A Picture of Present Ubicomp Research Exploring Publications from Important Events in the Field

In this work we use a dataset of papers published in top conferences focused on ubiquitous computing (ubicomp) to provide an overview and analysis of recent ubiquitous computing research performed internationally and in Brazil. The contributions of this study are twofold. First, we extracted useful information from our dataset such as representativeness of authors and institutions, and the formation of communities. Second, we analyzed all papers published between 2010 and 2011 in all top international conferences, creating a taxonomy of recent ubicomp research performed internationally. After that we mapped SBCUP papers (Brazilian ubicomp conference) according to this taxonomy, which enables the comparison of international and national research. This study is useful to guide novices in the field and it also provides experienced researchers with facts enabling the discussion of ubicomp research.


Introduction
The future world envisioned by Mark Weiser considers a computing environment in which each person is continually interacting with many wireless interconnected devices (Weiser, 1993).A fundamental aspect of this world will be ubiquitous computing (ubicomp or pervasive computing), as named by Weiser.As a good research area, ubicomp gave us more questions than answers (Weiser et al., 1999), and many of them are still open (Caceres and Friday, 2012).There are many researchers around the world working on different aspects of those ubicomp challenges, including Brazilians.Today there are three notable international conferences1 (TOPint group) devoted primarily to the area: Ubicomp2 (the oldest one, now in its 14 th edition), Pervasive3 , and Percom4 .The Brazilian event dedicated to this field is SBCUP5 , which is in its 4 th edition.
The ubicomp community, not only in Brazil, is in its infancy.In this case it is natural to have many newcomers to this community.With that in mind, this paper provides an overview of recent ubicomp research conducted abroad and in Brazil.
The contributions of this work are twofold.First, we collected information about all papers published in Ubicomp, Pervasive and Percom, and performed a data mining process, extracting statistics such as most productive authors and institutions.We also analyzed the collaboration among authors identifying, for instance, the formation of communities.Second, we analyzed all papers published in 2010 and 2011 in the TOPint group, creating a taxonomy of recent ubicomp research.We used this taxonomy to discuss the research being performed in Brazil by analyzing SBCUP papers and, thus, to compare recent research results.
The rest of this work is organized as follows.Section "Related Work" describes the related work.Section "Ubicomp in Numbers" presents statistics about authors, papers and institutions related to papers published in the aforementioned conferences.Section "Representativeness of Authors and Institutions" discusses the representativeness of authors and institutions.Section "Collaboration Network" analyzes the author's collaboration network.Section "Recent Ubicomp Research" presents the taxonomy of recent research on ubicomp performed abroad and in Brazil.Finally, Section "Conclusions" concludes the paper.

Related Work
Related to the first part of our work, i.e., statistics and analysis, we can cite the following studies.In Maia et al. (2012) the authors collected some data from all papers published between 1983 and 2012 in the SBRC 6 .With that dataset, they built an author's collaboration network, which was analyzed using complex network metrics.The studies of Procópio et al. (2011) and Freire and Figueiredo (2011) also built and analyzed a collaboration network.The first one considers authors from SBBD 7 .Besides applying complex network metrics, they also presented some statistics about SBBD.The second one considers two collaboration networks: global (composed of all authors who published in a paper listed in DBLP) and Brazilian (subset of the previous network, composed only of authors from Brazil).Besides the characterization of basic information on those networks, they created a metric, called ranking, which was used to measure the importance of individuals, as well as groups.Kaye (2009) presented statistical aspects of CHI and its sister conferences.Bazzan and Argenta (2011) studied the main features of the social network created by the Program Committee members of conferences sponsored by the Brazilian Computing Society.The relationships between nodes of this network were established according to co-authorship data extracted from the DBLP Computer Science Bibliography database.The authors show that social network does not fit any well-established pattern when compared to other networks discussed in the literature.Nascimento et al. (2003) analyzed the SIGMOD co-authorship graph, finding, for instance, that the SIGMOD community is a small world network.
Regarding the second part of our work (taxonomy of recent research), the closest study, to the best of our knowledge, is the work performed by Satyanarayanan (2001).That paper introduces a taxonomy and the way research problems in pervasive computing relate to those in mobile computing and distributed systems.Our taxonomy differs from Satyanarayanan (2001) because its concern is to present research that has been published recently in top events.As discussed in Section "Recent Ubicomp Research", many problems discussed in Satyanarayanan (2001) are being studied today, such as energy and location aware systems.

Ubicomp in Numbers
This section presents the dataset used throughout this work and several statistics for the conferences considered.

Dataset
We collected all articles published in all editions of Ubicomp, Pervasive, Percom (group called TOPint), and SBCUP, from the first edition until the edition held in 2011 (main track only for all events, not including, for example, demos).For each paper we collected: title, year of publication, authors, and their respective institutions and location.
We are not considering the SBCUP dataset in all statistics and analysis because sometimes it does not make much sense due to its small size.

