On the trade-off between delivery delay and power consumption in opportunistic scenarios Salvatore D’Oro, Laura Galluccio, Giacomo Morabito, Sergio Palazzo CNIT Research Unit at Dipartimento di Ingegneria Elettrica, Elettronica e Informatica - University of Catania (Italy) Email: {name.surname}@dieei.unict.it Abstract—Opportunistic networking is rapidly emerging as a suitable communication paradigm to characterize social network- ing applications and contents dissemination. However, due to the sparsity of the opportunistic network, it is expected that the considered mobility model will play a crucial role in the delivery performance of the opportunistic system. Accordingly in this paper we present an analysis of the impact of different conventional mobility models on the delivery performance of op- portunistic networks. In particular, we focus on the estimation of the packet delivery delay and the power consumption associated to the delivery procedure which are the two key metrics to be traded-off according to the application requirements. I. I NTRODUCTION In opportunistic networks nodes intermittently communicate with each other only when in closest proximity. Such a communication paradigm is suitable to characterize social interactions between users moving around and exchanging information without exploiting the telecommunication infras- tructure that can be provided by the telco operators. This opportunistic communication paradigm poses several challenging issues such as energy efficiency, optimal routing design and minimization of data delivery delay and losses. Numerous studies showed that mobility can increase the performance of wireless data dissemination mechanisms [1], [2], [3] both in terms of capacity and network connectivity. In particular, in [3] it was shown that even considering very simple mobility models (e.g. random walk), but associated to dynamic variations in nodes speed, positive effects on the connectivity of the network can be observed. Obviously mobility patterns are a relevant concern in mobile communications, especially in opportunistic networks, where it has been often proposed to consider static source nodes and add few mobile sinks to speed up the delivery process. For this reason, several research studies addressed the pattern optimization problem for the definition of the best trajectory for the mobile nodes so as to increase network efficiency and reduce delay and losses. The use of mobile sinks in opportunistic networks has been proposed in the past literature on Delay Tolerant Networks (DTNs) that were designed for those scenarios in which, in principle, delivery delay and packet losses are not an issue [4]. Even if DTNs were theorized to support large delay in interplanetary communications, DTNs have been also proposed for terrestrial communications in order to support nodes mobility [4]. The use of mobile nodes in DTNs was theorized as the Mobile Ferrying (MF), a proactive mobility- assisted approach proposed for ad hoc network scenarios which utilizes a set of special mobile nodes, called message ferries, to provide communication services for nodes in the network. Similarly the DataMule concept [5] has been also proposed where mobile entities (called MULEs) pick up data from sensors when in close range, store it, and drop it to wired access points. Inspired by these papers, in this work we present a study of the impact of different mobility schemes on packet delivery and power consumption in opportunistic scenarios. More in depth, we compare the performance achievable by using simple mobility approaches such as random waypoint, random mobility and Manhattan grid for the mobile sinks and figure out a trade-off between the reduction in power consumption and the prompt data delivery. We also investigate on the effect of partial knowledge of nodes’ mobility pattern and see how this impacts on the same performance metrics. The remainder of this paper is organized as follows. In Section II the addressed scenario is described. Section III is devoted to the discussion of the analytical framework. Performance results are presented in Section IV. Finally, the concluding remarks are given in Section V. II. SCENARIO OVERVIEW In this paper we address a network scenario where N sensor devices are deployed in the network area; a mobile sink then moves around and collects the data from the sensors upon coming in their proximity [6]. We assume that each sensor device is equipped with a wireless interface that allows to communicate with other nodes in the proximity, either static sensor devices or mobile sink(s). In this way multi-hop communications can be performed and nodes can choose if transmitting their data to the mobile sink upon/when coming into proximity or exploiting other sensor devices with which they can occasionally come into contact to deliver the data to the final destination. Moreover, due to the duty cycle implemented by sensor devices to save energy and increase network lifetime, nodes can be occasionally disconnected from the network. Accordingly, in order to perform multi- hop communications we assume that nodes execute neighbor discovery protocols to identify the network topology [7]. To deliver the data into the network, each node is able to identify the next hop neighbor which is closer to the mobile sink at each time instant. Hence, each sensor node is able