Mobility Models & Simulation Erika Rosas Olivos Departamento de Ingeniería Informática Universidad de Santiago de Chile

Introduction

● Mobility models describe the movement patterns of mobile users: ○ Location ○ Velocity ○ Acceleration ● Mobility models influence communication patterns ● Mobility models impact performance of communication protocols ● They should target real-life movement Synthetic Mobility Models: Random Models

● Random Waypoint Model ○ Nodes move independently ○ To a random chosen destination in the simulation field ○ At a randomly velocity [0, Vmax] ○ stops at destination, pause time Tpause ○ Simple, adjust parameters for stability vs dynamism in the network ● Random Walk Model ○ Proposed to emulate unpredictable movement of particles in physics ○ RWPM with zero pause time ○ Nodes change their speed and direction at each time interval t ○ Speed follows a uniform or Gaussian distribution from [0, Vmax] ● Nodes are more likely to gather in the center of the field

Mobility Models with temporal dependency

● In real life scenarios the speed of nodes accelerate incrementally and direction change is smooth. ● Gauss-Markov Mobility Model ○ Velocity of mobile node is assumed to be correlated over time as a Gauss-Markov stochastic process. ○ Memory level is parametrized ● Smooth Random Mobility Model ○ It is observed that mobile nodes in real life tend to move at preferred speeds, rather than randomly distributed in a range. ○ Speed change is a Poisson process. ○ Direction change is an exponential distribution. Gauss-Markov Mobility model

● Velocity vector at time t ● Based on velocity at time t-1 ● W uncorrelated random Gaussian process ● Alfa represents the memory level ● Asymtotic mean and asymtotic standar deviation.

Mobility models with spatial dependency

● In some scenarios, museum, battlefield movement is influenced by a leader node in its neighbor. ● In a freeway to avoid collision the speed of vehicle cannot exceed the speed of the vehicle ahead of it. ● Reference Point Group Mobility Model ○ Example: group of soldiers may move together in a group. ○ Each group has a center (group leader) and a number of members. ○ The movement of the group leader determines the mobility behaviour of the group. ○ Group Leader Movement: Motion Vector V (random or predefined) ○ Each member of the group deviated from V by some degree (independent and identically distributed random process) in [0,Rmax] maximum allowed deviation ● Ej: Convention MM, Pursue MM, Nomadic MM. Reference Point Group Mobility Model Mobility models with geographic restrictions

● Node’s movement is restricted for example: ○ Streets, freeways ○ Obstacles, buildings ● Pathway Mobility Model (fig) ○ Restricts the node movement to the pathways in the map (predefined). ○ Map is represented as a graph. ○ Each node moves towards its destination though the shortest path along the edges, then pauses and starts a new movement. ● Obstacle Mobility Model (fig) ○ Mobile nodes are required to change trajectory when an obstacle is found. ○ Obstacles also impact way radio system, could not propagate the signal through the obstacles without atttenuation. ○ If an obstacle is between two nodes, the link is considered broken. ○ Voronoi graph is computed to construct the pathways

Trace-based Mobility Models

● Challenge in obtaining real-life traces ○ Monitoring device location (GPS, GSM, 802.11) ○ Monitoring communication between the devices to base station or between them. Assumes correlation of signal strength and distance. ● Available traces ○ CRAWDAD http://crawdad.cs.dartmouth.edu/. ○ UNC/FORTH http://netserver.ics.forth.gr/datatraces/. ○ MobiLib http://nile.cise.ufk.edu/MobiLib/. ○ Similar traces, limited number of different scenarios and applications. ○ Conference, campus, city scenarios. ● Synthetic models are not appropriate to model heterogeneous environments.

Complex Mobility Models: Submodels

● Working Day Mobility Model ○ Being at home: evening and night. Node stay still. ○ Working: Office model is a 2-D movement inside an office. (They do not model signal attenuation for indoor). Uses Pareto distribution. Length of meetings follows a log-normal distribution. ○ Evening activity with friends: Each node is assigned a favorite spot. After work it is assigned a group. Uses transport to go to the spot, waits until all nodes in the group are present and move. ● Submodels repeat everyday ● Communities from with nodes doing the same activity in the same location. ● Wake up time with normal distribution. ● Transport methods: walk, car, bus (pre-defined routes). ● Map contains location of houses, office, meeting spots, bus routes and stops. Mobility Models for Disaster Scenarios Mobility Models for Disaster Scenarios

● Map-based approaches ○ Movements of people and vehicles described over a 2-D plane. ○ OpenStreetMap: layered representation of buildings, streets, water sources, obstacles, altitude and ground conditions. ○ In disaster scenario the infrastructure may be modified. Mobility Models for Disaster Scenarios

● Event-driven paradigm ○ Fire, tsunami, collapse of a building trigger direction changes or different movement patterns of the people using mobile devices. ○ Role-based approaches: Represent different response of an entity to different events. ● Ex 1: 3 classes of behaviour ○ Fleeing from the event ○ Approaching the event ○ Oscillating from an event to a predetermined location ○ Use of physical gravitational model to determine how entities to and away from an event and how entities closer to an event are more affected by it. ● Ex 2: Events that attract attention and repel the movement of first responders. ○ An event is modeled by its central location, the attending zone, the event time and lifetime, attending role list, resolution effort.

