Capacity Maximization in Wireless MIMO Networks with Receiver-Side - - PowerPoint PPT Presentation

capacity maximization in wireless mimo networks with
SMART_READER_LITE
LIVE PREVIEW

Capacity Maximization in Wireless MIMO Networks with Receiver-Side - - PowerPoint PPT Presentation

Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu IIT P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu Capacity Maximization in Wireless MIMO


slide-1
SLIDE 1

Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu

IIT

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 1 / 27

slide-2
SLIDE 2

Outline

Introduction Computational Hardness Practical Constant-Approximation Algorithms Summary

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 2 / 27

slide-3
SLIDE 3

Network Model

V : set of networking nodes

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 3 / 27

slide-4
SLIDE 4

Network Model

V : set of networking nodes

Each node v ∈ V has τ (v) half-duplex antennas

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 3 / 27

slide-5
SLIDE 5

Network Model

V : set of networking nodes

Each node v ∈ V has τ (v) half-duplex antennas

A: set streams of all node-level communication links

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 3 / 27

slide-6
SLIDE 6

Network Model

V : set of networking nodes

Each node v ∈ V has τ (v) half-duplex antennas

A: set streams of all node-level communication links

Along each node-level link (u, v), min {τ (u) , τ (v)} streams can be multiplexed

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 3 / 27

slide-7
SLIDE 7

Network Model

V : set of networking nodes

Each node v ∈ V has τ (v) half-duplex antennas

A: set streams of all node-level communication links

Along each node-level link (u, v), min {τ (u) , τ (v)} streams can be multiplexed

(V , A): stream-level communication topology

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 3 / 27

slide-8
SLIDE 8

Receiver-Side Interference Suppression under PrIM

When I ⊆ A transmits at the same time, the transmission by a stream a ∈ I from a sender u to a receiver v succeeds if the following constraints are satisfied:

1 Half-Duplex Constraint: u (resp. v) is not the receiver (resp.

sender) of any other stream in I.

2 Sender Constraint: u is the sender is at most τ (u) streams in I. 3 Receiver Constraint: v lies in the interference range of at most

τ (v) streams in I.

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 4 / 27

slide-9
SLIDE 9

Independent Streams

I ⊆ A is said to be independent if all streams in I succeed when they transmit at the same time.

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 5 / 27

slide-10
SLIDE 10

Independent Streams

I ⊆ A is said to be independent if all streams in I succeed when they transmit at the same time. ℐ: the collection of all independent subsets of A

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 5 / 27

slide-11
SLIDE 11

Independent Streams

I ⊆ A is said to be independent if all streams in I succeed when they transmit at the same time. ℐ: the collection of all independent subsets of A MWIS:

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 5 / 27

slide-12
SLIDE 12

Independent Streams

I ⊆ A is said to be independent if all streams in I succeed when they transmit at the same time. ℐ: the collection of all independent subsets of A MWIS:

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 5 / 27

slide-13
SLIDE 13

Independent Streams

I ⊆ A is said to be independent if all streams in I succeed when they transmit at the same time. ℐ: the collection of all independent subsets of A MWIS: Given a non-negative weight function w on A, find an independent subset I of A with maximum total weight ∑a∈Iw (a).

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 5 / 27

slide-14
SLIDE 14

Independent Streams

I ⊆ A is said to be independent if all streams in I succeed when they transmit at the same time. ℐ: the collection of all independent subsets of A MWIS: Given a non-negative weight function w on A, find an independent subset I of A with maximum total weight ∑a∈Iw (a). MIS: {0, 1}-weighted variant of MWIS

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 5 / 27

slide-15
SLIDE 15

Algorithmic Issues of MWIS

While MWIS is expected to be NP-hard, what are the major technical obstacles that have prevented the progress on provable approximations so far?

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 6 / 27

slide-16
SLIDE 16

Algorithmic Issues of MWIS

While MWIS is expected to be NP-hard, what are the major technical obstacles that have prevented the progress on provable approximations so far? Do there exist a PTAS?

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 6 / 27

slide-17
SLIDE 17

Algorithmic Issues of MWIS

While MWIS is expected to be NP-hard, what are the major technical obstacles that have prevented the progress on provable approximations so far? Do there exist a PTAS? Do there exist poly. time approx. algorithms with constant approximation bound and practical running time?

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 6 / 27

slide-18
SLIDE 18

Discoveries

NP-hard even when the input streams with positive weight are node-disjoint: due to the receiver-side interference suppression

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 7 / 27

slide-19
SLIDE 19

Discoveries

NP-hard even when the input streams with positive weight are node-disjoint: due to the receiver-side interference suppression APX-hard when the nodes have arbitrary number of antennas: due to the half-duplex constraint.

