are uplink-active-only If {UL_RCID_flag == 1}{ N_UL_RCID is the number of selected uplink-active-only N_UL_RCID 8 CPEs that are in this subband for i = 1: N_UL_RCID { Uplink RCID 8 Feedback_Control( ) variable } } CSIT_Collection_Request for active CPEs (to be cont’d) Submission Slide 74 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 CSIT Collection Request Message (2) 0: no CID is used CID_flag 1 1: CID is used If {CID_flag == 1}{ N_CID is the number of selected CPEs that are switched N_CID 8 to this subband for i = 1: N_CID { CID 16 Feedback_Control( ) variable } } } CSIT_Collection_Request for active CPEs (Cont’d) Submission Slide 75 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Feedback Control Message (1) Syntax Size (bits) Remarks Feedback_Control() { 0: estimate the downlink CSI of this subband Subband_change_flag 1 1: in the next frame estimate the downlink CSI of the subband specified by Subband Index If{Subband_ change_flag==1}{ Subband Index 8 At most 256 6MHz subband } Else{ 0: use default quantization level, L=a Quantization_level_flag 1 1: use specified quantization level If{ Quantization_level_flag ==1}{ Assume there are at most 4 additional quantization Quantization level, L=b 2 precision levels } 0: use default number of subchannels, N=c Feedback_ch_constraint_flag 1 1: use specified number of subchannels If{ Feedback_ch_constraint_flag==1}{ Number of subchannels, N=d 6 Assume 64 subchannels in a subband } Feedback_Control Message (to be cont’d) Jianwei Zhang, Huawei Submission Slide 76 March 2006 doc.: IEEE 802.22-06/0050r0 Feedback Control Message (2) 0: use default number of OFDM symbols, M=e Feedback_symb_constraint_flag 1 1: use specified number of OFDM symbols If{Feedback_symb_constraint_flag==1}{ Assume at most 4 OFDM symbols can be used to do Number of OFDM symbols, M=f 2 feedback } for j=1:N{ Subchannel Index 6 } } } Feedback_Control Message (cont’d) Submission Slide 77 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 CSIT Collection Request Message (1) Syntax Size (bits) Remarks CSIT_Collection_Request() { N_CID 8 N_CID is the number of selected inactive CPEs for i = 1:N_CID{ CID 16 Subband Index 8 At most 256 6MHz subband 0: use default quantization level, L=a Quantization_level_flag 1 1: use specified quantization level If{ Quantization_level_flag ==1}{ Assume there are at most 4 additional Quantization level, L=b 2 quantization precision levels } 0: use default number of subchannels, N=c Feedback_ch_constraint_flag 1 1: use specified number of sub-channels If{ Feedback_ch_constraint_flag==1}{ Number of subchannels, N=d 6 Assume 64 subchannels in a subband } CSIT_Collection_Request for inactive CPEs (to be cont’d) Submission Slide 78 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 CSIT Collection Request Message (2) 0: use default number of OFDM symbols, M=e Feedback_symb_constraint_flag 1 1: use specified number of OFDM symbols If{Feedback_symb_constraint_flag==1}{ Assume at most 4 OFDM symbols can be used to Number of OFDM symbols, M=f 2 do feedback } for j=1:N{ Subchannel Index 6 } } } CSIT_Collection_Request for inactive CPEs (Cont’d) Submission Slide 79 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Main Features of Our Proposed Structure • Overhead reduction For currently active CPEs, 8-bit RCID is used instead of the 16-bit CID to identify CPEs • Flexibility Default constraint on the number of subchannels and the number of OFDM symbols that a CPE should use to do feedback is known to both the BS and the CPEs BS has the option to allocate more or less subchannels and/or OFDM symbols for each CPE to do feedback, depend on the QoS requirement or the urgency of the downlink traffic Default CSIT quantization level is known to both BS and CPEs BS has the option to increase or decrease the quantization level to adjust the precision of the feedback Submission Slide 80 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Main Features of Our Proposed Structure • CPEs decide which subchannel CSIT to feedback based on the channel condition