Routing performance analysis of human-driven delay tolerant networks using the truncated levy walk model

Abstract

The routing performance of delay tolerant networks (DTN) is highly correlated with the distribution of inter-contact times (ICT), the time period between two successive contacts of the same two mobile nodes. As humans are often carriers of mobile communication devices, studying the patterns of human mobility is an essential tool to understand the performance of DTN protocols. From measurement studies of human contact behaviors, we find that their distributions closely resemble a form of power-law distributions called truncated Pareto. Human walk traces has a dichotomy distribution pattern of ICT; it has a power-law tendency up to some time, and decays exponentially after that time. Truncated Pareto distributions offer a simple yet cohesive mathematical model to express this dichotomy in the measured data. Using the residual and relaxation time theory [17] [4], we apply truncated Pareto distributions to quantify the performance of opportunistic routing in DTN. We further show that Truncated Levy walk (TLW) mobility model [22] commonly used in biology to describe the foraging patterns of animals [25], provide the same truncated power-law ICT distributions as observed from the empirical data, especially when mobility is confined within a finite area. This result confirms our recent finding that human walks contain similar statistical characteristics as Levy walks [22].

Self-stabilizing Mutual Exclusion and Group Mutual Exclusion for Population Protocols with Covering

“…Each request phase is followed by a finite but unbounded phase, which is controlled by BS and called an access phase. During an access phase, BS gives the token to the first registered agent visiting it (lines 9-14), waits for the token to be returned (remember that as a basic assumption, an agent uses its CS for a finite time), and then decrements req ctr (lines [15][16][17][18][19][20]. Then, BS waits for another registered agent and so on, until req ctr goes down to 0.…”

“…In case (a), if BS does not hold a token, then during the next event of BS with an agent in state out holding a token, token BS becomes true (line 16) and req ctr is decremented (line 20). From this point, BS can dispatch the token to different registered agents until req ctr ≤ 0 (lines [9][10][11][12][13][14][15][16][17][18][19][20]. Then, the end of the phase condition, in line 32, triggers.…”

“…⟨ x enters CS ⟩ 15: In case (B), BS has no token and cannot receive one, since the only possibility to receive a token is by executing lines [15][16][17][18][19][20]. In this case, these lines cannot be executed, since for any agent x, the condition token x ∧ state x = out (line 15) is false.…”

“…These data sets concern students in a campus [1,24], participants in a network conference [25], visitors at Disney World and more. All exhibit the fact that the Inter Contact Time (ICT -the time period between two successive contacts of the same two mobile agents) follows a so called truncated Pareto distribution [11,20,23]. In particular, this involves that ICT is practically bounded.…”

Accessibility and delay in random temporal networks

“…Specifically, they provide a general framework for analyzing delay problems in multi-hop networks (e.g. in delay tolerant networks [4]) using the bounds obtained in this paper.…”