Connectivity in large mobile ad-hoc networks

Similar documents
c n b n 0. c k 0 x b n < 1 b k b n = 0. } of integers between 0 and b 1 such that x = b k. b k c k c k

arxiv: v1 [math.mg] 25 Aug 2011

Minimizing movements along a sequence of functionals and curves of maximal slope

ONLINE APPENDIX: TESTABLE IMPLICATIONS OF TRANSLATION INVARIANCE AND HOMOTHETICITY: VARIATIONAL, MAXMIN, CARA AND CRRA PREFERENCES

One Class of Splitting Iterative Schemes

Lecture 12: Examples of Root Locus Plots. Dr. Kalyana Veluvolu. Lecture 12: Examples of Root Locus Plots Dr. Kalyana Veluvolu

Online Appendix for Managerial Attention and Worker Performance by Marina Halac and Andrea Prat

ON THE APPROXIMATION ERROR IN HIGH DIMENSIONAL MODEL REPRESENTATION. Xiaoqun Wang

ON A CERTAIN FAMILY OF QUARTIC THUE EQUATIONS WITH THREE PARAMETERS

Stochastic Optimization with Inequality Constraints Using Simultaneous Perturbations and Penalty Functions

New bounds for Morse clusters

Signaling over MIMO Multi-Base Systems: Combination of Multi-Access and Broadcast Schemes

CONGESTION control is a key functionality in modern

Given the following circuit with unknown initial capacitor voltage v(0): X(s) Immediately, we know that the transfer function H(s) is

The Dynamics of Learning Vector Quantization

C up (E) C low (E) E 2 E 1 E 0

EP2200 Queueing theory and teletraffic systems

into a discrete time function. Recall that the table of Laplace/z-transforms is constructed by (i) selecting to get

Theoretical Computer Science. Optimal algorithms for online scheduling with bounded rearrangement at the end

The continuous time random walk (CTRW) was introduced by Montroll and Weiss 1.

A CATEGORICAL CONSTRUCTION OF MINIMAL MODEL

Chapter 2 Sampling and Quantization. In order to investigate sampling and quantization, the difference between analog

SOME RESULTS ON INFINITE POWER TOWERS

ON A CERTAIN FAMILY OF QUARTIC THUE EQUATIONS WITH THREE PARAMETERS. Volker Ziegler Technische Universität Graz, Austria

μ + = σ = D 4 σ = D 3 σ = σ = All units in parts (a) and (b) are in V. (1) x chart: Center = μ = 0.75 UCL =

Social Studies 201 Notes for March 18, 2005

Amplify and Forward Relaying; Channel Model and Outage Behaviour

FUNDAMENTALS OF POWER SYSTEMS

OVERFLOW PROBABILITY IN AN ATM QUEUE WITH SELF-SIMILAR INPUT TRAFFIC

Social Studies 201 Notes for November 14, 2003

Energy-Distortion Tradeoff for the Gaussian Broadcast Channel with Feedback

EFFECT ON PERSISTENCE OF INTRA-SPECIFIC COMPETITION IN COMPETITION MODELS

Strong Stochastic Stability for MANET Mobility Models

Multi-dimensional Fuzzy Euler Approximation

Lecture 7: Testing Distributions

0 of the same magnitude. If we don t use an OA and ignore any damping, the CTF is

Robustness analysis for the boundary control of the string equation

two equations that govern the motion of the fluid through some medium, like a pipe. These two equations are the

Beyond Cut-Set Bounds - The Approximate Capacity of D2D Networks

Problem 1. Construct a filtered probability space on which a Brownian motion W and an adapted process X are defined and such that

Lecture 9: Shor s Algorithm

Assessment of Performance for Single Loop Control Systems

Codes Correcting Two Deletions

THE HAUSDORFF MEASURE OF SIERPINSKI CARPETS BASING ON REGULAR PENTAGON

Streaming Calculations using the Point-Kernel Code RANKERN

Lecture 15 - Current. A Puzzle... Advanced Section: Image Charge for Spheres. Image Charge for a Grounded Spherical Shell