Analysis of Publications
Organized by Karlsruhe University in Germany, the first edition of Ubicomp, called International Symposium on Handheld and Ubiquitous Computing (HUC), occurred in 1999.In 2001, the word 'Handheld' was omitted and the symposium began to be called Ubicomp.In 2011, the event was in its thirteenth edition.It is organized in full papers sections and notes, workshops, posters, panels, live demonstrations, video presentations, doctoral colloquia, and tutorials.Considering the submission and acceptance of full papers at the Ubicomp conference, the average acceptance rate is approximately 19%, as shown in Figure 1a.The Pervasive conference has a similar scope of Ubicomp and the first edition occurred in Switzerland in 2002.The acceptance rate of full papers is about 16%, as shown in Figure 1b.Both conferences have similar standards and criteria for acceptance.
The Percom conference presents advances in the field of pervasive computing and communications.The first edition was held in the USA in 2003.Considering full papers in the Percom conference, the average acceptance rate is of approximately 16%, as shown in Figure 1c.
Authors have published 397 papers in 13 editions of Ubicomp, 308 papers in nine editions of Percom, and 203 papers in nine editions of Pervasive.This amounts to a total of 908 papers and 2,239 unique authors.In SB-CUP we have 41 papers and 103 authors (77% in just one edition).
Percom, Pervasive and Ubicomp publish by year an average of 34 (standard deviation σ=9), 22 (σ =3) and 30 (σ=8.78) papers, respectively, while SBCUP publishes 14 papers on average.Figure 2a shows the evolution of published papers in the TOPint group.Percom decreased from 54 to 27 papers in the last year.On the other hand, Ubicomp increased from 23 in 1999 to 50 papers in 2011.Pervasive keeps stable in almost all editions − the maximum was 27 papers in 2008.
Figure 2b shows that almost 50% of the papers published were written by two or three authors (439 out of 908 papers).The number of authors has decreased over the years.In fact, after 2006, there was no paper published by a single author in TOPint.Ubicomp holds the record of number of authors per paper: two articles with 13 authors.
The CDF of the number of papers per authors in TOPint is presented in Figure 2c.Around 80% of the authors published only one paper in all conferences, 98% of authors published at most five papers, 0.013% published between six and ten papers and 0.007% published more than 10 papers.The top three authors in the TOPint group are Gregory D. Abowd, Anind K. Dey and Shwetak Patel.
Considering all papers from the TOPint group, most of the publications come from the USA (20%), and in the second to the fifth positions we have, respectively, the UK (5.4%), Germany (3.6%), Japan (2.3%), and Switzerland (2.3%).Brazil is the only country from Latin America, with just one paper.A curious fact is that the only Brazilian author (first author of a paper published in 2011 at Ubicomp) does not have ubiquitous computing as his main research area.

Analysis of Authors
The TOPint group has a total of 2,548 authors distributed as follows: 1,103 in Ubicomp, 613 in Pervasive, and 832 in Percom.However, some authors have published in more than one conference, and, thus, we have a total of 2,239 different authors.In that universe, 1,954 authors published only in one of the three conferences, 256 authors in two conferences, and 26 in all of them.Figure 3 shows the intersection between authors and conferences.Each circle represents a conference (with its authors), and the graph also shows the intersections between them.We see a major interest in Pervasive and Ubicomp, with 169 authors in common.We believe this occurs because Percom is a broader conference in computer networks, while Pervasive and Ubicomp focus on ubiquitous computing.The authors Anind K. Dey, Gregory D. Abowd and Shwetak Patel published at least one paper in nine editions of Ubicomp.Another interesting fact is that Shwetak Patel published in nine consecutive editions of Ubicomp.Gregory D. Abowd has the largest number of publications in Pervasive (seven), all consecutive, followed by Gaetano Borriello, Gerhard Troster, and Shwetak Patel (each one with six publications).In Percom, Lionel Ni has the largest number of publications (seven).It is interesting to note that Lionel Ni has no involvement in Ubicomp or Pervasive, and the top publishers of these conferences did not publish in Percom (except for Gerhard Troster with one paper in 2010).Among the top publishers, Abowd and Patel have twelve joint publications, and Abowd and Dey have two joint publications.
Figure 4a presents the number of authors by their streak length of publications per year.Most authors published a maximum of two years in a row.This is in accordance with the information observed in Figure 2c (most authors publish just one paper).Figure 4b con-siders the longest period of absence of publications by a given author, i.e., the maximum period between two publications.Most of the returning authors have a two-year maximum of absence.
Figure 4c shows the fraction of newcomers per year.Obviously, in the first edition of a conference the percentage of newcomers is 100%.It is interesting to see in Figure 4c that in every edition of the TOPint group there are, at least, 63% of newcomers.
This indicates that the number of returning authors is small, as we could observe previously in Sections "Analysis of Publications" and "Analysis of Authors".There are different conjectures for this fact.One of them is that there are few people who dedicate most of their time to research ubiquitous computing issues.Most of the authors might have their main research in other related areas, such as Artificial Intelligence, and occasionally work with some ubiquitous computing problems.

