978-1-4673-4404-3/12/$31.00 ©2012 IEEE Opportunistic Forwarding Throughout Customers or Sellers in Shopping Mall Environments Adriano Galati and Karim Djemame Collaborative Systems and Performance Group School of Computing University of Leeds LS2 9JT, UK Email: {a.galati, k.djemame}@leeds.ac.uk Chris Greenhalgh Mixed Reality Lab School of Computer Science University of Nottingham NG8 1BB, UK Email: cmg@cs.nott.ac.uk Abstract—In this paper, we seek to improve understanding of human mobility patterns in environments having definite and highly organized structure, such as shopping malls. We present a method to identify individuals expressing different mobility patterns. Besides, to understand better the role of groups of message carriers expressing different mobility patterns, we performed simulations of a derivative of the Epidemic protocol with real-world mobility traces, which distinguishes between two groups of carriers, and entrusts messages through either one or the other. We discuss the implications of our results and make recommendations to guide the design of ad-hoc forwarding algorithms for delay tolerant mobile ad-hoc networks in shopping mall environments and to help modeling realistic simulation scenarios. I. I NTRODUCTION Several studies have been performed to analyse nodes’ movement in different settings: a conference environment involving conference attendees at Infocom [3], in research labs and universities in Cambridge [4], and MIT [5] involving researchers and students, in the campuses WiFi access network of UCSD (University of California, San Diego) [6], Dartmouth College [7], and ETH Zurich [8], during a Paris roller blading event [9], in an entertainment theme park [11], and in a typical office environment [12]. Although there is a range of useful data sets available no one deals with shopping malls in partic- ular. Therefore, to design and test our network application for such an environment we have conducted a field trial aiming to gather data about contacts between devices carried by humans in a shopping mall. This data set includes handheld Bluetooth devices employed in a smaller scale environment with short granularity for six days. In this paper, which follows from our previous works [1], [2], we study relationships between two people in shopping mall environments. We provide a methodology to identify devices carried by visitors/customers and devices carried by shopping mall related people based on pure contact duration, inter-contact time and frequency. Such a method can also be employed in different structured environments. We define as structured environment, a scenario having definite and highly organised structure, where people are organised by characteristic patterns of relationship and mobility. Finally, we have conducted an initial study to un- derstand better the potential roles of customers and sellers in forwarding messages in shopping mall environments. For that, we evaluate the performance of two semi-Epidemic protocols with our real-world mobility traces, which forward messages exclusively to customers or sellers. The paper is structured as follows: in Section II, we present a method to identify devices carried by customers and sellers in shopping malls; in Section III we present our initial simulation results of two semi-Epidemic forwarding protocols with real-world mobility traces and Section IV concludes the paper. II. I DENTIFYING CUSTOMERS’ AND SELLERS’DEVICES From the collected contact traces, we cannot straightaway assume external contacts as between sellers and customers because they can occur between sellers and any other de- vice. Namely, external devices do not necessarily identify customers; they could be sellers’ personal devices. Here, we provide a methodology to identify devices carried by visi- tors/customers and devices carried by shopping mall related people based on pure contact duration, inter-contact time and frequency. Namely, if an internal device spends long time in contact with an external one or they see each other very often, then, we assume that they are in ”working relationship”. As such they have specific duties and relatively predictable mobility. Figure 1 plots the correlation between the number of contacts and the contact durations of each device with our smartphones, day by day, distinguishing between internal and external devices. Contact duration is a single set of consecutive sightings of the same node, i.e. a presumed period of continuous contact. In this plot we consider ”internal-like” nodes which are external devices behaving like internals. They express contact durations longer than three hours or number of contacts bigger than 20 (contact durations are derived from contiguous data logs). We conjecture that they show contacts with devices carried by other sellers or shopping mall employees which were not part of the experiment. To identify such devices we plot the correlation between the number of contacts and the longest contact duration of each device with all our smartphones in Figure 2. Here, we can see two clusters of devices: externals on the left bottom and internals on the right bottom of the quadrant. We infer that people in the first cluster do not spend more than roughly two hours in the