Using predefined modulation and coding scheme, given the number of subchannels, OFDM symbols that are used to convey CSIT, and the CSIT quantization level, each CPE knows it can feedback the CSIT of say c number of subchannels For FDD system, the CPE should choose c number of subchannels with the largest gains For TDD system, the CPE should choose c number of un-excited subchannels with the largest gains Size (bits) Remarks CSIT_Feedback_Format() { If Q-bit feedback is allowed, then c Q (6 x ) for i = 1:c { Subchannel Index 6 Subchannel Gain x } } Submission Slide 81 Jianwei Zhang, Huawei CSIT_Feedback_Format March 2006 doc.: IEEE 802.22-06/0050r0 Part 2: Radio Resource Allocation Downlink multiuser resource allocation algorithm for OFDMA-based QoS-enabled WRAN system Submission Slide 82 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Background • What’s challenging for 802.22? (1) Interference Avoidance to Incumbent Users (IU) – No cooperation possible between incumbent & WRAN systems Preventive measures should be chosen at the WRAN transmitter – Unknown BS-IU channels & incompatible system structure Isotropic transmission reduces the effective cell coverage Transmit-side interference pre-cancellation is impossible (2) Broad available spectrum for each cell: (~180MHz, 30 TV channels) – covered by multiple OFDM symbols instead of one – max. one subband per each CPE Simultaneous multi-band channel estimation is not possible Submission Slide 83 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Key Related Work Paper • [Wong99] C. Y. Wong, R. S. Cheng, K. B. Letaief, and R. Murch, “Multiuser OFDM with adaptive subcarrier, bit and power allocation,” IEEE Journal on Selected Areas of Communications, vol. 17, no. 10, pp. 1747-1758, Oct. 1999. US Patent • [Li05] X. Li, H. Liu, K. Li, and W. Zhang, “OFDMA with Adaptive Subcarrier-Cluster Configuration and Selective Loading,” US Patent, US6947748 B2, Sep-20 2005. US Patent Application • [Cho05] Y.-O. Cho, et al, “Method for Allocating Subchannels in an OFDMA Mobile Communication System,” US Patent Application, US2005/0180354 A1, Aug-18, 2005. Submission Slide 84 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Definitions Concept of band, subband, subchannel and subcarriers Submission Slide 85 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Proposed Algorithm • Features for (1) Interference Avoidance to Incumbent Users (IU) – Peak power constraint, namely power mask, for every subband. – Sectored antenna adopted for reducing the performance sensitivity to any nearby incumbent users (from a cell to only a sector). for (2) Broad available spectrum for each cell: (~180MHz, 30 TV channels) – Two-layer resource allocation algorithm: Layer-1: subband allocation – distribute users over subbands exploiting knowledge of power mask. avoid over-congestion of subbands. Layer-2: in-subband subchannel and power allocation – maximize subband throughput: QoS-enabled, priorities allowed. Submission Slide 86 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 2-Layer Algorithm Dynamic Frequency Selection Block Layer-1 Allocation Knowledge of transmit Subband Assignment power mask on every subband in every sector Layer-2 Allocation In-subband Knowledge of Subchannel, Power channel gain of the and Rate Allocation assigned subband The two-layer structure of the multiuser resource allocation algorithm Submission Slide 87 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer-1: Subband Assignment Intuition: Subband with smaller allowed maximum transmit power should handle less CPEs Step 1: For each sector, eliminate those unserviceable subbands, defined as those subbands with the power mask value smaller than a threshold. Step 2: Define Pmm,b,c as the average power mask per subchannel of subband b, i.e. the peak possible transmit power per subchannel, in sector c. Let Kc be the total number of users in sector c. For each sector c, the number of users allocated to subband b, represented by Kb,c, is done according to the following equation: K b ,c K c f subband