Wireless%20Ad%20Hoc%20/%20Sensor%20Networks:%20Energy%20Efficiency%20and%20Cooperativeness - PowerPoint PPT Presentation

About This Presentation
Title:

Wireless%20Ad%20Hoc%20/%20Sensor%20Networks:%20Energy%20Efficiency%20and%20Cooperativeness

Description:

... neighbor and delete other links conflicted with this ... unprocessed neighbor and delete the conflicted links and repeat till all nodes are processed ... – PowerPoint PPT presentation

Number of Views:448
Avg rating:3.0/5.0
Slides: 61
Provided by: Li7
Learn more at: http://www.cs.iit.edu
Category:

less

Transcript and Presenter's Notes

Title: Wireless%20Ad%20Hoc%20/%20Sensor%20Networks:%20Energy%20Efficiency%20and%20Cooperativeness


1
Wireless Ad Hoc / Sensor NetworksEnergy
Efficiency and Cooperativeness
  • Xiang-Yang Li
  • Illinois Institute of Technology

xli_at_cs.iit.edu
2
Acknowledgment
  • Colleagues
  • Ophir Frieder, Sanjiv Kapoor, Peng-Jun Wan, Gruia
    Calinescu, Ming-Yang Kao, Zheng Sun, Xiaowen
    Chu,.
  • PhD Students
  • Yu Wang, WeiZhao Wang, WenZhan Song, Kousha
    Moaveninejad, Chih-Wei Yi
  • Support
  • NSF CCR 0311174, 0342259

3
Organization
  • Achievement Summary
  • Research on Wireless Networks
  • Students Supervising, Supervised
  • Services
  • Research
  • Wireless networks
  • Energy efficiency
  • Cooperative issues
  • Algorithm design and analysis
  • Computational geometry
  • Algorithm mechanism design
  • Conclusion

4
Research on Wireless Networks
  • Published papers (since joined IIT at 2000)
  • Journals 31 (20 published, 11 accepted)
  • 10 IEEE Transactions, 8 ACM Journals
  • Referred Conferences 57
  • 2 ACM MobiCom, 4 ACM MobiHoc, 5 IEEE INFOCOM, 1
    ACM SODA.
  • Best paper awards
  • COCOON 2001
  • IEEE HICSS 35 (2002)
  • ACM MobiCom 2005
  • one of three best paper candidates other 2 from
    MIT
  • Funding
  • NSF for Wireless CDMA assignment (co-PI)
  • NSF for workshop on Algorithms in Wireless
    Networks

5
Students
  • Students Supervised
  • Yu Wang (PhD 2004, Assistant professor at CS,
    UNCC)
  • WenZhan Song (PhD 2005, Assistant professor at
    CS, WSU)
  • Ovidiu Cristea (MS 2004), Mihai Moldovan (MS
    2005)
  • Students Supervising
  • Kousha Moaveninejad (PhD expected 2006)
  • Weizhao Wang (PhD expected 2006)
  • Ashraf Nusairat (PhD, 2004--?)
  • Yanwei Wu (PhD, 2005--?)
  • QiZhong Hu (MS)
  • Thesis Committee
  • A number of PhD and MS students

6
Services
  • To the discipline
  • Guest editor of ACM MONET, IEEE JSAC
  • Editor Ad hoc Sensor Wireless Networks An
    International Journal
  • TPC member of a number of conferences, e.g.,
  • ACM MobiHoc 2005, IEEE INFOCOM 2005, IEEE ICCCN
    2005, IEEE MASS 2005, IEEE RTSS 2004
  • Invited review of
  • NSF proposals
  • Articles for numerous well-known journals
  • Gave more than 15 invited colloquiums worldwide
  • HongKong, China, Mexico, USA
  • Invited tutorial at ACM MobiHoc

7
Services
  • To the university, department
  • Graduate student admission (2000-present)
  • Graduate Study Committee (2002-present)
  • Undergraduate Study Comm. (2000-2002)
  • Undergraduate CAMRAS Award Interviewer
  • Sophomore Leadership Retreat

