a b s t r a c tThe information dissemination problem in large-scale networking environments like wireless sensor networks and ad hoc networks is studied here considering random geometric graphs and random walk based approaches. A new type of random walk based agent is proposed in this paper and an analytical expression with respect to coverage (i.e., the proportion of the network nodes visited by the random walk agent) as a function of the number of the agent movements is derived. It is observed that the cover time of many of already existing random walk based variants is large in random geometric graphs of low degree (as it is commonly the case is wireless environments). As this inefficiency is attributed (as discussed in the paper) to the inability of existing random walk based solutions to move away from already likely covered areas, a mechanism for directional movement (i.e., jumping) of the random walk based agent is proposed and studied, that allows the agent to jump to different network areas, most likely not covered yet. The proposed mechanism (Jumping Random Walk) is studied analytically and via simulations and the parameters (of the network topology and the mechanism) under which the proposed scheme outperforms existing random walk based variations are determined.

Random Walks (RWs)

Abstract-Wireless Sensor Networks are typically bound to operate autonomously on a field, under severe energy constraints and without any centralized control. It is thus essential to develop self-organization protocols/algorithms which enable the autonomous, distributed and energy efficient network selforganization. Budget-based clustering approaches have recently been proposed for this purpose, by specifying rules for distributing a given budget of tokens to neighbors. In this paper, two strictly localized, budget-based clustering algorithms are proposed: the Directed Budget-Based (DBB) and Directed BudgetBased with Random Delays (DBB-RD). The basic, innovative idea is to utilize clustering status information that can be readily available (e.g. through the HELLO exchanges) to reduce or eliminate token distribution contentions (both intra-and intercluster) that severely limit the effectiveness of earlier budgetbased approaches. Simulation results are presented demonstrating a substantial improvement over the earlier approaches with respect to the achieved clustersizes and time to complete network decomposition.

Abstract-A mobile sink is widely considered to facilitate the data collection from energy constrained sensor fields, by having the sink come close to the sensors and conserving precious sensor node energy. The effectiveness of such a data collection approach can be measured in terms of the sensor energy conserved and the time required to collect the sensor data from the field (or, equivalently, the length of the trajectory implemented by the mobile sink).In this paper we explore two important dimensions in the design of mobile sink-based data collection schemes. One dimension refers to how close to the sensor nodes the sink moves to, to collect the data, which impacts on the transmission energy expenditure by the sensor node. The other dimension refers to the way the sink moves through the sensor field, to collect the data, which impacts on the delay in collecting the data. To capture the first dimension, the 0-hop and 1-hop data collection schemes are considered and studied; at the same time, two "extreme" approaches to the sink mobility process are considered: a (topology unaware) random walk-based sink mobility scheme and a (topology aware, optimal) deterministic sink mobility scheme. Through the analytic and simulative study presented in this paper, an understanding of the level of the trade-offs involved between the energy spent by the sensor nodes and the delay in completing the data collection process is obtained.

No abstract