Pm m,b ,c | Pm m,1,c ,, Pm m, Nb ,c f sectorPm m,b ,c | Pm m,b ,c ,, Pm m,b , L P | Pm m,1,c ,, Pm m, Nb ,c f sectorPm m,b ,c | Pm m,b ,c ,, Pm m,b , L Nb f b 1 subband m m,b ,c Submission Slide 88 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer-1: Subband Assignment Step 2: (cont’d) where Nb is the number of serviceable subbands and L is the number of sectors. Both f sectorPmm,b,c | Pmm,b,c ,, Pmm,b, L and f subbandPmm,b,c | Pmm,1,c ,, Pmm, N ,c b should be non-decreasing functions of Pm m,b,c . Example functions: If the objective is to maximize the minimum average user data rate, we can use: (i) f subband Pmm,b,c | Pmm,1,c ,, Pmm, Nb ,c log 1 b Pmm,b,c where b can be set to the average channel power gain to noise ratio. f sectorPm m,b ,c | Pm m,b ,c ,, Pm m,b , L Pm m,b ,c / Pm m,b ,c L (ii) c 1 - (i) approximates the rate of each subchannel in subband b of sector c. - (ii) reflects the relative number of possible subchannel allocation across different sectors for that subband b. Submission Slide 89 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer-1: Subband Assignment Step 3: Randomly select Kb,c users for subband b in sector c. Remarks: Step 3 is indeed up to the vendors. e.g. Assignment can be done based on user classes so that users of higher class may be distributed to a subband with larger power mask. Example: Advantages of exploiting one-dimensional (within sector) and two-dimensional (across sector & subband) power mask against equal user allocation. Objective: maximize the minimum average user data rate. System Settings: 3 sectors, 2 subbands, 40 subchannels per subband, 60 users per sector. (i) 1-D (Single-sector) allocation f sectorPmm,b,c | Pmm,b,c ,, Pmm,b, L 1 (ii) 2-D (Multi-sector) allocation f sectorPm m,b ,c | Pm m,b ,c ,, Pm m,b , L Pm m,b ,c / Pm m,b ,c L c 1 Submission Slide 90 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer-1: Performance Subchannel power masks in the example for the Layer-1 algorithm: Subband 1 Subband 2 Subchannel Power Subchannel Power Mask Mask Sector 1 20 20 Sector 2 40 20 Sector 3 40 0 Submission Slide 91 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer-1: Performance Subchannel allocation and subchannel data rate for the Layer-1 algorithm example with 40 subchannels per subband: Equal Allocation 1-D (single-sector) 2-D (multi-sector) Allocation Allocation Subband 1 Subband 2 Subband 1 Subband 2 Subband 1 Subband 2 Sector #Sub- Data #Sub- Data #Sub- Data #Sub- Data #Sub- Data #Sub- Data ch. Rate ch. Rate ch. Rate ch. Rate ch. Rate ch. Rate Alloc- per Alloc- per Alloc- per Alloc- per Alloc- per Alloc- per ated subch. ated subch. ated subch. ated subch. ated subch. ated subch. 1 8 4.3923 20 4.3923 8 4.3923 20 4.3923 8 4.3923 20 4.3923 2 16 5.3576 20 4.3923 16 5.3576 20 4.3923 16 5.3576 20 4.3923 3 16 5.3576 0 0 16 5.3576 0 0 16 5.3576 0 0 Submission Slide 92 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer-1: Performance Effect of different user allocation algorithms on the subband data rate per user with 60 users per sector: (Differences are highlighted) Equal Allocation 1-D (single-sector) 2-D (multi-sector) Allocation Allocation Subband 1 Subband 2 Subband 1 Subband 2 Subband 1 Subband 2 Sector No. of Bits No. of Bits No. of Bits No. of Bits No. of Bits No. of Bits users per users per users per users per users per users per Alloc- User Alloc- User Alloc- User Alloc- User Alloc- User Alloc- User ated ated ated ated ated ated 1 30 1.1713 30 2.9282 30 1.1713 30 2.9282 17 2.0670 43 2.0429 2 30 2.8574 30 2.9282 33 2.5976 27 3.2536 30 2.8574 30 2.9282 3 30 2.8574 30 0 60 1.4287 0 -- 60 1.4287 0 -- Submission Slide 93 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer-1: Performance Advantage of 1-D allocation over Equal Allocation: - realized in Sector 3: min. average rate per user increases from 0 to 1.4287. Advantage of 2-D allocation over its 1-D counterpart (also Equal Allocation): - realized in Sector 1: min. average rate per user increases from 1.1713 to 2.0429. Submission Slide 94 