8
Research
  • Main research area
  • Wireless networks
  • Energy efficiency
  • Efficient distributed algorithm design
  • Cooperative issues
  • Algorithm Design and Analysis
  • Algorithm mechanism design
  • Computational geometry
  • High quality mesh generation

9
Organization
  • Achievement Summary
  • Research on Wireless Networks
  • Students Supervising, Supervised
  • Services
  • Research
  • Wireless networks
  • Energy efficiency
  • Cooperative issues
  • Algorithm design and analysis
  • Computational geometry
  • Algorithm mechanism design
  • Conclusion

10
Wireless Ad Hoc Network
  • No wired infrastructure
  • Self-organized
  • All nodes act as routers
  • Broadcasted signal
  • Powered by battery (majority)
  • Mobile (maybe)
  • Potential Multi-hop routes

11
Energy Efficiency at Routing
  • Many routing protocols proposed
  • Metric Based Routing
  • DSR, AODV, .
  • Location Based Routing
  • GPSR, GFG, AFR,.
  • Content Based Routing
  • Which links to use
  • Shorter links more stable, thus less
    retransmission
  • Save energy possibly

12
Location Based Routing
  • Each node forwards message to best neighbor
  • E.g., best ? closest to target

t
s
13
Greedy Routing?
  • Fails to deliver

t
?
w
s
What should node w do?
14
Get out of local minimum
  • Find a planar graph
  • Gabriel Graph, for example
  • Face Routing or Right Hand Rule

t
?
w
s
15
Topology Control
  • Topology control is to select some nodes and/or
    some available links as candidates for routing
  • Backbone based structures select some nodes
  • Mainly used for broadcast, multicast
  • Typically assume that nodes power fixed
  • thus minimize the number of backbone nodes
    (MCDS)
  • Flat structures select some links, e.g., GG,LMST
  • Used for unicast, or broadcast
  • Typically assume that nodes power adjustable
  • ---thus minimize the total power (so called
    low-weight), or power to connect any pair of
    devices (so called spanner)

16
Backbone Structure
  • Select some nodes
  • Form a backbone (Connected Dominating Set)
  • each other node is connected to some node in
    backbone
  • Backbone needs to be connected
  • Our efficient distributed methods
  • Using only O(n) total messages, find a backbone
    at most 12 times optimum
  • Proved to be power spanner (fixed or adjustable)
  • Published at IEEE ICDCS02, then IEEE TPDS03

17
Flat What we want to achieve?
  • Build a single structure efficiently with a
    number of nice properties
  • Power efficient Unicast (majority operations)
  • Power efficient broadcast (widely used in WSN)
  • Bounded node degree (logical, physical)
  • Planar structure (support greedy routing)
  • Separated neighbors (directional antenna, reduce
    signal interference)
  • All these properties are achieved in a single
    structure
  • After a sequence of results

18
What do we mean by efficiently?
  • Best scenario
  • Localized method (run in constant rounds) to
    build such structure
  • Each node u quickly determines which links uv to
    keep locally
  • Our achievement
  • A semi-localized method with total communication
    cost O(n log n) bits with wireless broadcast
    model
  • Worst case still O(n) rounds

19
Our Network Model
  • A set V of n wireless nodes in 2D region
  • All nodes with same transmission power (fixed
    power)
  • Ideal case, ?
  • It induces a unit disk graph UDG
  • Two nodes are connected directly if distance at
    most one unit
  • Each node knows the position of its one-hop
    neighbors
  • Localization techniques assumed already in place

20
Adjustable Power Model
  • Power needed to support a link uv is proportional
    to
  • This model
  • Only accounts for emission power
  • Good only if long range communication, or
    techniques are used to reduce the receiving power