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 7 / 27

slide-20
SLIDE 20

Discoveries

NP-hard even when the input streams with positive weight are node-disjoint: due to the receiver-side interference suppression APX-hard when the nodes have arbitrary number of antennas: due to the half-duplex constraint. PTAS when the maximum number of antennas at all nodes is bounded by a constant.

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 7 / 27

slide-21
SLIDE 21

Discoveries

NP-hard even when the input streams with positive weight are node-disjoint: due to the receiver-side interference suppression APX-hard when the nodes have arbitrary number of antennas: due to the half-duplex constraint. PTAS when the maximum number of antennas at all nodes is bounded by a constant. Practical constant-approx. algorithms when all streams have uniform interference radii or all nodes have uniform number of antennas

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 7 / 27

slide-22
SLIDE 22

Roadmap

Introduction Computational Hardness Practical Constant-Approximation Algorithms Summary

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 8 / 27

slide-23
SLIDE 23

NP-Hardness

Theorem

The problem MIS is NP-hard even when restricted to node-disjoint streams with uniform interference radii and when all nodes have uniform and fixed number of antennas.

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 9 / 27

slide-24
SLIDE 24

NP-Hardness

Theorem

The problem MIS is NP-hard even when restricted to node-disjoint streams with uniform interference radii and when all nodes have uniform and fixed number of antennas. Reduction from maximum k-restricted independent set (MAX k-RIS) in UDGs

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 9 / 27

slide-25
SLIDE 25

APX-Hardness

Theorem

With uniform but arbitrarily many antennas at each node, the problem MIS is NP-hard and APX-hard even when restricted to pairwise conflicting streams.

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 10 / 27

slide-26
SLIDE 26

APX-Hardness

Theorem

With uniform but arbitrarily many antennas at each node, the problem MIS is NP-hard and APX-hard even when restricted to pairwise conflicting streams. Reduction from the maximum directed cut (MAX Di-Cut),

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 10 / 27

slide-27
SLIDE 27

PTAS

Theorem

When maxv∈V τ (v) is bounded by a constant, the problem MWIS admits a PTAS.

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 11 / 27

slide-28
SLIDE 28

PTAS

Theorem

When maxv∈V τ (v) is bounded by a constant, the problem MWIS admits a PTAS. shifting strategy + dynamic programming

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 11 / 27

slide-29
SLIDE 29

PTAS

Theorem

When maxv∈V τ (v) is bounded by a constant, the problem MWIS admits a PTAS. shifting strategy + dynamic programming Sparsity of IS: If a point o lies in the interference ranges of an independent set I of streams, then ∣I∣ = O (1).

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 11 / 27

slide-30
SLIDE 30

Roadmap

Introduction Computational Hardness Practical Constant-Approximation Algorithms

Relaxation on Half-Duplex Constraint Divide And Conquer LP-Relaxation And Rounding

Summary

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 12 / 27

slide-31
SLIDE 31

Relaxation

A set S of streams is said to be weakly independent if it satisfies the Sender Constraint and Receiver Constraint.

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 13 / 27

slide-32
SLIDE 32

Relaxation

A set S of streams is said to be weakly independent if it satisfies the Sender Constraint and Receiver Constraint. Relaxation of independence on Half-Duplex Constraint

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 13 / 27

slide-33
SLIDE 33

Extraction

Algorithm ExtractIS: extracts an independent set I from a weakly independent set S s.t. w (I) ≥ 1

4w (S) .

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 14 / 27

slide-34
SLIDE 34

Extraction

Algorithm ExtractIS: extracts an independent set I from a weakly independent set S s.t. w (I) ≥ 1

4w (S) .

Reduction to Max-Weighted Dicut

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 14 / 27

slide-35
SLIDE 35

Roadmap

Introduction Computational Hardness Practical Constant-Approximation Algorithms

Relaxation on Half-Duplex Constraint Divide And Conquer LP-Relaxation And Rounding

Summary

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 15 / 27

slide-36
SLIDE 36

Divide-And-Conquer

All streams have length ≤ one and (uniform) interference radii r > 1.

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 16 / 27

slide-37
SLIDE 37

Divide-And-Conquer

All streams have length ≤ one and (uniform) interference radii r > 1. A simple spatial divide-and-conquer algorithm with constant approximation bound.

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 16 / 27

slide-38
SLIDE 38

Divide-And-Conquer

All streams have length ≤ one and (uniform) interference radii r > 1. A simple spatial divide-and-conquer algorithm with constant approximation bound. A+: the set of streams in A with positive weight.