Chapter 1 Basic Description of Laser Diode Dynamics by Spatially Averaged Rate Equations: Conditions of Validity

List coloring hypergraphs

Characterizing the Effect of Channel Estimation Error on Distributed Reception with Hard Decision Exchanges

Lecture 10 Filtering: Applied Concepts

Root Locus Diagram. Root loci: The portion of root locus when k assume positive values: that is 0

WELL-POSEDNESS OF A ONE-DIMENSIONAL PLASMA MODEL WITH A HYPERBOLIC FIELD

OPTIMAL STOPPING FOR SHEPP S URN WITH RISK AVERSION

Optimal Strategies for Utility from Terminal Wealth with General Bid and Ask Prices

IEOR 3106: Fall 2013, Professor Whitt Topics for Discussion: Tuesday, November 19 Alternating Renewal Processes and The Renewal Equation

A Short Note on Hysteresis and Odd Harmonics

DISCRETE ROUGH PATHS AND LIMIT THEOREMS

Moment of Inertia of an Equilateral Triangle with Pivot at one Vertex

Lecture 23 Date:

MATEMATIK Datum: Tid: eftermiddag. A.Heintz Telefonvakt: Anders Martinsson Tel.:

Overflow from last lecture: Ewald construction and Brillouin zones Structure factor

Selfishness need not be bad: a general proof

A NEW LOAD MODEL OF THE PEDESTRIANS LATERAL ACTION

Volume 29, Issue 3. Equilibrium in Matching Models with Employment Dependent Productivity

Center Manifolds Optimal Regularity for Nonuniformly Hyperbolic Dynamics 1

Equivalent Strain in Simple Shear Deformations

Chapter 7. Root Locus Analysis

Preemptive scheduling on a small number of hierarchical machines

Hybrid Projective Dislocated Synchronization of Liu Chaotic System Based on Parameters Identification

Department of Mechanical Engineering Massachusetts Institute of Technology Modeling, Dynamics and Control III Spring 2002

Some Sets of GCF ϵ Expansions Whose Parameter ϵ Fetch the Marginal Value

[Saxena, 2(9): September, 2013] ISSN: Impact Factor: INTERNATIONAL JOURNAL OF ENGINEERING SCIENCES & RESEARCH TECHNOLOGY

The Impact of Imperfect Scheduling on Cross-Layer Rate. Control in Multihop Wireless Networks

Çankaya University ECE Department ECE 376 (MT)

Random Sparse Linear Systems Observed Via Arbitrary Channels: A Decoupling Principle

Pulsed Magnet Crimping

Math 273 Solutions to Review Problems for Exam 1

Learning Multiplicative Interactions

MODERN CONTROL SYSTEMS

Connectivity of a Gaussian Network

Constant Force: Projectile Motion

Primitive Digraphs with the Largest Scrambling Index

White Rose Research Online URL for this paper: Version: Accepted Version

SUPPLEMENTARY INFORMATION

Automatic Control Systems. Part III: Root Locus Technique

Sampling and the Discrete Fourier Transform

Long-term returns in stochastic interest rate models

Hyperbolic Partial Differential Equations

Clustering Methods without Given Number of Clusters

What Signals Do Packet-Pair Dispersions Carry?

Avoiding Forbidden Submatrices by Row Deletions

Analytical estimates of limited sampling biases in different information measures

Hylleraas wavefunction for He. dv 2. ,! r 2. )dv 1. in the trial function. A simple trial function that does include r 12. is ) f (r 12.

arxiv: v3 [math.pr] 21 Aug 2015

Non-stationary phase of the MALA algorithm

SOLVING THE KONDO PROBLEM FOR COMPLEX MESOSCOPIC SYSTEMS

SAMPLING. Sampling is the acquisition of a continuous signal at discrete time intervals and is a fundamental concept in real-time signal processing.