21
Priory Arts Some Structures
RNG
GG
MST
Yao
22
Priori Arts
published Topology Planar Unicast Spanner Low weight Degree Bound Comm. Cost
INFOCOM 01 YAOGG Yes Yes NO NO O(n)
PODC 01 CBTC No Yes No Yes O(n)
MobiHoc 01 RDG Yes Yes No No O(n2)
ICCCN 02 Yao No Yes No No O(n)
INFOCOM 03, LMST Yes No No 6 n
MobiCom 04 FLSS No Yes No No O(n)
Not completed here, due to space limit
23
Our Results
published Topology Planar Unicast Spanner Low weight Degree Bound Comm. Cost
ICCCN 01 RNG Yes No No No n
ICCCN 01 GG Yes Yes No No n
ICCCN 01 Yao No Yes No 7 (2K1)n
INFOCOM 02, TPDS LDel Yes Yes No No 60n
INFOCOM 04, TPDS LMST2 Yes No Yes 6 700 n
MONET IMRG Yes No Yes 6 7n
DialM 03, MONET BPS Yes Yes No 27 700n
MobiHoc 04, MONET OrdYaoGG Yes Yes No 12 24n
MobiHoc 04, MONET SYaoGG Yes Yes No 9 3n
MobiCom 05, LSQGG Yes Yes Yes 9 12 n
Not completed here, due to space limit
24
Power Efficient Unicast Structure
  • Assume GG has been constructed. All nodes marked
    unprocessed initially.
  • Once a node u has smallest ID among unprocessed
    neighbors, then
  • If it has processed neighbors, then it keeps the
    nearest processed neighbor and delete other links
    conflicted with this
  • Otherwise, it selects the nearest unprocessed
    neighbor and delete the conflicted links and
    repeat till all nodes are processed
  • Let SqGG be the final structure

w
Q-region
25
Structure Illustration
e
d
c
f
b
g
h
a
i
u
j
r
k
l
q
p
m
n
o
26
Properties
  • We can prove that the resulting topology is
  • Planar
  • Power efficient for unicast
  • Bounded logical node degree
  • Neighbor q-separation
  • What we miss is (counter example omitted)
  • Power efficient for broadcast (low weight)

27
Energy Efficient Broadcast
Any broadcast can be viewed as an arborescence
rooted at s
s
28
Priori Arts On Efficient Broadcast
  • Several Structures Proposed
  • MST, BIP, SPT, RNG, etc.,
  • Theoretically Good INFOCOM, WINET
  • MST, BIP within constant of optimum
  • But,
  • not localized, or even not efficient in a
    distributed way
  • Not efficient for unicast

29
Broadcast Low-weight Optimal
A structure is called low-weighted if its total
link length is within O(1) of MST
Proved previously (INFOCOM03, TPDS04) Given
any low-weighted structure H, the total power
consumption for broadcast is asymptotically best
among all locally constructed structures
Proposed several localized methods with O(n)
messages that construct a low-weighted structure
----(TPDS04, WINET05)
30
Add Low-Weight Property
  • Our previous approach
  • Given a structure, such as RNG, LMST

v
u
y
x
Any node x removes the longest link of any
quadrilateral xyvu
31
Not Efficient for Unicast Anymore
May break connectivity for graph SQGG constructed
previously
32
Our New Unified Structure
  • Build SqGG graph
  • Each node x collects 2-hop links E(x) in SqGG
  • Node x picks an incident link xy
  • with smallest ID (xy, maxID(x,y), minID(x,y))
  • If exits uv such that xygtmax(uv,3ux,3vy)
  • removes link xy from E(x)
  • Otherwise
  • keeps link xy forever
  • Let LSqGG be the final structure

33
Properties
  • We can prove that the resulting topology
  • Planar
  • Power efficient for unicast
  • Bounded logical node degree
  • Neighbor q-separation
  • Power efficient for broadcast (low weight)
  • Can be constructed efficiently using O(n) messages

34
Expected Interference
  • Interference
  • The physical degree of node u

I(u)7
35
Random Deployment
  • When nodes are of Poisson distribution
  • The maximum node interference is at least O(log
    n) for any connected structure almost surely
  • Proof omitted
  • Thus our structures also are bad in terms of the
    maximum node interference

36
Random Deployment
  • When nodes are of Poisson distribution
  • The average node interference is only O(1) for
    the following structures
  • RNG, GG, LMST, SqGG, LSqGG,..,

