1 / 22

Scheduled Spatial Reuse with Collaborative Beamforming

Scheduled Spatial Reuse with Collaborative Beamforming. Date: 2010-04-30. Authors:. Abstract. Spatial reuse is a key aim for 802.11ad [1] potentially a major increase in aggregate intra-BSS data throughput multiple STAs transmit simultaneously (co-channel)

peigi
Download Presentation

Scheduled Spatial Reuse with Collaborative Beamforming

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Scheduled Spatial Reuse with Collaborative Beamforming Date: 2010-04-30 Authors: Thomas Derham, Orange Labs

  2. Abstract • Spatial reuse is a key aim for 802.11ad [1] • potentially a major increase in aggregate intra-BSS data throughput • multiple STAs transmit simultaneously (co-channel) • mutual interference mitigated by directional antennas • Beamforming capability of phased-array antennascan be used to our advantage • a small amount of CSI feedback to the BSS coordinator allows • collaborative beamforming to minimize mutual interference • centralized scheduling to guarantee QoS for all links • A low-complexity scheme can optimize spatial reuse in TGad Thomas Derham, Orange Labs

  3. 802.11.ad usage scenario • in dense networks, multiple links share the same channel • one uncompressed HD video link can exhaust all bandwidth in one channel • video links tend to be continuously active over extended periods • spatial reuse necessary to support all links (e.g. video + data) [2] repeater Media server FTTH BSS Set-top box Home Gateway Thomas Derham, Orange Labs

  4. Scheduled spatial reuse • beacon interval “superframe” beacon scheduled access EDCA-likeaccess CSMA/CA “directional CAP” TDMA “CTAP” • most data transmissions are scheduled • since CSMA/CA is inefficient with directional antennas [3] • BSS coordinator (PCP) schedules and transmits beacon • spatial reuse: PCP may co-schedule links in same time slot (e.g. 2x spatial reuse) PCP = PBSS Control Point [4] STA1 to STA2 STA3 to STA4 STA3 STA1 STA2 STA4 Thomas Derham, Orange Labs

  5. How should spatial reuse be controlled? • scheduling • choose which links to co-schedule based on mutual interference metric • guarantee QoS requirements for each link • beamforming • conventional beamforming gives some “natural” interference reduction • but interference may be high e.g. when multiple Rx are close together • collaborative beamforming jointly optimizes beam patterns to take into account interference due to co-scheduling • allows much higher degree of spatial reuse • requires a small amount of CSI feedback to PCP • with which scheduling can also be performed • => beamforming and scheduling of STA-STA links managed by PCP Thomas Derham, Orange Labs

  6. Scheduling process with spatial reuse • STA sends Channel Time Request to PCP for a scheduled slot • if no free slot, PCP requests certain CSI with Channel Information Request • STAs calculates and sends CSI to PCP • PCP uses CSI to perform collaborative beamforming and scheduling STA1 to STA2 < x Channel Time Request Src STA = 3, Dest STA = 4, Length = x Channel Information Request Links: STA1->STA2, STA3->STA4 Cross-links: STA1->STA4, STA3->STA2 Channel Information H1,2, H3,4, H1, 4, H3, 2 perform collaborative beamforming and scheduling STA1 to STA2 STA3 to STA4 PCP STA3 STA1 STA2 STA4 Channel Time Allocation & Beamforming vectors (1) Src STA = 1, Dest STA = 2, Start = 0, Length=x, TxBeam = w12, RxBeam = p12 (2) Src STA = 3, Dest STA = 4, Start = 0, Length=x, TxBeam = w34, RxBeam = p34 Thomas Derham, Orange Labs

  7. Stage 1: PCP determines subset of “candidate” links for co-scheduling • reduce overhead by only obtaining CSI for a subset of links • based on two metrics that approximately indicate spatial separation • already known by PCP, so no additional overhead • large receiver spatial separation => greater chance links can be co-scheduled range separation is channel strength forPCP<=>Rx STA of ith link angular separation is beamforming vector used by PCP for PCP<=>Rx STA of ith link PCP Thomas Derham, Orange Labs

  8. Stage 2: Rx STAs calculate requested CSI • using 60 GHz phased-array beam training mechanism • multiple repetitions of a beam training sequence • each time a different Tx/Rx beam training vector combination is used • beam training vectors defined in orthogonal codebook matrices • a TDMA slot is allocated to each Tx for transmission of sequences • all Rx STAs listen to beam training sequence from all Tx STAs • MIMO CSI determined between every Tx and Rx in subset link cross-link link cross-link STA3 STA1 STA2 STA4 Thomas Derham, Orange Labs

  9. Stage 2: Rx STAs calculate requested CSI (2) • MIMO channel matrices determined • MIMO system model: • element is the complex channel gain on subcarrier i withTx/Rx beam training vectors in the kth and lth columns of W and D • MIMO channel matrix estimate calculated: • limited feedback CSI sent to PCP • quantized or (for certain subcarriers) [preferred] • or, to further reduce overhead • transmit-side covariance matrix • or certain eigenvectors/eigenvalues of MIMO channel Rx training codebook Tx training codebook Thomas Derham, Orange Labs

  10. Stage 3: PCP calculates beamforming vectors • collaborative beamforming based on CSI from all STAs • beamforming vectors are (in general) not entries in training codebook • Tx beamforming vectors • maximize Signal to Leakage plus Noise Ratio (SLNR) [5] • “leakage” is interference that a given Tx causes to all other Rx • Rx beamforming vectors • maximize Signal to Interference plus Noise Ratio (SINR) • conditional on Tx beamforming vectors above own link between Tx of pth link and Rx of qth link eig{X} = dominant eigenvector of X Thomas Derham, Orange Labs

