According to these objectives of the project and the activities performed during the 1^st year of the EMBRACE project, the 2^nd year was dedicated to works related to tasks 1, 2 and 3. It is worth mentioning a considerable amount of time was required to clean, filter, and complete the datasets mentioned in the previous report, i.e., Call Detail Record (CDRs) datasets (cellular datasets collected by a major telecom operator) and to MACACO datasets (GPS datasets collected in the context of EU CHIST-ERA MACACO project). Inria has published some papers on these activities and, together with UTFPR, discussed the importance of using location-based on-line networks to study dynamic of cities.

Regarding the activities related to task 2 for “Leveraging patterns in contextual ICT enabled human behavior”, the following are the works related to human behavior analysis:

1. Mobility behavior of users:we are revisiting the seminal works on the limits of predictability of human behavior to be adapted to short-term mobility, where spatial granularity is increased; we are extracting motifs in human mobility; regularity, confinement and the preference for “paths of least resistance” of movements are also being investigated to adapt the sampling rate of human mobility as well as to optimize resource allocation in cellular networks; we are also investigating the influence of human motion on visits of locations through Twitter datasets.
2. Content demand of users: we have studied the entropy and predictability of generated volume of traffic in cellular networks; we have studied the performance of different predictors of traffic demand.
3. Contact behavior of users: we are investigating contact behaviors of users in the MACACO dataset.
4. Personality traits of users: we have performed a deep analysis on the correlation between personality traits and users’ behaviors, in terms of mobility and context extracted from MACACO dataset (the related paper is under submission to a double blinded conference and for this reason, is not registered as a technical report).

As mentioned in the previous report, we are also using other public available datasets in our analysis, such as: Geolife dataset, Shanghai CDR dataset, Twitters and Foursquare datasets.

The above-mentioned analysis is used to profile behavior of users in different scenarios and according to different needs. Such profiles will be leveraged at the design of Device-to-Device (D2D) communication, resource allocation as well as service orchestration models. Concerning such research activities, hereafter is mentioned the on-going ones:

• Inria and UFMG have started research activities on the verification of influences between visits of places in a city. In particular, we would like to identify dependencies of places on the existence of other places in a city: what is the dependency level of a bar with nearby restaurants? If such restaurants go bankrupt, will the bar lose clients and be obliged to close too? This research activity is closely related to the way people move around places. Novelty: To the best of our knowledge, this is the first work studying physical location dependability from the point of view of human mobility, using large-scale Twitters datasets. This work is related to task 2 but can also feed research on task 3, since the grouping of flow of people in different localities in a city can reveal interesting properties for spatiotemporal network resource allocation. This activity involves a PhD student from UFMG who spent 6 months visiting Inria as in internship (May-October 2018).
• UFMG and Inria have just started a collaboration on the identification of motifs in human mobility. The goal is twofold (1) to be able to improve prediction of future locations by considering the motif the user has performed in the previous timestamps and (2) to be able to identify changes in habits of movement of users. Novelty:  Literature works mostly present models to capture long-term mobility behavior, i.e., mobility features are only detectable after months of observation or when are applied to datasets with large granularity (e.g., CDRs where cell tower coordinates are known, not users’ GPS location). Here, we intend to investigate the daily human mobility patterns (i.e., short-term mobility behavior) when considering datasets with a finer granularity. This activity involves a PhD student from UFMG who spent 2 weeks visiting Inria in October/November 2018. The project of the 3-month visiting of Prof. Loureiro at Inria (December 2018-February 2019) also concerns such activity.
• Inria and UTFPR have performed research activities on adaptive channel allocation strategies for WiFi networks (i.e., corresponding to task 3 of EMBRACE work plan). Novelty: The proposed strategy considers users’ behavior in terms of mobility, traffic, and popularity in order to adaptively decide in which channel a transmission should occur. Two papers were published on the subject which concern the PhD thesis (in “quotutele” between the 2 institutions) of Roni SHIGETA, defended in July 2018.
• Inria, UTFPR, and UFMG are performing research activities on user profiling (i.e., task 2of EMBRACE work plan) on online social networking data, characterizing user visits behavior and its associated preferences with different venues in the city. A tutorial paper was recently accepted for publication in a major journal with the highest ACM impact factor: ACM Computing Surveys. Novelty: In this work, we discuss fundamental concepts of urban computing leveraging Location-Based Social Networks (LBSNs), a source of data that offers unprecedented geographic and temporal resolutions. We present a survey of recent urban computing studies that make use of LBSN data, which is helpful to exemplify research trends and techniques commonly used. Besides, we point out the opportunities and challenges that those studies open up. Some of the challenges are being addressed in the context of EMBRACE.
• UFG have continued the work on optimization models involving wireless resources, mainly 5G, and data, mainly concerning user information. Papers have been published or are under development. The work started last year in collaboration with Inria, which is related to the leveraging of mobility, content demand, and interactions of mobile users in the design of user-aware optimization models for resource allocation in 5G wire- less networks. Novelty:  Related literature made optimistic assumptions about the resource availability and synergies between the services, what is inappropriate in the context of 5G networks. We will bring to resource allocation strategies, the capability to capture uncertainty and heterogeneity of edge wireless networks. The Synthetic Mobile Data Traffic Generator (our MDTGen generator, designed by E. Mucceli) is being used here. This work involves a PhD student from UFG who spent 2 weeks last year visiting Inria. Another work started this year, involving a PhD student from UFG and Inria, which is about recommendation algorithms for tourist itinerary that consider network and computing resources. Novelty: The proposed methodology introduces the context of 5G networks, leveraging its enhanced networking and computing infrastructure, which to the best of our knowledge, is the first work to approach the problem in this context. Such activities correspond to task 3 of EMBRACE work plan.

Besides the described research collaboration and as mentioned in the previous report, partners have the common motivation of writing a position paper discussing the main outcomes of our research as well as the implications of human behavior in the future of mobile networking. The paper will regroup the expertise of all partners on the “Human and Mobile Network” loop and provide discussions and main insights on the evolution and future of mobile networking research. Although we have advanced on the writing of some sections, we need more time to finish the paper.

Next year’s work program

Next year activities involve the on-going ones described in section 2, which we plan to finish/submit the papers mentioned in the previous sections. In particular, the writing/submission of the position paper structured after the kickoff meeting of October 2017 at UFMG.