Pengyuan Du Mario Gerla Department of Computer Science UCLA USA What is the paper all about Consider commerce in medieval France A citizen lives in a community he moves around to offer his productsservices to others and expects productsservices in turn ID: 535103
Download Presentation The PPT/PDF document "Promotion of Cooperation in Public Goods..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Promotion of Cooperation in Public Goods Game by Socialized Speed-Restricted Movement
Pengyuan Du, Mario
Gerla
Department
of Computer Science,
UCLA,
USASlide2
What is the paper all about?Consider commerce in medieval FranceA citizen lives in a community, he moves around to offer his products/services to others and expects products/services in turn
Public Good increases with commerce; but there is the risk of no returns
So, Cooperators (entrepreneurs) take risks; Defectors (
ie, conservatives) do not trade We want to understand how the movement of citizens beyond their homes can help the Public Good
2
of 26Slide3
OutlineIntroductionSystem ModelSSRM Approximated Degree Distribution
Simulation Study of Cooperation
Conclusion
3 of 26Slide4
IntroductionThe evolution of social interactions: from medieval France to Mobile CommunicationsRapid growth of mobile devices and users re-proposes the social collaboration at different levels and different Public GoodsWireless CooperationMobile users can benefit from collaboration
Collaboration may help overcome critical challenges in mobile environment, e.g. energy, connectivity, spectrum scarcity, global info scarcity, etc.
Example Applications
Packet Forwarding in DTNCooperative video streamingCrowdsourcing4 of 26Slide5
IntroductionCooperation is not freeRisk of non reciprocation by diffidents/selfishSelfishness leads to the tragedy of commonsCredit or reputation based systems stimulate cooperation
We study it with
Evol
. Game Theory (EGT)A branch of Game Theory that investigates cooperative behaviorsNo centralized management and information exchangeMobile users are involved in pairwise/group interactions, and make decisions rationallyFocusing on the evolutionarily stable strategy5 of 26Slide6
Introduction
EGT Setup
Two strategies: Cooperator (C), or Defector (D)
There is payoff from playing the social dilemma game
Reproduction model favors the “reproduction”
ie
mimicking of strategies of successful individuals (with higher payoff)
Similar to Darwinian species evolution: the fittest prevail
6
of 26
Social
Dilemma Games
Reproduction
model
Next
round
strategies
Current
strategies
1
st
Round
10000
th
RoundSlide7
Related WorkStatic Network[1]Network structure; Payoff heterogeneity; Strategic complexityMobile Network[2][3]
Random
but
homogeneous movement7 of 26
1] Perc
,
Matjaž
, et al. "Evolutionary dynamics of group interactions on
structured
populations: a review."
Journal of the royal society interface
10.80 (2013): 20120997.
[2]
Cardillo
,
Alessio
, et al. "Velocity-enhanced cooperation of moving agents playing public goods games." Physical Review E 85.6 (2012): 067101.
[
3] Antonioni
, Alberto, Marco
Tomassini
, and Pierre
Buesser. "Random diffusion and cooperation in continuous two-dimensional space." Journal of theoretical biology 344 (2014): 40-48.Slide8
ContributionsWe propose a Socialized Speed-Restricted Mobility (SSRM) model to represent realistic human movement
We show that two common social network structures, with
Power-law
and Exponential degrees, are generated by SSRMShow via analysis and simulation that SSRM mobility promotes the emergence of cooperation8 of 26Slide9
System ModelSSRM model: node i move within area A1
9
of 26
Homogeneous:
Uniform
Heterogeneous:
Exponential
-Pareto
Slide10
System ModelNetwork Large 2D circular spaceMobile users are initially distributed in the space according to Poisson Point Process with density 1 (home location)
SSRM model
Each user has a circular moving area around the home location
drawn from specific probability distributionRandom and independent movement within10 of 26Slide11
System ModelNeighbor Collection ProcessEvery user g
time steps, and communicates with neighbors within range
It “collects” neighbors met at each time
step (=
degree
)
11
of 26
Neighbor Collection Process
For Slide12
System ModelPublic Goods Game (PGG)Users initially random pick a strategy: C/D Each user plays the PGG with collected neighbors at the end of collection processCooperators contribute after each PGG game Defectors contribute nothing yet get benefit (free loaders)
12
of 26Slide13
Public Good IncreaseIn a single PGG game, the community gathers the contributions of Cooperators and multiplies them by an enhancement factor r.