Mobility Models for Disaster Scenarios

● Obstacle avoidance ○ Obstacles may be modeled as polygons, using a visibility graph to find the optimal path. ○ Using Dijkstra’s algorithm to compute the shortest movement. ○ Nodes may use the shortest safe path. ○ The Voronoï diagram may also be used to determine movement around obstacles (FIG)

Mobility Models for Disaster Scenarios

● Group Movement ○ First responders may be organized into microgroups. ○ A leader is defined and a set of followers move in its proximity. ○ A displace radio is defined to allow movement diversity. ○ May use RPGM. ○ Use of Levy Walk model: ■ Movement lengths and pause times follow a power law distribution. ● Cluster mobility model ○ Entities move around a particular point, called cluster center (POI). ● Dependency: ○ May assume rescue workers tend to physically separate from each other to scout unexplored areas, while maintaining in-range communication to at least one person in the group to avoid isolation. Cluster Based Model

POI: Point Of Interest

Match real-world places, attractions. Mobility Models for Disaster Scenarios

● Target Areas, DA example: ○ Incident location ○ Patient waiting treatment zone ○ Casualties clearing station ○ Transport zone ○ Hospital Zoe ○ Technical operational command. ● Areas have entry and exit point to allow movement between them ● Other example: ○ Ground Zero area (disaster area) ○ Event horizon area (reaction to the event)

Disaster Area Model Mobility Models for Disaster Scenarios ● Heterogenous Movements ○ Separation between pedestrian and vehicles (speed and movement). ○ Separation between first responders and people affected by the event (police, firefighter approach the event, ambulances may oscillate) ● Uddin example: ○ Recurrent motion between centers to simulate transportation. ○ Localized random motion for rescue workers. ○ Recurrent path motion though multiple neighborhoods for police patrols. ○ Motion switching from center to and back to a random location for ambulances. Mobility Models for Vehicular Networks

● Freeway Model: One-dimension, single or bi-directional road. ○ Road composed by lines (1 to 3) on each direction. ○ Mobile node velocity constant or variable with temporal dependency. Tools for simulating mobility

● NS-2: setdest tool from Monarch group ● BonnMotion (RWP, RW, GM, RPGM, DA, etc.) ○ Number of nodes, duration, simulation area, specific parameters of the model. ○ .params .movements.gz ○ Node-by-line waypoint based. Position at which the movement of a node changes (direction, velocity) The ONE Cellular Networks

Radio network made of a number of radio cells, each served by at least one fixed-location transceiver known as base station BTS in 2G.

● BSC: Base Station Controller ● MSC: Mobile Switching Center (telephone network operator) Cellular Networks: 3G Cellular Networks: 4G/LTE

● Mobile user: mobile device with SIM card. Called UE (User Equipment). ● Access network: manages communication between mobile user and . E-UTRAN (Evolved Universal Terrestrial Radio Access Network). ● EPC Evolved Packet Core: manages authentication, mobility, handover, routing, etc. Cellular Networks: 4G/LTE Cellular Networks: LTE handover

Handover (HO): The process of transferring an ongoing call or data session from one channel to another in a cellular network.

● The user has moved or the BS is full.

Simulating Cellular Networks

● Layout for the BSs: ○ Hexagonal, Manhattan and linear. ○ The simulation should allow any position for a BS. ● The area covered by a single antenna is called cell. ● Cell boundaries are defined by transmission power, radiation diagram, antenna gain, antenna height, etc. ● For simplicity: transmission power decays with the power of distance. ● For simplicity: devices are covered by the BS closest to them. ● BS has a limit in the number of users connected to it. Simulating Mobility for Cellular Networks

Example of Simulation LTE

● Latency: Average time in sending a packet from source to destination. ● Transfer rate: How many bits per unit of time. ● Handover decision: Which is the new cell of the mobile device? ● Handover time

For simplicity average values found in literature may be used for each time.

Fixed number of participants? Simulating Cellular Networks

Power Budget Handover Algorithm

● HOM: Handover margin is the threshold of difference in received signal strength between the source and destination cell. ● TTT: Time to Trigger timer is the timer interval that is required for satisfying the HOM condition. ● RSRP_dest > RSRP_source + HOM ● HOTrigger >= TTT Simulating Cellular Networks

Path Loss Lp(dB) - Models for specific scenarios COST 231 - Frequency (fc) 1500-2000 Mhz - Height BS (hb) 30 -200 m - Height US (hm) 1-10 m - Transmission Distance (d) 1-20km - Lp(dB) = A+B*log(d) +c

Donde: A=46,3 + 33,9log(fc) - 13,28(hb) - a(hm) B= 44,9 - 6,55 log (hb) C=0 (in small cities) a(hm)= (1,1 log(fc) -0,7)hm - (1,56 log(fc) -08)

Simulating Cellular Networks

● Increasing mobility leads to lower performance: devices that move faster need more handovers of their sessions between different BSs. ● Smaller cells lead to more handovers for the same speed. ● Others to consider: Slow fading, channel featuring,