Topology Control for Effective Interference Cancellation in MultiUser

Topology Control for Effective Interference Cancellation in Multi-User MIMO Networks E. Gelal, K. Pelechrinis, T. S. Kim, I. Broustis Srikanth V. Krishnamurthy, B. Rao IEEE INFOCOM 2010

Problem Motivation & Contributions • MIMO communications are becoming prevalent w Multiple antenna elements robust links • 802. 11 n utilizes MIMO PHY w CSMA/CA no exploitation of MIMO capabilities Successive w At most one transmission each time instance Interference Cancellation • How can we realize multi-user MIMO communications? • Precoding techniques can be used w Accurate channel estimation, feedback from receiver. 2

Problem Motivation & Contributions • We design MUSIC (Multi-User MIMO with Successive Interference Cancellation) w Uses SIC for enabling Multi-user MIMO communications • Centralized and distributed approaches • Evaluation on a variety of settings w Our approach scales and the decoding error probability is bounded w MUSIC outperforms Do. F approaches. 3

Roadmap • Problem motivation & Contributions • Background • SIC w Problem formulation • Our approach • Evaluations • Conclusions 4

Background • Multi-user MIMO w Precoding techniques § § Tx sends pilot signals Rx receives pilot signals channel coefficients estimation Rx feedbacks channel coefficients to Tx Tx assigns weights at the antennas w Successive Interference Cancellation (SIC) § Receiver iteratively extracts high interfering signals § SINR requirement should be satisfied for every interferer. 5

Background • Selective diversity at Tx w Feedback from Rx to Tx for the best transmission element w One element used for subsequent transmission w Feedback is required less often than with precoding • Degrees of Freedom = k #antenna elements = k w k simultaneous transmissions are possible 6

Roadmap • Problem motivation & Contributions • Background • SIC w Problem formulation • Our approach • Evaluations • Conclusions 7

SIC • Spatial multiplexing enables multi-user MIMO with SIC Node 4 Node 3 SIC tries to remove first the stronger interferers and then decode the weaker intended signal. Node 2 Node 1 SIC 8

Models • Selection diversity and SIC • Two kinds of interferers w Strong: signal strength higher than the intended w Weak: signal strength weaker than the intended • Path loss and multipath w htr follows Rayleigh distribution, α is the path exponent, P the transmission power 9

Dealing with Weak Interferers • Maximum weak interference tolerated on link (u, v): • We want to assure that: • Assuming all interferers at the same distance as of the strongest one Aggregate weak interference follows Erlang distribution with parameters w n: number of intreferers w σ: variance of the Rayleigh distributed variable h 10

Dealing with Strong Interferers Strongest interferer P 1 Second strongest interferer P 2 d. Bm P 1/(N+P 2+P 3+…. +Pk) > γ P 2/(N+P 3+P 4+…. +Pk) > γ … Intended signal ((k-1) strongest) Pk-1/(N+Pk) > γ k stronger interferer (weak) PSUCCESFUL DECODING !! k Compact rule: Iteratively for correct decoding on link (y, z), there must be at most one interferer u, with the following interfering power: 11

Problem Formulation • Interference Graph, w Directed, edge and vertex weighted w V’ : set of links, with weight the mean value of the received signal strength w E’ : set of directed edges among the links/vertices, with weight the mean value of interference among the links connected. u v x y a(x, y) Pxy Pxv Puy b(u, v) Puv 12

Problem Formulation Time Slot 1 Time Slot 2 V 1’ links V 2’ links. . . Time Slot m Vm’ links • V 1’ V 2 ’ … V k’ = V ’ • TDMA scheme w In every time slot: ALOHA – like access with probability of failure at most δ. • Objective: minimize m NP - Hard 13

Roadmap • Problem motivation & Contributions • Background • SIC w Problem formulation • Our approach • Evaluations • Conclusions 14

C-MUSIC • The centralized algorithm is iterative. • Global knowledge of the topology • Main steps w Priority to links not scheduled w Include links that do not require SIC for decoding w Add links that can be decoded with SIC w Try to pack more links among those already scheduled 15

C-MUSIC • Two interfering links cannot belong to the same subtopology if: u u The weak interferer causes more interference than the weak interference budget The strong interference cannot be removed The two links have the same transmitter (selection diversity) A node is the transmitter for one of the links and a receiver for the other. 16

D-MUSIC Transmitter Receiver Overhearing Nodes 17

Roadmap • Problem motivation & Contributions • Background • SIC w Problem formulation • Our approach • Evaluations • Conclusions 18

Simulation Set Up OPNET simulations Traffic load: 10 -30 pkt/sec, 1500 bytes packets Path loss (α=4) and Rayleigh fading Simulations with different w Node density, SINR requirement, number of antenna elements • Metrics of interest: w Number of time slots, average decoding success probability, throughput • • • Comparison with: w Optimal (small topologies), Do. F based topology control 19

Evaluation results • MUSIC is efficient in terms of number of time slots formed Optimal C-MUSIC D-MUSIC 7. 83 9. 18 9. 64 • Density does not significantly decrease the probability of successful decoding 20

Evaluation results • Do. F based link activation cannot effectively exploit the benefits of multi-user MIMO w Do. F-based link activation leads to more decoding errors w MUSIC provides better throughput as compared with Do. F 21

Roadmap • Problem motivation & Contributions • Background • SIC w Problem formulation • Our approach: C-MUSIC • Evaluations • Conclusions 22

Conclusions • Identify the conditions for SIC to allow multi-packet reception in multi-user MIMO settings. • Design a framework for exploiting SIC • Demonstrate through simulations the applicability of our approach 23

THANK YOU ! QUESTIONS? 24