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer-2: In-subband Allocation Objective: – Maximize subband throughput by subchannel (a group of pre-selected subcarriers) and power allocation. – Support differentiated-QoS service. – Allow flexible tradeoff between max. throughput and fairness among users. Problem Formulation: – divided into two cases: (i) individual subcarrier power gain is known. (ii) average channel power gain is known, – The proposed algorithm is optimal for case (i), and almost optimal for case (ii) if every subchannel is within the coherence bandwidth. Submission Slide 95 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer-2: In-subband Allocation Problem Formulation: Subchannel Sharing Factor Power allocated to user k on subchannel i K M max w k ,i {0 ,1} , k k ,i f k ,i ( Pk ,i ) P 0 k 1 i 1 k ,i lQoS _ Class ( k ) M K Ni e.g. wk k subject to P i 1 k 1 k ,i n ( i ) 1 k ,n ( i ) PTotal rate control (0 k 1) K k ,i 1 i {1,2,, M } priority control k 1 (lQoS_Class(k) 0) 0 Pk ,n (i ) Pkmnask PTotal , (i ) i ,2,, M , k ,2,, K , n(i) ,2,, N i 1 1 1 Power mask #subchannels #users #subcarriers in subchannel i Submission Slide 96 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer-2: In-subband Allocation Define the rate function f k ,i ( Pk ,i ) as: Ni f k ,i ( Pk ,i ) f n ( i ) 1 k ,n ( i ) ( Pk ,n ( i ) ) h 2 Pk ,n ( i ) Ni Ni log 2 1 Pk ,n (i ) Pk ,i k ,n ( i ) with n ( i ) 1 n 2 n ( i )1 where n 2 is the noise power, hk ,n ( i ) 2 is the average channel power gain of subcarrier n(i) in subchannel i, and 1.5 is a factor bridging the gap between ideal minimum ln( 5BER) power required (using mutual information) and actual required transmission power (using practical modulation schemes for a given rate) Submission Slide 97 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer-2: In-subband Allocation Proposed algorithm - by relaxing k ,i {0,1} to k ,i [ 0,1] , the problem becomes convex and method of Lagrangian can be applied to obtain the optimal solutions. Algorithm Details: Step 1: Initialization. Initialize Ω 0 . Step 2: Select the optimal CPE for each subcarrier for a given value of Ω CPE k is selected ( k ,i 1 ) for subcarrier i according to the following criterion: * 1 if Gk ,i (Ω ) maxGk ,i (Ω ) * k i 0 k ,i otherwise where ~ ~ ~ Ni Ω Ω Ni Ω Gk ,i (Ω ) wk f k ,n (i ) ( f ' 1n (i ) w ) w 1 f ' 1,n(i ) w n (i ) 1 k, k k k n (i ) k Submission Slide 98 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer-2: In-subband Allocation ~ with Ω defined as 1 Ω m ask wk f' k,n(i ) Pk,n(i ) if f' k,n(i ) w Pk,nask m (i ) k ~ 1 Ω Ω Ω if 0 f' k,n(i ) Pk,nask w m (i ) k 1 Ω wk f' k,n(i ) 0 if f' k,n(i ) 0 w k Step 3: Compute the optimal allocated power for each CPE for a given value of Ω. The optimal average power for user k on subchannel i is: Ω ~ * wk n 2 * c k .n ( i ) k .i f ' 1n ( i ) * k, w k .i ~ Ω 2 k hk ,n (i ) Submission Slide 99 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer-2: In-subband Allocation Step 4: Coarse adjustment of Ω. If ( Ω 0 ), K M Ni If ( c k 1 i 1 n ( i ) 1 k ,n (i ) PTotal ), Set Ω 0 for some small Ω . Go back to Step 2. Else Optimal solutions obtained; algorithm terminated. End Else K M Ni If ( c k 1 i 1 n ( i ) 1 k ,n (i ) PTotal ), Ωlower Ω; Ω 2Ω. Go to Step 2. Submission Slide 100 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer-2: In-subband Allocation Step 4 (Cont’d): K M Ni Elseif ( c k ,n (i ) PTotal ), k 1 i 1 n ( i ) 1 Go to Step 5. Else Optimal solutions obtained; algorithm terminated. End End Step 5: Fine adjustment of Ω. K M Ni While ( c k 1 i 1 n ( i ) 1 k ,n (i ) PTotal ) for some predefined tolerance level , Submission Slide 101 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer-2: In-subband Allocation Step 5 (Cont’d): Repeat Step 2 and 3. K M Ni If ( c k 1 i 1 n ( i ) 1 k ,n (i ) PTotal ), Ωlower Ω; K M Ni Elseif ( c k ,n (i ) PTotal ), k 1 i 1 n ( i ) 1 Ωupper Ω; End Ω Ωlower Ωupper / 2; Repeat Step 2 and Step 3. End Submission Slide 102 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer-2: In-subband Allocation In case of oscillations between two assignment profiles, K M K M Pu c k ,i PTotal and Pl c k ,i PTotal , k 1 i 1 k 1 i 1 a time-sharing ratio ( u : l ) for these profiles can be calculated: PTotal Pl u where l 1 u Pu Pl so that on average the total power constraint is satisfied. Submission Slide 103 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer-2: Channel Quantization In our numerical results, the following simple channel quantization algorithm is used: Quantization lookup table construction: 1. Acquire the channel power gain distribution. 