  11. Stage 4: PCP determines co-scheduling • based on SINR of each link assuming subset is co-scheduled • links are co-scheduled if for all q, where is a threshold • chosen according to QoS requirement (e.g. bit rate) for qth link • if SINRs too low, remove link with lowest SLNR from subset and recalculate • repeat until SINRs above thresholds • PCP co-schedules links in same time slot • informs STAs of scheduling (in beacon) and of beamforming vectors Thomas Derham, Orange Labs

  12. System-level simulation setup • conference room model • inter-cluster parameters between all pairs of STAs from ray-tracing [6] • correctly models interference between all STAs • TGad channel model code used for intra-cluster parameters [7] Link-n Thomas Derham, Orange Labs

  13. STA positions randomized on table • 10 pairs of STAs on 2.5 x 1 m table (random orientation) ==> dense network • scheduler rule: co-schedule all links where link SINR > 4 dB (lowest MCS) • beamforming types: (a) conventional training, (b) collaborative beamforming complementary CDF of aggregate throughput • at Pr=0.5, collaborative beamforming increases aggregate throughput byapprox. 4 Gbps (+40%) Thomas Derham, Orange Labs

  14. STA positions randomized on table (2) • 10 pairs of STAs on 2.5 x 1 m table (random orientation) ==> dense network • scheduler rule: co-schedule all links where link SINR > 4 dB (lowest MCS) • beamforming types: (a) conventional training, (b) collaborative beamforming complementary CDF of number of co-scheduled links • at Pr=0.5, collaborative beamforming increases number of links that can be co-scheduled by 50% Thomas Derham, Orange Labs

  15. STA positions as per Evaluation Methodology • 3 STA-STA pairs on table [8]: STA2=>1 (LoS), STA3=>5 (NLoS), STA7=>8 (LoS) • scheduler rule: co-schedule all links where link SINR > 4 dB (lowest MCS) • beamforming types: (a) conventional training, (b) collaborative beamforming complementary CDF of aggregate throughput complementary CDF of number of co-scheduled links no. of co-scheduledlinks increased for approx. 40% ofchannel instances Thomas Derham, Orange Labs

  16. STA positions as per Evaluation Methodology (2) • 3 STA-STA pairs on table [8]: STA2=>1 (LoS), STA3=>5 (NLoS), STA7=>8 (LoS) • scheduler rule: co-schedule all links where link SINR > 16 dB (16QAM, r=3/4) • beamforming types: (a) conventional training, (b) collaborative beamforming complementary CDF of aggregate throughput complementary CDF of number of co-scheduled links probability of spatial reuse increased 9% ==> 26% spatial reuse region Thomas Derham, Orange Labs

  17. Channel tracking and overhead • channel varies over time due to human blockage, mobility, etc • beamforming and co-scheduling should adapt by tracking channel • procedure is repeated at intervals defined by PCP • overhead of beam training mechanism similar to 802.15.3c • “clustered” tracking (reduced overhead) updates just certain elements of • in many usage cases, spatial channel characteristics static over extended time • overhead of CSI feedback to PCP is small • e.g. feedback Y or H on 8 evenly spaced subcarriers, 4x4 antenna arrays,2 bytes per complex element ==> 4 Kbytes per link • feedback of covariance matrix (or its eigenvectors) reduces overhead further • tracking updates may have much lower overhead if is fed back Thomas Derham, Orange Labs

  18. Implementation complexity • additional STA complexity is negligible • regular frequency-domain channel estimation (to obtain Y) • additional PCP complexity reduced due to efficient algorithms • calculate beamforming vectors • division by Hermitian matrix, e.g. Cholesky factorization • find dominant eigenvector, e.g. power iteration method • calculate SINRs for scheduling • matrix multiplication Thomas Derham, Orange Labs

  19. Supporting scheduled spatial reuse with collaborative beamforming in 802.11ad • STAs with beamforming capability shall support • channel information request (action frame) • channel information feedback (IE) • beamforming vector feedback (IE) • cross-link beam training mechanism • PCP shall also support • calculation of collaborative beamforming vectors • calculation of SINR for co-scheduling Thomas Derham, Orange Labs

  20. Conclusion • A method of scheduled spatial reuse with collaborative beamforming is proposed • significantly increases aggregate throughput • significantly increases the number of concurrent links that are supported • scheduling guarantees the QoS of each link • Low complexity and overhead • limited feedback and closed-form beamforming solutions • flexible to allow STAs with low capabilities to “opt out” • This method optimizes TGad to make the most of itsbeamforming capabilities Thomas Derham, Orange Labs

  21. Strawpoll • Do you support inclusion of the technique “Scheduled Spatial Reuse with Collaborative Beamforming” as described in 10/0487r0 in the TGad draft amendment? • Yes • No • Abstain Thomas Derham, Orange Labs

  22. References • [1] C. Cordeiro et al, “Spatial Reuse and Interference Mitigation in 60 GHz”, 802.11-09/0782r0 • [2] M. Park et al, “QoS Considerations for 60 GHz Wireless Networks, Globecom 2009 • [3] S. Nandagopalan et al, “MAC Channel Access in 60 GHz”, 802.11-09/0572r0 • [4] C. Cordiero et al, “Implications of Usage Models on TGad Network Architecture”, 802.11-09/0391r0 • [5] M. Lim et al, “Spatial Multiplexing in the Multi-user MIMO Downlink Based on Signal-to-Leakage Ratios”, Globecom 2007 • [6] M. Park et al, “TGad Interference Modeling for MAC Simulations”, 802.11-10/0067r0 • [7] A. Maltsev et al, “Channel Models for 60 GHz WLAN Systems”, 802.11-09/0334r7 • [8] E. Perahia et al, “Evaluation Methodology”, 802.11-09/0296r16 Thomas Derham, Orange Labs

More Related