Recent progress in fire-structure analysis

Unavoidable Cycles in Polynomial-Based Time-Invariant LDPC Convolutional Codes

ME2142/ME2142E Feedback Control Systems

Transcription:

Weiertraß-Intitut für Angewandte Analyi und Stochatik Connectivity in large mobile ad-hoc network WOLFGANG KÖNIG (WIAS und U Berlin) joint work with HANNA DÖRING (Onabrück) and GABRIEL FARAUD (Pari) Mohrentraße 39 10117 Berlin el. 030 20372 0 www.wia-berlin.de

A probabilitic model for large mobile ad-hoc network large bounded domain D in R d (telecommunication area) many particle X (1),..., X (N) (the uer) move randomly and independently relay principle: meage are tranmitted via a equence of hop from uer to uer; each hop ditance i R Our main quetion today: Connectivity I there a hop-path from a given uer to any other uer? How many pair of uer have a connection at a given time, over a given time lag? (in term of percentage) What i the amount of time during a given time lag over which a large percentage of the uer are connected with each other? What i the large-time aymptotic of the probability that a non-zero percentage ha a bad ervice (i.e., low connectivity)? We will attack uch quetion for N in term of a law of large number. Connectivity in large mobile ad-hoc network Edinburgh, January 2016 Seite 2 (10)

Connection time hermodynamic limit: D = D N R d with volume N. Equivalently: D independent of N, but the tranmiion radiu R i replaced by R N = N 1/d. Communication zone at time : Connectivity at time : x Connection time: D (N) = D N i=1 B ( X (i), N 1/d) N y x and y lie in the ame component of D (N). τ (N) := d 1l{X (1) N X (2) }, 0 We will invetigate the aymptotic of τ (N) in the limit N in probability w.r.t. P( X (1), X (2) ) for almot all (X (1), X (2) ). Connectivity in large mobile ad-hoc network Edinburgh, January 2016 Seite 3 (10)

Movement cheme and continuum percolation Aumption on the movement cheme he location of X (1) ha a continuou denity f : D [0, ), and P(X (1) = x X (1) = y) = 0 for <. Sufficient: the exitence of a jointly continuou denity for (X (1), X (1) ). Notion from continuum percolation θ(λ) percolation probability for a Poion point proce in R 2 with intenity λ. λ c = inf{λ > 0: θ(λ) > 0} the critical threhold. Write x > y : there exit a path from x to y within {f > λ c}. (Analogouly with and and < intead of >.) For {, >}, define τ ( ) (X (1), X (2) ) = 0 d 1l{X (1) X (2) }θ ( ) ( f(x (1) ) ) θ ( ) ( f(x (2) ) ), with θ (>) (λ) = θ(λ ) and θ ( ) (λ) = θ(λ+) the left- and right-continuou verion of θ. Connectivity in large mobile ad-hoc network Edinburgh, January 2016 Seite 4 (10)

Main reult: aymptotic of the connection time heorem: Bound on the connection time For almot every path X (1), X (2), in probability with repect to P 1,2 := P( X (1), X (2) ), τ (>) (X (1), X (2) ) lim inf τ (N) N lim up τ (N) τ ( ) (X (1), X (2) ). (1) N Comment: Global (determinitic) effect: he two walker are only connected at time if {X (1) X (2) }, i.e., if they belong to the ame component of the region where the denity of uer i high enough. Local (tochatic) effect: hey have a connection only if they locally belong to the infinitely large cluter, which ha probability θ(f (X (1) ))θ(f (X (2) )). Doe the limit exit? Under many abtract condition, all the four expreion in (1) are equal to each other almot urely, but it i cumberome to formulate and prove reaonably general and explicit one. (Difficulty here: geometry of the et {f = λ c}) Connectivity in large mobile ad-hoc network Edinburgh, January 2016 Seite 5 (10)

