| Abstract: |
Opportunistic, mobility-assisted, or encounter networking is a method based on ad hoc networking and introduced to disseminate data in a store-and-forward manner by means of spontaneously
connecting mobile devices. Although, in many networked systems, mobility is treated as a challenge requiring additional management, in opportunistic networks, movement facilitates networking as
it creates additional contacts between devices. These new networking opportunities can be exploited in addition to traditional wireless infrastructure networks or in absence of these networks.
Hereby, algorithms for opportunistic data dissemination make use of information about social ties, regularities in movement, and the future path of mobile entities. The availability of this
information is reasonable for areas such as campuses or conference venues, where social or professional ties are strong or when traveling by, for example, public transport lines or vehicles
following a navigation system. Other movement activities of humans in larger areas often lack this information, and new techniques are required to derive similar useful movement information. By
observing movement characteristics of network users such as average velocities or revisiting patterns, estimates about the likelihood of getting in contact with other devices can be estimated.
Our approach goes one step further by introducing users’ movement activities derived from movement patterns typical in, for example, tourist movement, shopping activities, or evening
activities. Movement activities are notions summarizing a particular movement situation that is meaningful to users and can be used to further estimate user needs and user-generated network
traffic. In case movement patterns are uncertain or fragmentary, knowledge about activities may help to faster estimate average movement characteristics. The main objective of this paper is to
detail the approach of relating activities to observed multivariate mobility characteristics on the basis of the Naïve Bayes classifier. The approach is applied to four typical urban movement use
case activities including pedestrian and vehicular movement. Results are presented on the basis of two different experimental training sets consisting of GPS outdoor traces: first, a training set
of emulated movement activities and, second, a training set consisting of labeled real-world daily activities over one month tracked by volunteers. The results of the classification study confirm
that movement can be characterized as proposed. By using mobility activities and corresponding distributions of movement characteristics, the impact of activities on opportunistic forwarding
performance in terms of contact and inter-contact time, forwarding distance and coverage of an area, and predictability of the future path of a moving device is investigated. |