2. Identify the range of the channel power with a desirable probability of occurrence, say 90%. 3. Equally partition the corresponding range in the logarithm domain. 4. Set up the thresholds as the middle points of each interval in the logarithm domain. 5. Transform the thresholds into their corresponding thresholds in the original domain. Submission Slide 104 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer-2: Rho Quantization When time-sharing cannot be implemented, the following two algorithms can be used: Algorithm 1: Step 1: Select the assignment profile closest to the Total Power Constraint. Step 2: Perform optimal power allocation for that assignment set. Algorithm 2: (shown good enough through numerical evaluation) Select the assignment profile with the total power smaller than the Total Power Constraint. In practice, perfect channel information feedback may not be possible but limited number of bits is used instead. Submission Slide 105 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer-2: Performance Sum rate comparison of (i) optimal SPA, (ii) random SA & optimal PA and (iii)random SA & equal PA with effects of channel quantization: Legend: - Perfect - 3-bit quantization - 1-bit quantization Submission Slide 106 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer-2: Performance Percentage loss of sum rate for the optimal subchannel and power allocation due to channel quantization: Submission Slide 107 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer-2: Performance Sum Rate Performance: Optimal Subchannel and Power Allocation: • 3-bit Channel Quantization is sufficiently good (~ 1% loss). • 1-bit Channel Quantization is fairly good (~ 9% loss). Random Subchannel Assignment with Optimal/Equal Power Allocation: • Even 1-bit Channel Quantization gives apparently the same performance. Submission Slide 108 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer-2: Performance Number of iterations required for convergence with 3-bit channel quantization and power constraint accuracy of 99.999998%: Submission Slide 109 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer-2: Performance Complexity Issue: Optimal Subchannel and Power Allocation (i) Number of operations1 required: (Number of users)*(Number of subcarriers or subchannels2) *(Number of iterations3) Random Subchannel Assignment with Optimal Power Allocation: (i) Number of operations1 required: (Number of subcarriers or subchannels2)*(Number of iterations3) Random Subchannel Assignment with Equal Power Allocation: • Two steps: random subchannel assignment + peak power clipping according to the Power Mask values. Remarks: 1. includes mainly the calculation of power and rate. 2. when the same channel gain and power mask are used in a subchannel. 3. fairly independent of the total number of users, of order O(log(FFT Size)) assuming same #subchannels for all FFT sizes. Submission Slide 110 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer-2: Performance Percentage of the occurrence of subchannel sharing with the application of 3-bit channel quantization. Submission Slide 111 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer-2: Performance Percentage loss of sum rate among the cases of subchannel sharing with sharing factor quantization for the optimal subchannel and power allocation: Submission Slide 112 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer-2: Performance - Sharing rarely occurs (~2%). - Actual loss due to rho-quantization in total data rate is negligible (~0.01% loss with rho-quantization Algorithm 2 among scenarios with time-sharing). Submission Slide 113 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Conclusion • Developed a two-layer resource allocation algorithm for the downlink IEEE 802.22 WRAN Systems, featuring – interference avoidance to incumbent users – user pre-distribution over subbands in a cell, avoiding over-congestion of subbands in a way that subband with a larger power mask (max. transmit power possible) should handle more CPEs – efficient in-subband subchannel and power allocation for: (i) maximizing subband throughput at affordable complexity, (ii) allowing QoS to be guaranteed, (iii) allowing prioritized transmission and flexible tradeoff between maximum throughput and fairness among users. Submission Slide 114 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Part 2: Radio Resource Allocation Joint dynamic frequency selection and power control with user specific transmit power mask constraints in uplink WRAN system using OFDMA scheme Submission Slide 115 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Background (1) • Principles of WARN systems shares the VHF/UHF TV bands between 47MHz-910MHz which are being used by the licensed operators and other license-exempt (LE) devices. a main constraint is to avoid interference to incumbent services such as TV broadcasting (analog and digital) and Public Safety systems. • Role of Dynamic Frequency Selection performs multiple-access control to provide QoS-guaranteed services required in the WRAN standard while not disturbing the service quality of the licensed users. involves user selection, rate adaptation as well as transmit power control (TPC). Submission Slide 116 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Background (2) Role of Dynamic Frequency Selection (Cont’) The spectrum occupation information, called geographical spectrum state information (GSSI), is obtained by data fusion and acts as the input information for dynamic frequency selection (DFS). – Usually full GSSI may not be easy to obtain. – Instead of full GSSI, one possible form of partial GSSI is transmit power masks imposed on all WRAN transmitters. CPE2 CPE1 × ×{n , n }={1, 1} {n , n }={1, 6} b c b c CPE3 × {nb, nc}={2, 3} {nb, nc}={5, 5} BS ×j ×CPE k CPE {nb, nc}={3, 7} {nb, nc}={5, 8} nb: Index of subbands nc: Index of subchannels × CPEn Submission Slide 117 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Related Works · In patent US2005180354 “Method for allocating subchannels in an OFDMA mobile communication system”, Cho et al. proposed resource allocation algorithms to maximum the transmission rates of all users by allocating subchannels and bits. · The scheme introduced an adaptive modulation using linear programming into an existing scheme for a system including a single kind of users, thereby enabling simultaneous execution of the adaptive modulation for all users in a system including two kinds of users. max z ck ,n , k ,n N s.t. Rk ck ,n k ,n z for all k n 1 K N k 1 n 1 f k (ck ,n ) k ,n / k2,n PT K Submission k 1 k ,n 1 for all n Slide 118 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Related Works · In paper “Multiuser OFDM with adaptive subcarrier, bit and power allocation,” Wong et al. considered a subcarrier, bit and power allocation problem in OFDM system. · The objective is to the minimize the total transmitted power, given the minimum data rate requirement of each user. K N k ,n min ck ,n , k ,n 2 f k (ck ,n ) k 1 n 1 k ,n N s.t. Rk ck ,n k ,n for all k n 1 K k 1 k ,n 1 for all n Submission Slide 119 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Drawbacks of the Related Work • For the patent US2005180354, the problem considered here is actually a rate adaptive problem which maximizes a lower bound of all users’ throughput with respect to a transmit power budget. • Delay constraints and users’ priorities were not considered in this invention. • It cannot be applied in WRAN systems since it does not employ any technique to guarantee free interference to the