On the proof: convergence of the expectation (I) Auxiliary event: { G (i) N,,δ = X (i) N [ X (i) + ( δ/2, δ/2) d]}. hen connection of X (1) and X (2) i roughly equal to G (1) G (2), if they are in the ame component of {f > λ c} or {f < λ c}: Lemma 1: Approximation with G (i) For P-almot all X (1) and X (2), for almot any [0, ] and on the event {f (X (1) ) λ c} {f (X (2) ) λ c} {X (1) X (2) }, ) ( 0 = lim lim X (1) N X (2) G (1) N,,δ G(2) (1) N,,δ {X [( δ 0 N P1,2 )] > X (2) }. Connectivity in large mobile ad-hoc network Edinburgh, January 2016 Seite 6 (10)

On the proof: convergence of the expectation (I) Auxiliary event: { G (i) N,,δ = X (i) N [ X (i) + ( δ/2, δ/2) d]}. hen connection of X (1) and X (2) i roughly equal to G (1) G (2), if they are in the ame component of {f > λ c} or {f < λ c}: Lemma 1: Approximation with G (i) For P-almot all X (1) and X (2), for almot any [0, ] and on the event {f (X (1) ) λ c} {f (X (2) ) λ c} {X (1) X (2) }, ) ( 0 = lim lim X (1) N X (2) G (1) N,,δ G(2) (1) N,,δ {X [( δ 0 N P1,2 )] > X (2) }. For ufficiently mall δ, the δ-boxe around x = X (1) and y = X (2) are dijoint. Ue reult of [PENROSE (1995)]: With probability tending to one, the only way to realie G (1) i that X (1) lie in the unique giant cluter of it δ-box. Analogouly for X (2) and G (2). Contruct a equence of mutually overlapping δ-boxe from x to y inide {f > λ c} and argue imilarly for thi et. Connectivity in large mobile ad-hoc network Edinburgh, January 2016 Seite 6 (10)

On the proof: convergence of the expectation (II) Lemma 2: Convergence of the probability of G (i) Under the ame aumption, θ(f (X (1) ) )θ(f (X (2) ) ) lim inf δ 0 lim up δ 0 ( ) lim inf N P1,2 G (1) N,,δ G(2) N,,δ ( lim up P 1,2 G (1) N,,δ G(2) N,,δ N θ(f (X (1) )+)θ(f (X (2) )+). ) Aymptotic independence of G (1) and G (2) in the limit N, followed by δ 0. Connectivity in large mobile ad-hoc network Edinburgh, January 2016 Seite 7 (10)

On the proof: convergence of the expectation (II) Lemma 2: Convergence of the probability of G (i) Under the ame aumption, θ(f (X (1) ) )θ(f (X (2) ) ) lim inf δ 0 lim up δ 0 ( ) lim inf N P1,2 G (1) N,,δ G(2) N,,δ ( lim up P 1,2 G (1) N,,δ G(2) N,,δ N θ(f (X (1) )+)θ(f (X (2) )+). ) Aymptotic independence of G (1) and G (2) in the limit N, followed by δ 0. Sofar, we have proved that the expectation of τ (N) and τ (N,δ, ) = d 1l (1) G 1l (2) 0 N,,δ G 1l{ X (1) X (2) 3δ} 1l{X (1) X (2) } N,,δ are aymptotically equal in the limit N, followed by δ 0 (in term of an upper bound with = and a lower bound for =>). Connectivity in large mobile ad-hoc network Edinburgh, January 2016 Seite 7 (10)

On the proof: Vanihing variance We ue the econd-moment method and till have to prove: Propoition: τ (N,δ, ) i aymptotically determinitic Almot urely, under P 1,2 := P( X (1), X (2) ), the variance of τ (N,δ, ) N, followed by δ 0. vanihe in the limit Write out the variance, uing integral 0 d and 0 d. Left to how: independence of G (1) N,,δ G(2) N,,δ and G(1) N,,δ G(2) N,,δ. Connectivity in large mobile ad-hoc network Edinburgh, January 2016 Seite 8 (10)

