Mobility Management Tyler Ngo Mobility Models for Systems Evaluation A Survey By Musolesi and Mascolo Mobility Patterns Categories Trace Models Obtained by measurements of deployed systems Consists of logs of connectivity or location information Related to a specific scenario Publicly available traces are limited Synthetic Models Mathematical models sets of equations that try to capture the movement of the devices Design Synthetic Models from Real Traces Capture and model key statistical properties of the traces Reproduce and generalize them into simulators input data Social Network Theories Mobile devices are carried by humans Purely Synthetic Mobility Models Represent movements of a single node Random Walk Mobility Model Purely random movement Not suitable for wireless environment Random Waypoint Mobility Model Addition of pauses to the Random Walk Not realistic Initial placement of nodes Nodes concentrate in the middle of a bounded area Speed decay over time Random Trip Mobility Model Generalization of Random Walk and Random Waypoint Sample the initial simulation state from the stationary regime to solve the problem of reaching time stationary Synthetic Group Mobility Model Model behavior of groups that move together i e platoon of soldiers groups of students or colleagues Structured Group Mobility Model Link movements of node to the position of a subset of the other nodes of the network Not realistic Groups move randomly Memberships are hard wired nodes cannot join other groups during simulation Heterogeneous Random Walk Model Reproduce the presence of clusters that are observed in realworld traces to study the emergence of clustered networks Trace based Mobility Model Exploit available measurements to generate synthetic traces characterized by the same statistical properties of the real ones WLAN Campus Usage Traces Movements between areas of campus represented by means of Markov model Session duration data follow power law distribution Pause time and speed follow log normal distribution Movements inside downtown Osaka No reliance on any wireless measurements Empirical methodology to analyze characteristics of crowds in instants of time and maps of the city Characterization and Analytical Models of Human Connectivity Power law distributions to represent contacts duration and inter contacts time Beyond characteristic time of 12 hours the CCDF exhibits exponential decay Do not abandon Random Waypoint Model yet User registration patterns exhibit a distinct hierarchy and WLAN APs can be clustered based on registration patterns Heavy tailed Wei bull distribution to model cluster size distributions intra cluster transition probabilities and trace lengths Modeling of animal movements such as animal foraging behavior Social Network Based Mobility Models Devices are carried by humans so the movement of such devices is based on human decisions and social behavior Presence of clusters dependent on the relationships among members of the social group Fundamentally different from other types of networked systems clustering is far greater than in networks based on the stochastic models Social Network Based Mobility Models Community Based Mobility Model Hosts grouped together based on social relationships among individuals Grouping mapped to a topographical space with topography biased by the strength of social ties Movements of hosts are driven by social relationships among them Fundamental parameters such as distribution of contacts duration and inter contacts time provide good approximation of real movements Approximate power law holds over a large range of values for inter contacts time Contacts duration distribution follows a power law for a more limited range of values Social Network Based Mobility Models Community Based Mobility Model Cont Weighted graph to present social networks 0 1 0 no interactions between nodes person 1 strong social interaction Network represented by interaction matrix M mij interaction between individual i and j mij 1 where i j Interaction matrix to generate connectivity mtrix C cij 1 if mij specific threshold A host is initially positioned in a certain square in the grid A goal is assigned to host to drive movement Host i is associated to square Spq if its goal is inside Spq Each Spq exerts a certain social attractivity to a certain host The social attractivity of a square is a measure of its importance to host in terms of the social relationships Social Network Based Mobility Models Community Based Mobility Model Cont Social attractivity of square Spq towards host i SApq i w the cardinality of CSpq i e the of hosts associated to the square Spq 2 mechanisms for selection of next goal Deterministic based on selection of square that exerts the highest attractivity goals are chosen only inside squares associated to their community Probabilistic based on probability of selection of a goal in a certain square proportional to their attractivities Social Network Based Mobility Models Community Based Mobility Model Cont Probabilistic mechanism hosts randomly select next goals in other squares of the simulation area with a certain non zero probability New goals are chosen inside the same area when the input social network is composed by loosely connected communities Host may also be attracted to a different square when it has strong relationships with both communities Probability of selecting square Spq i d random value greater than 1 Connectivity Models Focuses on the evolution of the emergent connectivity graph that is changing over time as nodes move Connectivity Trace Generator CTG Probability distributions describing the patterns of collocation of mobile users i e contact duration and inter contacts time are direct inputs of a synthetic traces generation tool Input of CTG is a set of real traces processed by a trace analyzer to generate parameters describing user connectivity Parameters are the coefficients of the curves used to approximate the distributions of Inter contacts time Contract durations Link degrees characterizing the social graph of the contacts among the users Connectivity graph as basis for time varying graph of instant connectivity for each instant t Testing Tools and Mobility Modeling Most popular are NS 3 Glomosim Opnet Discrete event simulators OMNeT Parsec Different simulators show significant differences Various modeling techniques and assumptions Different methodologies followed by researchers Use of unrealistic mobility models Absence of
View Full Document