陳泳宏

Mirco Musolesi and Cecilia Mascolo “ Designing Mobility Models basedon Social Network Theory ” SIGMOBILE Mobile Computing and Communications Review, July 2007

Mirco Musolesi and Cecilia Mascolo “ Designing Mobility Models basedon Social Network Theory ” SIGMOBILE Mobile Computing and Communications Review, July 2007

Introduction Design of the Mobility Model Implementation and Evaluation Conclusion

Introduction

Design of the Mobility Model

Implementation and Evaluation

Conclusion

Real movement traces have been rarely used for evaluation and testing of protocols and systems for mobile networks Synthetic models have been largely preferred the available data are limited these traces are related to very specific scenarios and their validity is difficult to generalize for simulation

Real movement traces have been rarely used for evaluation and testing of protocols and systems for mobile networks

Synthetic models have been largely preferred

the available data are limited

these traces are related to very specific scenarios and their validity is difficult to generalize

for simulation

all synthetic movement models are suspect -> it is difficult to assess to what extent they map reality random mobility models generate behaviour that is most unhuman-like movement of carried devices is necessarily based on human decisions and socialisation behavior

all synthetic movement models are suspect

-> it is difficult to assess to what extent they map reality

random mobility models generate behaviour that is most unhuman-like

movement of carried devices is necessarily based on human decisions and socialisation behavior

Using Social Networks as Input of the Mobility Model Modelling Social Relationships Detection of Community Structures Establishment of the Model: Placement of the Communities in the Simulation Space Dynamics of the Mobile Hosts

Using Social Networks as Input of the Mobility Model

Modelling Social Relationships

Detection of Community Structures

Establishment of the Model: Placement of the Communities in the Simulation Space

Dynamics of the Mobile Hosts

one of the classic ways of representing social networks is weighted graphs Modelling Social Relationships(1/3)

one of the classic ways of representing social networks is weighted graphs

the weights graphs can be represented by the 10 × 10 symmetric matrix M (Interaction Matrix) element m i,j represents the interaction between two individuals i and j

the weights graphs can be represented by the 10 × 10 symmetric matrix M (Interaction Matrix)

element m i,j represents the interaction between two individuals i and j

the Interaction Matrix is also used to generate a Connectivity Matrix the idea behind this is that we have an “interaction” threshold above which we say that two people are interacting as they have a strong relationship

the Interaction Matrix is also used to generate a Connectivity Matrix

the idea behind this is that we have an “interaction” threshold above which we say that two people are interacting as they have a strong relationship

use the algorithm proposed by Newman and Girvan to detect the presence of community structures in social networks represented by matrices (Connectivity Matrix) this algorithm is based on the calculation of the so-called betweenness of edges in order to extract the communities from the network, one of the edges of the host with the highest centrality is removed at each run the algorithm severs one edge and measures the value of Q

use the algorithm proposed by Newman and Girvan to detect the presence of community structures in social networks represented by matrices (Connectivity Matrix)

this algorithm is based on the calculation of the so-called betweenness of edges

in order to extract the communities from the network, one of the edges of the host with the highest centrality is removed

at each run the algorithm severs one edge and measures the value of Q

C1 = {A,B,C}, C2 = {D,E, F,G} and C3 = {H, I,L}. the symbol Sp,q to indicate a square in position p, q

C1 = {A,B,C}, C2 = {D,E, F,G} and C3 = {H, I,L}.

the symbol Sp,q to indicate a square in position p, q

in order to drive movement, a goal is assigned to the host a host i is associated to a square Sp,q if its goal is inside Sp,q When a goal is reached, the new goal is chosen according to the social attractivity ,w is the cardinality of C Sp,q If w = 0, the value of SA p,q i is set to 0

in order to drive movement, a goal is assigned to the host

a host i is associated to a square Sp,q if its goal is inside Sp,q

When a goal is reached, the new goal is chosen according to the social attractivity

,w is the cardinality of C Sp,q

If w = 0, the value of SA p,q i is set to 0

compared the results with the real movement patterns provided by Intel and synthetic traces generated using a Random Way-Point model a scenario composed of 100 hosts in a simulation area of 5 km × 5 km, divided into a grid composed of 625 squares We chose a relatively large simulation scenario, with a low population density

compared the results with the real movement patterns provided by Intel and synthetic traces generated using a Random Way-Point model

a scenario composed of 100 hosts in a simulation area of 5 km × 5 km, divided into a grid composed of 625 squares

We chose a relatively large simulation scenario, with a low population density

each device is equipped: an omnidirectional antenna (transmission range 250 m) free space propagation model speeds of the nodes were randomly generated according to a uniform distribution in the range [ 1 −6 ] m/s. the duration of the simulation is one day the reconfiguration interval is equal to 8 hours Goal: to verify whether the movement patterns observed in Intel traces were reproduced by our mobility mode

each device is equipped:

an omnidirectional antenna (transmission range 250 m)

free space propagation model

speeds of the nodes were randomly generated according to a uniform distribution in the range [ 1 −6 ] m/s.

the duration of the simulation is one day

the reconfiguration interval is equal to 8 hours

Goal: to verify whether the movement patterns observed in Intel traces were reproduced by our mobility mode

the social network is built starting from K full connected graphs every edge of the initial network in input is re-wired to point to a node of another cave with a certain probability p

the social network is built starting from K full connected graphs

every edge of the initial network in input is re-wired to point to a node of another cave with a certain probability p

contact duration : the time interval in which two devices are in radio range inter-contacts time : the time interval between two contacts

contact duration :

the time interval in which two devices are in radio range

inter-contacts time :

the time interval between two contacts

the influence of speed, in range [1 − 6], [1 − 10] and [1 − 20] m/s approximated with a power law function for a wide range of values

the influence of speed, in range [1 − 6], [1 − 10] and [1 − 20] m/s

approximated with a power law function for a wide range of values

the impact of the density of the population

the impact of the density of the population

Influence of the number of initial number of groups (100 hosts for 10, 20, 25 groups)

Influence of the number of initial number of groups (100 hosts for 10, 20, 25 groups)

Simulation Description: 50 hosts 1000 m × 1000m simulation area maximum node transmission range equal to 250 m two-ray pathloss model at the MAC layer bandwidth equal to 2 Mbps using CBR traffic 512 bytes(packet size), 4 packets/second (sending rate)

Simulation Description:

50 hosts

1000 m × 1000m simulation area

maximum node transmission range equal to 250 m

two-ray pathloss model

at the MAC layer

bandwidth equal to 2 Mbps

using CBR traffic

512 bytes(packet size), 4 packets/second (sending rate)

compared with RWP, the decreasing trend of the delivery ratio is less evident

compared with RWP, the decreasing trend of the delivery ratio is less evident

presenting a new mobility model based on social network theory mobility patterns are driven by the fact that devices are carried by humans and that the movements are strongly affected by the relationships between them show that the mobility model(CM) generates traces that present characteristics similar to real ones, in terms of inter-contacts time and contacts duration.

presenting a new mobility model based on social network theory

mobility patterns are driven by the fact that devices are carried by humans and that the movements are strongly affected by the relationships between them

show that the mobility model(CM) generates traces that present characteristics similar to real ones, in terms of inter-contacts time and contacts duration.