On the proof: Vanihing variance We ue the econd-moment method and till have to prove: Propoition: τ (N,δ, ) i aymptotically determinitic Almot urely, under P 1,2 := P( X (1), X (2) ), the variance of τ (N,δ, ) N, followed by δ 0. vanihe in the limit Write out the variance, uing integral 0 d and 0 d. Left to how: independence of G (1) N,,δ G(2) N,,δ and G(1) N,,δ G(2) N,,δ. Let C (,N) x,δ be the bigget component of the N 1/d -ball around X (1),..., X (N) in X (1) + ( δ, δ). Abbreviate x = X (1), x = X (1), y = X (2) and ỹ = X (2). Idea: the dependence of C (,N) x,δ C (,N) y,δ of the walker (only O(δ 2d N)). and C (,N) x,δ C (,N) ỹ,δ come from only very few Main argument: our aumption that P(X (1) = x S = y) = 0 implie that the ma of thoe walker that are at time cloe to y and at time cloe to x i mall. Connectivity in large mobile ad-hoc network Edinburgh, January 2016 Seite 8 (10)

Drawback of the model For mathematical reaon, we made a number of unrealitic aumption: Number of hop for tranmitting a given meage i unbounded. All travel of any uer are global. here are no fixed additional relay intalled. No interference nor capacity problem are conidered, jut the poibility of tranferring a ingle meage. Connectivity in large mobile ad-hoc network Edinburgh, January 2016 Seite 9 (10)

Improved model More realitic movement cheme: Contained in our aumption: the random waypoint model with arbitrary dependence on time and arbitrary denity of the tarting ite. Improved verion: Pick i.i.d. home in D and an N -dependent way point meaure centred at the home with action radiu N α and infinite upport after caling. Connectivity in large mobile ad-hoc network Edinburgh, January 2016 Seite 10 (10)

Improved model More realitic movement cheme: Contained in our aumption: the random waypoint model with arbitrary dependence on time and arbitrary denity of the tarting ite. Improved verion: Pick i.i.d. home in D and an N -dependent way point meaure centred at the home with action radiu N α and infinite upport after caling. More realitic relay ytem: Bounded-hop percolation: Inert bae tation on a grid D S N Z d with ome S N 0, forbid equence of more than k hop for a meage, and look only at connection with ome of the bae tation. (Some firt reult on ergodicity of the tatic model in [HIRSCH (2015)].) Connectivity in large mobile ad-hoc network Edinburgh, January 2016 Seite 10 (10)

Improved model More realitic movement cheme: Contained in our aumption: the random waypoint model with arbitrary dependence on time and arbitrary denity of the tarting ite. Improved verion: Pick i.i.d. home in D and an N -dependent way point meaure centred at the home with action radiu N α and infinite upport after caling. More realitic relay ytem: Bounded-hop percolation: Inert bae tation on a grid D S N Z d with ome S N 0, forbid equence of more than k hop for a meage, and look only at connection with ome of the bae tation. (Some firt reult on ergodicity of the tatic model in [HIRSCH (2015)].) Include interference: Invetigate how many ignal can be uccefully received if many ignal are floating around. Criterion: Signal-to-interference ratio (SIR) and variant. Connectivity in large mobile ad-hoc network Edinburgh, January 2016 Seite 10 (10)

Improved model More realitic movement cheme: Contained in our aumption: the random waypoint model with arbitrary dependence on time and arbitrary denity of the tarting ite. Improved verion: Pick i.i.d. home in D and an N -dependent way point meaure centred at the home with action radiu N α and infinite upport after caling. More realitic relay ytem: Bounded-hop percolation: Inert bae tation on a grid D S N Z d with ome S N 0, forbid equence of more than k hop for a meage, and look only at connection with ome of the bae tation. (Some firt reult on ergodicity of the tatic model in [HIRSCH (2015)].) Include interference: Invetigate how many ignal can be uccefully received if many ignal are floating around. Criterion: Signal-to-interference ratio (SIR) and variant. Conider capacity: Handle each trajectory of a meage like a tochatic proce and inert upper bound for the number of meage per relay and time unit. Connectivity in large mobile ad-hoc network Edinburgh, January 2016 Seite 10 (10)