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 16 / 27

slide-39
SLIDE 39

Spatial Division

(a) (b) 1 2 3 4 5 6 7 8 11 9 10 11 1 2 3 4 5 6 7 8 9 10 11 1 2 3 4 5 6 7 8 9 10 11 1 2 3 4 5 6 7 8 9 10 11 1 2 3 4 5 6 7 8 9 10 11 1 2 3 4 5 6 7 8 9 10 11 1 2 3 4 5 6 7 8 9 10

Figure: Tiling of the plane into half-open half closed hexagons of diameter diameter r − 1. A stream is said to be associated with a cell if its sender lies in this cell

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 17 / 27

slide-40
SLIDE 40

Conquer

Computes a heaviest weakly independent subset S of the streams B in A+ associated with a non-empty cell. ⟨τ1, τ2, ⋅ ⋅ ⋅ , τk⟩ ← the distinct values of the number of antennas of receivers of the streams in B sorted in the ascending order

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 18 / 27

slide-41
SLIDE 41

Conquer

Computes a heaviest weakly independent subset S of the streams B in A+ associated with a non-empty cell. ⟨τ1, τ2, ⋅ ⋅ ⋅ , τk⟩ ← the distinct values of the number of antennas of receivers of the streams in B sorted in the ascending order For each 1 ≤ j ≤ k, Sj ← a weakly independent subset of streams in B whose receivers have ≥ τj antennas

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 18 / 27

slide-42
SLIDE 42

Conquer

Computes a heaviest weakly independent subset S of the streams B in A+ associated with a non-empty cell. ⟨τ1, τ2, ⋅ ⋅ ⋅ , τk⟩ ← the distinct values of the number of antennas of receivers of the streams in B sorted in the ascending order For each 1 ≤ j ≤ k, Sj ← a weakly independent subset of streams in B whose receivers have ≥ τj antennas

a matroid greedy algorithm

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 18 / 27

slide-43
SLIDE 43

Conquer

Computes a heaviest weakly independent subset S of the streams B in A+ associated with a non-empty cell. ⟨τ1, τ2, ⋅ ⋅ ⋅ , τk⟩ ← the distinct values of the number of antennas of receivers of the streams in B sorted in the ascending order For each 1 ≤ j ≤ k, Sj ← a weakly independent subset of streams in B whose receivers have ≥ τj antennas

a matroid greedy algorithm

S ← the heaviest one among S1, S2, ⋅ ⋅ ⋅ , Sk

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 18 / 27

slide-44
SLIDE 44

Combination

Cell labelling: all cells with the same label are apart at a distance > r + 1

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 19 / 27

slide-45
SLIDE 45

Combination

Cell labelling: all cells with the same label are apart at a distance > r + 1

lattice coloring with labels {0, 1, ⋅ ⋅ ⋅ , λ − 1}

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 19 / 27

slide-46
SLIDE 46

Combination

Cell labelling: all cells with the same label are apart at a distance > r + 1

lattice coloring with labels {0, 1, ⋅ ⋅ ⋅ , λ − 1}

For each 0 ≤ i ≤ λ − 1, Ji ← the union of the heaviest weakly independent subsets of the streams associated with non-empty cells with label i.

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 19 / 27

slide-47
SLIDE 47

Combination

Cell labelling: all cells with the same label are apart at a distance > r + 1

lattice coloring with labels {0, 1, ⋅ ⋅ ⋅ , λ − 1}

For each 0 ≤ i ≤ λ − 1, Ji ← the union of the heaviest weakly independent subsets of the streams associated with non-empty cells with label i. J ← the heaviest one among J0, J1, ⋅ ⋅ ⋅ , Jλ−1

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 19 / 27

slide-48
SLIDE 48

Combination

Cell labelling: all cells with the same label are apart at a distance > r + 1

lattice coloring with labels {0, 1, ⋅ ⋅ ⋅ , λ − 1}

For each 0 ≤ i ≤ λ − 1, Ji ← the union of the heaviest weakly independent subsets of the streams associated with non-empty cells with label i. J ← the heaviest one among J0, J1, ⋅ ⋅ ⋅ , Jλ−1 I ← the IS I extracted from J by ExtractIS

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 19 / 27

slide-49
SLIDE 49

Combination

Cell labelling: all cells with the same label are apart at a distance > r + 1

lattice coloring with labels {0, 1, ⋅ ⋅ ⋅ , λ − 1}

For each 0 ≤ i ≤ λ − 1, Ji ← the union of the heaviest weakly independent subsets of the streams associated with non-empty cells with label i. J ← the heaviest one among J0, J1, ⋅ ⋅ ⋅ , Jλ−1 I ← the IS I extracted from J by ExtractIS

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 19 / 27

slide-50
SLIDE 50

Combination

Cell labelling: all cells with the same label are apart at a distance > r + 1

lattice coloring with labels {0, 1, ⋅ ⋅ ⋅ , λ − 1}

For each 0 ≤ i ≤ λ − 1, Ji ← the union of the heaviest weakly independent subsets of the streams associated with non-empty cells with label i. J ← the heaviest one among J0, J1, ⋅ ⋅ ⋅ , Jλ−1 I ← the IS I extracted from J by ExtractIS

Theorem

The output I is a 4λ-approximate solution.