See our ACM MobiCom 2005 paper for more details
about these structures and proofs
37
Other Results
  • Determine Transmission Range (MobiHoc03)
  • So the induced graph has some properties almost
    surely for certain random distribution
  • The critical range for connectivity and
    k-connectivity
  • OVSF/CDMA Code Assignment (DialM03, COCOON05)
  • PTAS for IS, VC in some graphs
  • To maximize the bottleneck and throughput
  • Build CDS Efficiently (ICDCS02, TPDS03)
  • Linear messages, 12-approximation
  • Some Geometry Results
  • New structure Local Delaunay Graph (INFOCOM02,
    WADS04)
  • Spanning ratios of Beta-Skeleton (CCCG)
  • High Quality mesh generation (STOC00, SODA01)

38
Organization
  • Achievement Summary
  • Research on Wireless Networks
  • Students Supervising, Supervised
  • Services
  • Research
  • Wireless networks
  • Energy efficiency
  • Cooperative issues
  • Algorithm design and analysis
  • Computational geometry
  • Algorithm mechanism design
  • Conclusion

39
New Dimension
  • Previously, Efficient topology control
  • Time, space, communication efficiency
  • Assumption
  • Participants act as instructed
  • Not always true
  • Faulty ones ? Fault-tolerant computing
  • Malicious ones ? Security, and Trusted computing
  • Selfish ones ? Truthful computing

40
How to deal with selfish nodes?
  • Reputation based methods
  • Nodes are rated by peers
  • Detecting/punishing/avoiding
  • Pay each node its declared cost
  • Node will manipulate its declared cost to
    increase its profit
  • May reach a stable point no node will
    unilaterally change its declared cost---Nash
    Equilibrium
  • Pay each node some payment
  • Node maximizes its profit when it reports cost
    truthfully--- Dominant Strategy
  • So relieve nodes from manipulating declared cost

41
Non-Cooperative Networks
SP 1
3
4
4.5
4.8
4.9
6
SP 2
5
7
SP 3
42
Non-Cooperative Networks
  • Network Agent
  • Selfish Only interested in its own benefit
    instead of system performance
  • Rational Do what will maximize its own benefit
  • Non-Cooperative Networks
  • A set of n agents which are selfish and rational
  • For each agent, it has a set of strategies
  • Algorithm mechanism design
  • Mechanism M(O,P)
  • O determines who to be selected
  • P determines how much to pay the agents

43
Unicast
  • Node vk costs ck to relay (private knowledge)
  • Each node vk is asked to report a cost dk
  • Find the least cost path from node v0 to node v9
    based on reported costs d
  • Compute a payment pk for node vk based on d

8
7
6
7
9
5
1
7
  • Objective Find a payment pk(d) so node
    maximizes utility when dk ck

44
Truthful Unicast Scheme
  • Output O
  • Least cost path from s to t, by LCP(s, t, G)
  • Payment to a relay node vk (VCG mechanism 2nd
    price auction)
  • Remove it and its incident links
  • Compute the shortest path from s to t
  • The payment to vk is
  • Otherwise the payment is 0
  • Present a centralized method with time O(mn log
    n) to compute payment to all nodes
  • Clearly asymptotically optimum
  • IEEE Transaction on Mobile Computing, 2005

45
Multicast
  • K receiving nodes R and a source
  • Node vk costs ck to relay (private knowledge)
  • Each node vk is asked to report a cost dk
  • Find the minimum cost tree spanning all receivers
    and source node based on reported costs d
  • Compute a payment pk for node vk based on d
  • Objective Find a payment pk(d) so node
    maximizes utility when dk ck

46
LCPT Based Mechanism
  • Structure (node or link or both)
  • Calculate all shortest paths from source node to
    receivers
  • Combine these shortest paths
  • The structure is a tree called Least Cost Path
    Tree (LCPT)
  • Payment Scheme
  • Calculate the payment for node vk based on every
    LCP containing vk
  • Choosing the maximum of these payments as the
    final payment