The contributions are
equally distributed among all the
participants (including Defectors)The entire system is driven by the goal of increasing the PG Steady PG increase is guaranteed by strong fraction of Collaborators 13 of 26Slide14
System ModelStrategy Update at the end of every PG GameEach user i randomly picks one of its neighbors j, and borrows its strategy with probability P[
i
takes j’s strategy]=
Where is the payoff i obtained from all its PGG games. What does the MAX at numerator mean if you have already selected you neighbor??? What are you maximizing over?14 of 26Slide15
Degree DistributionWorking ApproximationApproximate with the number of users whose home locations reside in , denoted by Main result
Homogeneous case
Heterogeneous caseExponential CCDFPareto CCDF
15
of 26Slide16
Approximating SSRM Degree DistributionNumerical Validation
16
of 26Slide17
Approximating SSRM Degree Distribution
Numerical validation
17
of 26Slide18
How does Cooperation evolve?Matlab Simulation setup 10000 epochs of
time
steps
Cooperation is promoted if the fraction of cooperators is close to 1 at relatively small In theory, cooperation prevails at in an infinite, well-mixed population [1]Every data point is the average of 5 repeated simulations
18
of 26
[1] Santos
, Francisco C., Marta D. Santos, and Jorge M. Pacheco. "Social diversity promotes the emergence of cooperation in public goods games."
Nature
454.7201 (2008): 213-216.Slide19
Evolution of Cooperation Simultaneous evolution of Mobility and Strategy, the homogeneous SSRM mobility is consistent with previous random mobility studies
19
of 26
High cooperation when mobility is moderate
Low cooperation when mobility is increased toSlide20
Evolution of Cooperation Comparing homogeneous and heterogeneous SSRMThe average moving radius for cooperation to prevailHeterogeneity does not help
20
of 26Slide21
Evolution of Cooperation Cooperation is promoted under the same settings Heterogeneous SSRM is better than homogeneous case
21
of 26
Heterogeneous
Heterogeneous
Homogeneous
HomogeneousSlide22
Explaining the promotion1. Stability metricAverage number of neighbors in M epochs over the number of distinct neighbors in M epochs
22
of 26Slide23
Explaining the promotion2. Degree HeterogeneityExponential and Power-law tails are heavier than PoissonMore high degree users in the heterogeneous modelThe high degree users and their neighbors form clusters similar to social communities
23
of 26Slide24
Cooperative Behaviors of Heterogeneous Mobile UsersTwo instances of heterogeneous mobile network , Exponential withPareto withWe examine
degree, cooperation rate and strategy update frequency
Group mobile users based on the moving area
Data points are the average of 100 epochs after the evolution enters enquilibruim24 of 26Slide25
Degree in different moving area groupsCooperation rate in different moving area groupsStrategy update frequency in moving area groups
Cooperative Behaviors of Heterogeneous Mobile Users
25
of 26Slide26
Cooperative Behaviors of Heterogeneous Mobile UsersSnapshots in exponential mobile networkC-blueD-red6 “hub”s
C-green
D-
yello26 of 26Slide27
ConclusionWe employ the EGT framework to study cooperative behavior in mobile networksA socialized mobility model SSRM to drive the movementPGG and strategy update occur after every neighbor collection processWe verify that SSRM produces Exponential and Power-law degree networks
Cooperation is promoted in mobile social networks due to the
degree heterogeneity
and regular moving patterns27 of 26