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 19 / 27

slide-51
SLIDE 51

Roadmap

Introduction Computational Hardness Practical Constant-Approximation Algorithms

Relaxation on Half-Duplex Constraint Divide And Conquer LP-Relaxation And Rounding

Summary

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 20 / 27

slide-52
SLIDE 52

LP-Relaxation And Rounding

Uniform number τ of antennas but arbitrary interference radii

Receiver Constraint ⇒ Sender Constraint.

LP based constant-approximation approximation algorithm

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 21 / 27

slide-53
SLIDE 53

Interference Factor

For any a, b ∈ A+, ρ (a, b) = 1/τ if the receiver of b lies within the interference range of a, ρ (a, b) = 0 otherwise.

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 22 / 27

slide-54
SLIDE 54

LP Relaxation

Compute an optimal solution x to the LP: max ∑

a∈B

w (a) x (a) s.t. ∑

b∈A+∖{a}

ρ (a, b) x (a) ≤ 1

2, ∀b ∈ A+

0 ≤ x (a) ≤ 1

2, ∀a ∈ A+.

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 23 / 27

slide-55
SLIDE 55

{0,1}-Rounding

B ← {a ∈ A+ : x (a) > 0}

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 24 / 27

slide-56
SLIDE 56

{0,1}-Rounding

B ← {a ∈ A+ : x (a) > 0} While B ∕= ∅:

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 24 / 27

slide-57
SLIDE 57

{0,1}-Rounding

B ← {a ∈ A+ : x (a) > 0} While B ∕= ∅:

pick a ∈ B arbitrarily and remove it from B.

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 24 / 27

slide-58
SLIDE 58

{0,1}-Rounding

B ← {a ∈ A+ : x (a) > 0} While B ∕= ∅:

pick a ∈ B arbitrarily and remove it from B. if ∑

b∈A+∖{a}

(w (b) w (a) ρ (a, b) + ρ (b, a) ) x (b) < 1, set x (a) ← 1; otherwise set x (a) ← 0.

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 24 / 27

slide-59
SLIDE 59

Extraction of Weak IS

J ← {a ∈ A+ : x (a) = 1}

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 25 / 27

slide-60
SLIDE 60

Extraction of Weak IS

J ← {a ∈ A+ : x (a) = 1} While J is not weakly IS:

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 25 / 27

slide-61
SLIDE 61

Extraction of Weak IS

J ← {a ∈ A+ : x (a) = 1} While J is not weakly IS:

a ← an arbitrary one in J violating Receiver Constraint

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 25 / 27

slide-62
SLIDE 62

Extraction of Weak IS

J ← {a ∈ A+ : x (a) = 1} While J is not weakly IS:

a ← an arbitrary one in J violating Receiver Constraint remove a from J

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 25 / 27

slide-63
SLIDE 63

Extraction of IS

I ← the IS I extracted from J by ExtractIS

Theorem

Suppose that the interference radius of each stream is at least η times its

  • length. Then, I is a 16

(⌈ π/ arcsin 1−1/η

2

⌉ − 1 )

  • approximate solution.

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 26 / 27

slide-64
SLIDE 64

Summary

NP-hardness and APX-hardness even in very simple settings

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 27 / 27

slide-65
SLIDE 65

Summary

NP-hardness and APX-hardness even in very simple settings PTAS in case of constant-bounded antennas

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 27 / 27

slide-66
SLIDE 66

Summary

NP-hardness and APX-hardness even in very simple settings PTAS in case of constant-bounded antennas Practical constant-approx. algorithms in case of uniform interference radii or uniform number of antennas

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 27 / 27

slide-67
SLIDE 67

Summary

NP-hardness and APX-hardness even in very simple settings PTAS in case of constant-bounded antennas Practical constant-approx. algorithms in case of uniform interference radii or uniform number of antennas Open Problem: whether there exists a practical constant-approximation algorithm in the most general setting of arbitrary interference radii and arbitrary number of antennas

P.-J. Wan, B. Xu, O. Frieder, S. Ji, B. Wang, X. Hu (IIT) Capacity Maximization in Wireless MIMO Networks with Receiver-Side Interference Suppression 27 / 27