4
3
7
3
2
47
Other Structures
  • VCG Mechanism generally does not work
  • Since finding minimum cost spanning tree is
    NP-hard.
  • Virtual Minimum Spanning tree
  • Construct the virtual complete graph K(G)
  • Nodes are receivers, plus source node
  • Edges are LCP between two end-points
  • Find the MST on K(G), say VMST(G)
  • All agents on VMST(G) are selected
  • General link weighted Steiner Tree
  • NP-Hard, constant approximation methods exist
  • Efficient computing of payments
  • General Node weighted Steiner Tree
  • NP-Hard, best approximation ratio O(ln k)
  • Efficient computing of payments

See our ACM MobiCom 2004 paper for more details
48
Multicast Cost/Payment Sharing--- cooperative
games
49
General Cost Sharing
  • Given a set of players N
  • The cost of C(S) for every is known
  • The cost is cohesive C(ST)lt C(S)C(T)
  • Fair Cost Sharing
  • For all players
  • Budget balance
  • For every subset of players S
  • Core
  • Cross-monotone

50
Multicast Cost Sharing(fixed tree)
  • Given a structure for multicast
  • The cost of each relay agent is known
  • A fixed tree from the source to all receivers
  • Share the cost among receivers
  • Budget balance, core, Cross-monotone
  • Methods
  • Equally share for downstream receivers (ELDS)

Alice
Digital Classic 20
Comcast
Bob
51
Cost Sharing (no fixed tree)
  • All receivers must get the data
  • Find an efficient tree as output
  • Share the cost of tree among receivers fairly?
  • Various concepts of fair core, etc
  • ?-Core
  • ?-Budget balance
  • core
  • Tight bound
  • No core allocation can recover more than
    fraction of cost
  • Conjecture A core allocation can recover
    fraction of cost

See STACS05 for more details
52
Cost Sharing (no fixed tree)
  • Cross monotonic ?-Core
  • ?-Budget balance
  • Core
  • Cross monotone
  • Tight bound
  • No CM ?-Core allocation can recover more than
    fraction of cost
  • of Shapley value on LCPT can recover
    fraction of cost and being a CM ?-Core!

See STACS05, INFOCOM05 for more details
53
Multicast Payment Sharing
  • Multicast payment sharing (IEEE INFOCOM 2005)
  • Given a mechanism M(O,P)
  • Example Truthful Payment for LCPT
  • How much each receiver should pay?
  • Fair Payment Sharing Scheme
  • Budget balance the payment is all agents is
    recovered
  • Cross-monotonic more receivers, less sharing
  • No negative transfer The sharing is positive
  • No free rider sharing of each receiver is within
    some bound of what it has to pay in its unicast

54
Recall LCPT Payment
  • Payment for agent vk is maxqi PkUNI(s,qi).

vk
Payment vk to is p3
55
Simple Sharing Not Works
  • Fair Sharing ELSD?

Digital Classic 20
Alice
Digital Classic HBO 60
Digital classic with HBO 60
Comcast
Bob
56
Illustration of Fair Sharing
Digital Classic with HBO 60
Digital Classic 20
Alice
Comcast
Bob
Digital Classic HBO 60
57
Sharing LCPT Payment
  • Payment for agent ek is maxqi PkUNI(s,qi).

vk
q1
q2
q3
58
Properties
  • No negative transfer
  • Budget balance
  • Cross-monotonic
  • No-free rider
  • Dummy
  • sharing is its cost if marginal payment payment
    of unicast
  • Symmetry
  • shared payments are same if two are
    interchangeable

59
Other Results
  • Algorithm mechanism design
  • General framework for binary demand games (ACM
    EC05)
  • AMD and cost sharing for set cover games
    (STACS05, TCS05)
  • Sets or elements are agents
  • DiffServ Multicast (AAIM05, COCOON05)
  • AMD design and payment sharing
  • Nash equilibrium
  • Nash equilibrium and AMD for unicast and
    multicast (ISAAC05)

60
Questions and Comments
Write a Comment
User Comments (0)
About PowerShow.com