1 / 22

An Optimal Link Layer Model for Multi-hop MIMO N etworks

An Optimal Link Layer Model for Multi-hop MIMO N etworks. Yi Shi Virginia Tech, Dept. of ECE (with Jia Liu, Canming Jiang, Cunhao Gao , and Thomas Hou ). IEEE INFOCOM 2011 – Shanghai, China. MIMO. Multiple antennas at both transmitter and receiver Benefits

javan
Download Presentation

An Optimal Link Layer Model for Multi-hop MIMO N etworks

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. An Optimal Link Layer Model for Multi-hop MIMO Networks Yi Shi Virginia Tech, Dept. of ECE (with Jia Liu, CanmingJiang, CunhaoGao, and Thomas Hou) IEEE INFOCOM 2011 – Shanghai, China

  2. MIMO • Multiple antennas at both transmitter and receiver • Benefits • Increase throughput, mitigate interference • Without additional bandwidth or transmit power IEEE INFOCOM 2011 2

  3. Current Status • Two modeling approaches • Matrix-based model • Degree of freedom (DoF)-based model • DoF-based model • Simple: Use DoF to identify a feasible rate region • Not optimal: Existing DoF-based models cannot achieve the maximum rate region An optimal DoF-based model for multi-hop MIMO networks • Matrix-based model • Accurate: Characterize MIMO channel by a matrix • High complexity: Due to matrix manipulations IEEE INFOCOM 2011 3

  4. ZFBF Scheme • DoF-based model is for the zero-force beam-forming (ZFBF) scheme • An effective MIMO technique • Two benefits associated with ZFBF • Spatial multiplexing(SM) • Enables multiple data streams on the same link • Interference cancellation(IC) • Enables more links to transmit at the same time IEEE INFOCOM 2011 4

  5. Spatial Multiplexing – An Example • Two data streams S1 and S2 • Transmitter uses two transmit weight vectors and • Transmitted signal is • Signal arriving at receiver is • Receiver uses two receive weight vectors and 1 1 0 0 IEEE INFOCOM 2011 5

  6. Interference Cancellation – An Example • Link causes interference at link • Interference for stream on link is 0 IEEE INFOCOM 2011 6

  7. Matrix-Based Model • For a time slot based scheduling, denote # of data streams on link in time slot t as • Assume each data stream has one unit rate • Link ’s average rate is • For SM, we need • For IC (if interferes with ), we need IEEE INFOCOM 2011 7

  8. Troubles with Matrix-Based Model Networking research using matrix-based model has very limited success • Need to verify the feasibility of each set of values for • The number of these sets is exponential with L • Verifying the feasibility of a particular set requires to solve a bilinear problem • A general solution to bilinear problems remains unknown IEEE INFOCOM 2011 8

  9. Understanding DoF • DoF is associated with each transmit/receive vector • Initially, each vector has no constraint • Each element in a vector can be adjusted to optimize network performance • Feasible region of this vector includes all possible values • # of DoFsof this feasible region is equal to # of elements in a vector(or # of antennas at the node) IEEE INFOCOM 2011 9

  10. Understanding DoF Consumption- An Example - • Consider a transmit vector for a node with five antennas • Initially, there is no constraint: DoFs = 5 • Consider two constraints and • The vector becomes • Remaining DoFs = 3 • Consumed DoFs = 5-3 = 2 IEEE INFOCOM 2011 10

  11. Understanding DoF Consumption- A Second Example - # of consumed DoFsdue to a set of constraints is equal to # of independent constraints • Consider three constraints • Since (7) is a linear combination of (5) and (6), we have only two independent constraints • The vector becomes • Remaining DoFs= 3 • Consumed DoFs = 5-3 = 2 IEEE INFOCOM 2011 11

  12. DoF Consumption by SM • All constraints in (8) and (9) are independent • The DoFconsumption for is • Similarly, the DoF consumption for is also Transmit vector needs to satisfy IEEE INFOCOM 2011 12

  13. DoF Consumption by IC • Interference can be cancelled by either transmit or receive vector • Which vector? • To answer this question, we need an order among vectors IEEE INFOCOM 2011 13

  14. IC DoFConsumption Under An Order • For IC, vector must satisfy for • Consider one constraint • If is determined before , uses one DoF • If is determined after , the above constraint will be satisfied by in the future -- no DoFconsumption for • Similar results hold for IEEE INFOCOM 2011 14

  15. Vector-Level to Node-Level- A Transformation - • To achieve the maximum rate region, we prove that we only need an order among nodes • An order among vectors is unnecessary • We need an order between and • If is behind , # of DoFsconsumed at is and is 0 • If is behind , # of DoFs consumed at is and is 0 IEEE INFOCOM 2011 15

  16. Total DoFs Consumed by SM & IC • Is the total number of consumed DoFs a simple sum of those by SM and IC? • The answer is Yes! • Show that there is no dependency among SM and IC constraints IEEE INFOCOM 2011 16

  17. DoF-Based Model Half-duplex constraint Constraints for node activity Ordering constraints IEEE INFOCOM 2011 17

  18. DoF-Based Model (Cont’d) DoFconsumption constraints IEEE INFOCOM 2011 18

  19. Matrix-Based Model vs. New DoF-Based model Consider a three-link network Two models achieve the same rate region Complexity comparison IEEE INFOCOM 2011 19

  20. An Application Example • Objective: Maximize the sum of weighted session rates • A linear optimization problem • Similar complexity to that for single-antenna networks IEEE INFOCOM 2011 20

  21. Node Ordering Results in Each Time Slot IEEE INFOCOM 2011 21

  22. Summary • The matrix-based MIMO model is too complex for network performance analysis • Results based on the matrix-based model are very limited • Developedan optimal DoF-based model • Retains the similar simplicity as single-antenna networks • Offers the same achievable rate region as that by the matrix-based model • Showed how to use our optimal DoF-based model for a multi-hop MIMO network problem IEEE INFOCOM 2011 22

More Related