On Group Mobility Patterns and their Exploitation to Logically Aggregate Terminals in Wireless Networks Michele Rossi †‡ , Leonardo Badia †‡ , Nicola Bui ‡ , Michele Zorzi †⋆ † Consorzio Ferrara Ricerche (CFR), ‡ University of Ferrara, Department of Engineering, via Saragat 1, 44100, Ferrara, Italy. ⋆ DEI University of Padova, via Gradenigo 6B, 35131, Padova, Italy. Abstract— Improved mobility and connectivity is one of the many challenges to be faced in future generation wireless net- works. In this paper, we first propose a novel mobility model whose aim is to describe the group mobility behavior in quite a general manner. Moreover, we present a possible on–line algo- rithm to aggregate terminals whose links remain stable over the time, i.e., showing correlated mobility patterns. Finally, we present simulation results to validate our terminal aggregation algorithms. I. I NTRODUCTION AND MOTIVATIONS Modelling mobility in wireless networks is a challenging task. New types of mobile entities are appearing everyday, with recursive or aggregate structures moving together. One of the open issues of new telecommunication systems is to support and take advantage of these mobility patterns by improving per- formance whilst at the same time increasing system efficiency. That is, data retrieval by the end user should be independent of the physical location, mobility pattern and could even take advantage of the mobility itself. For example, user mobility patterns could be exploited to predict future movements and provide resource reservation in advance. More than this, users moving in a group could elect a leader, which is typically the most powerful device, and rely on this entity to route their pack- ets. This would allow for a partial centralization of network resources with a subsequent room for increased performance. We stress that this argument is somewhat similar to what was long studied, e.g., for clustering in Ad Hoc networks, however, there are at least the following notable difference: mobility is a key issue for the presented research as we try to take advan- tage of correlated mobility structures to improve connectivity; the aim is to logically aggregate neighboring and stable users. A first investigation of the effectiveness of such a logical de- vice aggregation can be found in [1], where the authors show that substantial connectivity improvements are indeed possible. However, while the authors in [1] present a study of such a de- vice aggregation by following an analytical approach, here we focus on practical algorithms by therefore complementing their research. The intense activity to support and promote such new concepts is clearly demonstrated by the many running projects with focus on mobility. Among them, we cite here the Ambient Networks project [2]. In the current literature, several models can be found, which aim at capturing the behavior of sets of mobile entities in het- erogeneous environments [3–14]. Trying to represent users’ mobility is pivotal to provide good simulation tools and cor- rectly evaluate protocol performance. A survey on mobility models proposed up to year 2000 can be found in [3], whereas the most recent solutions are discussed in [4]. Often, mobility models try to “mimic” real mobility patterns, as well as to an- alyze the properties of these models from a statistical point of view. The simplest mobility models are the so called entity mo- bility models [3], where users move independently of each other by following random patterns. Examples falling in this category are the Random way-point [6] and the Random Walk model [8]. More refined mobility patterns are given by group mobility models, like the Reference Point Group Model (RPGM) [9] and the Structured Group Mobility Model (SGMM) [12]. Mobiles terminals are now starting to possess multiple inter- faces and therefore new possibilities open up for both operators and final users. In this respect, the aim of our current research is to promote terminal aggregation by possibly enhancing con- nectivity and, where possible, exploiting the inherent correla- tion of physical mobility patterns. In fact, as users usually tend to move in groups, we could think of exploiting their vicinity over the time to improve the connectivity of some of them and, in particular, of the ones experiencing bad channel conditions and/or whose radio technologies does not allow for e.g. a di- rect connection to a network access point (AP). In such a case, it should indeed be beneficial to elect a leader terminal, which is in the position of managing a stable communication with at least one AP and whose function will be to provide the con- nectivity to disconnected terminals through multi–hop commu- nication. Clearly, these techniques involve the thorough under- standing of mobility behaviors, the creation of logical structures that should be able to make the terminals aware of their close and stable neighbors and of their available resources. The first contribution in the present paper focuses on a gen- eral framework, in which different mobility patterns can be framed. To this end, we propose a tunable model, where several “knobs” can be regulated to adjust the behavior of the terminal mobility. The strong point of the model is a compact and clear representation of the group mobility mechanics, which is de- scribed by means of attractivity between nodes without fixed constraints. Finally, other superstructures can be imposed on top of the model to make it better adhere to a specific situa- tion, i.e., users moving along streets and/or in the presence of obstacles. 0-7803-9152-7/05/$20.00 (c) 2005 IEEE