incumbent users. • Subband allocation among multiple OFDM symbols was not investigated. Submission Slide 120 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Our Proposed Algorithm • Two-Layers’ Design Dynamic Frequency Selection Block Layer-1 Allocation Knowledge of transmit Subband Assignment power mask on every subband in every sector Layer-2 Allocation In-subband Knowledge of Subchannel, Power channel gain of the and Rate Allocation assigned subband Submission Slide 121 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer 1 (Subband Allocation) • Method 1: (Sum-Rate-Max Strategy) Step 1: For each 6-MHz subband b, create a list of CPEs in descending order of their transmit power mask values. CPEs with power mask values smaller than a serviceable threshold predefined a priori are eliminated. Step 2: Create a list of CPEs in descending order of their maximum power ~ mask values across subbands. Define Pkmbask as the normalized power mask per , subchannel of user k on subband b. For k = Lmax (1) to Lmax ( K total )where k Lmax , (i) b k arg max f ~ m ask ~ m ask ~ m ask subch Pk ,b | P1,b , , PK total ,b f rate Pk ,b ~ m ask b1,, N b kL ~ , ~ b ~ ask ~ b f subch Pkmbask | P1,m ask ,, PKmtotal ,b f rate Pkmbask , (ii) Remove CPE k from Lb for all b’s except bk. End Jianwei Zhang, Huawei Submission Slide 122 March 2006 doc.: IEEE 802.22-06/0050r0 Layer 1 (Subband Allocation) (Cont’) The functions f subch Pkmask | P1,mask ,, PKmaskb and f rate Pkmask should be non-decreasing ~ ~ ~ ~ ,b b , total ,b functions. For example, ~ ,b ~ f rate Pkmask log 1 b Pkmask ,b b ~ m ask ~ m ask ~ m ask ~ m ask L ~ m ask f subch Pk ,b | P ,b , , PKtotal ,b Pk ,b / Pk ,b 1 k 1 where b can be set to the average channel gain to noise ratio. Step 3 (Optional): Perform subband re-assignment starting from the CPE with the minimum f subchPkmask | P,mask,, PKmaskb f rate Pkmask . ~ ,b ~ 1b ~ , total ~ ,b Submission Slide 123 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Layer 1 (Subband Allocation) • Method 2: (Round-Robin-Max Strategy) Step 1: For each 6-MHz subband b, create a list of CPEs Lb in descending order of the transmit power mask values. CPEs with power mask values smaller than a serviceable threshold predefined a priori are eliminated. Step 2: Sort the subbands in descending order of their maximum power mask. Starting from index 1, i.e. the subband with the largest maximum power mask, each subband takes turn to pick up one CPE with the maximum transmit power mask. Any CPE selected in the previous subband will be subtracted from the list of the latter subbands. Repeat Step 2 until the lists of all the subbands are empty. Submission Slide 124 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 An example for Layer-1 Algorithm (1) Subchannel Power Masks and the Approximated Subchannel Data Rate Subband 1 Subband 2 CPE Subchannel Power Approx. Subchannel Subchannel Power Approx. Subchannel Mask Data Rate Mask Data Rate 1 50 5.6725 0 0 2 50 5.6725 0 0 3 60 5.9307 50 5.6725 4 30 4.9542 20 4.3923 5 10 3.4594 30 4.9542 6 40 5.3576 20 4.3923 Submission Slide 125 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 An example for Layer-1 Algorithm (2) CPE Assignment and Corresponding Subchannel Data Rates Sum-Rate-Max Round-Robin-Max CPE-Max Subband 1 Subband 2 Subband 1 Subband 2 Subband 1 Subband 2 CPE Data CPE Data CPE Data CPE Data CPE Data CPE Data selected Rate per selected Rate per selected Rate per selected Rate per selected Rate per selected Rate per subch. subch. subch. subch. subch. subch. - - 3 5.6725 3 5.9307 - - 1 5.6725 - - 1 5.6725 - - - - 5 4.9542 2 5.6725 - - 2 5.6725 - - 1 5.6725 - - 3 5.9307 - - 6 5.3576 - - - - 4 4.3923 4 4.9542 - - - - 4 4.3923 2 5.6725 - - - - 5 4.9542 - - 5 4.9542 6 4.3923 6 5.3576 - - Submission Slide 126 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 An example for Layer-1 Algorithm (3) Subchannel Allocation and Corresponding Subchannel Data Rates Sum-Rate-Max Round-Robin-Max CPE-Max Subband 1 Subband 2 Subband 1 Subband 2 Subband 1 Subband 2 #Subch. Data #Subch. Data #Subch. Data #Subch. Data #Subch. Data #Subch. Data allocate Rate allocate Rate allocate Rate allocate Rate allocate Rate allocate Rate d per d per d per d per d per d per subch. subch. subch. subch. subch. subch. - - 20 5.6725 15 5.9307 - - 9 (8.7) 5.6725 - - 14 17 5.6725 - - - - 4.9542 9 (8.7) 5.6725 - - (14.3) (17.2) 14 13 10 5.6725 - - 5.6725 - - 5.9307 - - (14.3) (12.5) (10.4) CPE 12 12 Order 5.3576 - - - - 4.3923 5 (5.2) 4.9542 - - (11.4) (11.4) 12 - - 8 4.3923 5.6725 - - - - 40 4.9542 (12.5) 11 - - 12 4.9542 4.3923 7 5.3576 - - (11.4) Min. CPE 35.14 48.32 24.77 Rate Sum 431.16 416.02 421.85 Jianwei Zhang, Huawei Rate Submission Slide 127 March 2006 doc.: IEEE 802.22-06/0050r0 Layer-2 Allocation • Objective maximize the weighted system capacity given the QoS requirements and power constraints • Problem Formulation K Nc max k ,nc w k k , nc f k , nc ( Pk , nc ) wk k lQoS _ Class ( k ) k 1 nc 1 Pk ,n c subject to k ,n {0,1} k , nc c K k 1 k , nc 1 nc 0 Pk ,nc Pkmask , nc k , nc Nc nc 1 k , nc Pk ,nc Pkt k Submission Slide 128 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Proposed Algorithm (1) • Our proposed algorithm to solve Layer-2 problem is described a follows: Step 1: Initialize all the Lagrangian multipliers u k to be zeros and set ck ,n k ,n Pk ,n . c c c Step 2: Selection of temporarily optimal CPE for each subchannel given the values of u k . For every subchannel and every CPE, compute '1 uk o uk '1 uk o o Gk ,nc (u ) wk f k ,nc f k ,nc o k f k ,nc where wk wk wk u ' wk f k ,nc (0) f k',n1c ( k ) 0 wk u uk uk o 0 f k',n1c ( k ) Pkmask , nc wk ' 1 u wk f k ,nc ( Pk ,nc ) f k ,nc ( k ) Pk ,nc . ' mask mask wk Then for each subchannel, we select the CPE k ' such that k ' arg max Gk ,nc (uk ) o k and accordingly set k ,n 1, k ,n 0 for all k k ' * ' * c c Submission Slide 129 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Proposed Algorithm (2) Step 3: Compute the temporarily optimal power allocation. For each CPE in each subchannel, compute o uk c* k , nc * k , nc f ' 1 k , nc ( ) wk Step 4: Examine whether the total power limitation for each CPE is satisfied or not. Given the temporarily optimal values of ck ,nc and k ,n . * * c Nc If c nc 1 * k , nc Pkt has been satisfied for each CPE, stop. The optimal solutions have been obtained. Else, go to Step 5. Step 5: Adjust the values of u k to satisfy the total power limitations. Denote as the precision of the power allocation within a tolerance error. Submission Slide 130 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Proposed Algorithm (3) (Cont’) Nc While Pkt ck ,nc Pkt * haven’t been satisfied for all the CPE’s, n 1 c Choose the CPE k that exceeds the most the total power limitation. Set that the lower bound uk(l ) to be the current value uk and the upper bound ( uku ) to be wk f k',n (0) , where nc arg max wk fk',n (0) c c nc (ukl ) uku ) ) ( ( Set u (m) k and repeat step 2 and step 3 using u ( m ) . 2 k Nc If c nc 1 * k , nc Pkt , set u ( l ) u ( m ) ; k k Nc Elseif ck Pkt , * , nc nc 1 set uk(u ) ukm). ( Nc Repeat until P ck ,n Pkt * k t is satisfied. c nc 1 End Submission Slide 131 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 Channel Quantization · In practice, perfect channel information feedback may not be possible but limited number of bits is used instead. · A simple channel quantization algorithm is provided where the index of a quantization table based on the estimated channel power gain is used as the channel feedback. Quantization lookup table construction: Step 1: Acquire the channel power gain distribution. Step 2: Identify the range of the channel power with a desirable probability of occurrence, say 90%. Step 3: Equally partition the corresponding range in the logarithm domain. Step 4: Set up the thresholds as the middle points of each interval in the logarithm domain. Step 5: Transform the thresholds into their corresponding thresholds in the Submission original domain. Slide 132 Jianwei Zhang, Huawei March 2006 doc.: IEEE 802.22-06/0050r0 End Submission Slide 133 Jianwei Zhang, Huawei