Representativeness of Authors and Institutions
Figures 5a and 5b depict the occurrence of authors and institutions in the TOPint group, respectively.In this analysis authors and institutions are counted just once by paper.Gregory D. Abowd is the author who published the largest number of papers (Figure 5a), and Universities of California and Washington, and Intel are the most productive institutions (Figure 5b).In SBCUP, both the South region (UFRGS, UCPel, Unisinos, UFSM) and North region (UFAM, and FUCAPI) have a strong participation.
Figure 6 displays the institutions according to their occurrence in papers (institutions are counted just once in each paper), considering Ubicomp, Pervasive and Percom individually

Collaboration Network
A natural process when conducting research is to form groups and establish partnerships (e.g., by affinity) in order to improve a work or to share experiences.In this regard, we identified the main institutions that collaborated with each other in the TOPint group.Ubicomp has 40% of papers with collaboration (more than one institution), Pervasive 38% and Percom 32%.
In the rest of this section we represent the scientific collaboration network of the TOPint group as a graph.In this network, nodes represent authors, while an edge between two nodes indicates a joint publication between authors.In Section "Connected Components" we discuss the connected components, and in Section "Communities Visualization" we identify some communities of authors.

Connected Components
Figure 7a presents the evolution in the number of connected components over the years.In the first editions of Ubicomp, Pervasive, and Percom, the network had 21, 19, and 46 connected components, respectively.After five editions, Ubicomp, Pervasive, and Percom had 80, 71, and 144 authors, i.e., an increase of 280%, 273%, and 213%, respectively.In 2011, Ubicomp, Pervasive, and Percom had 128, 89, 178 components, i.e., an increase of 60%, 25%, 23%, respectively.In this way, the number of components had a significantly greater increase in the first years.Furthermore, the number of components in Percom is much higher than in Ubicomp and Pervasive.
Figure 7b shows the evolution of the biggest connected component (BCC), and the second biggest connected component (SBCC).In the second half of that period, BCC starts

Communities Visualization
One of the most relevant characteristics of graphs representing real systems is the structure of communities.For such identification, we grouped the final network (total) of the TOPint group and applied the k-clique community algorithm, according to Palla et al., 2005.A community is defined as the union of all clicks of size k that can be reached through adjacent k-cliques (two k-cliques are adjacent if they share k nodes).
Figure 8 depicts communities with members connected to universities.If an author was affiliated to more than one institution throughout his/her history, we considered all recorded affiliations.We found 66 communities (using k=3).Figure 8 shows only the four biggest communities (due to space limitations) with 29, 16, 14 and 12 authors.Figure 8a (biggest community) shows that authors from the University of Washington, and Intel are very popular in this community.

Recent Ubicomp Research -Taxonomy
In order to provide a snapshot of the research being conducted in the field of ubicomp, we did a careful reading of the abstract, introduction, conclusion, and, in some cases, other sections as well, of papers published in the TOPint group editions of both 2010 and 2011.This study gave us an overview of the present research in the area, leading to our proposed taxonomy.We do not intend to say that we offer a complete view of recent research, but we do believe it is a good approximation about recent research topics.This taxonomy is to increase much more than SBCC, for both Ubicomp and Pervasive.This does not happen for Percom, where BCC does not increase significantly in any edition.Authors who were supposed to increase BCC may end up increasing the number of new components (Figure 7a).very useful for two reasons.First, it can guide novices in the field.Second, it can be used to compare and discuss the ubicomp research performed in Brazil.
We identified seven categories (see Figure 9), named: Context-aware; HCI Evaluation; Invisible Computing; Innovative Technology; Resource Limitation; Privacy/Security; and Software Design.Sometimes a paper can belong to more than one category, but we consider its dominant entry.Categories have subcategories.They were created when we identified at least three papers related to a "specialization" of the category.This is a way to highlight topics that are receiving considerable attention, which could indicate a research trend or a hot topic.Figure 9 also shows the number of papers found in each category.For instance, category Invisible Computing has 23 papers (15 of them were classified in a subcategory, which means that the remaining eight papers were not sub-categorized).
In the next sections we describe all categories of the taxonomy.In order to increase the readability, we did not cite all papers analyzed.

Context-Aware
Context is defined by Dey and Abowd (2000) as any information that can be used to characterize the situation of an entity.An entity can be a person, place, or object that is considered relevant to the interaction between a user and an application, including the user and applications themselves.The category named "Context-aware" groups studies that use context to automatically adapt a service behavior according to the current context, such as environmental conditions or location.This is a very active and important field in pervasive computing with a considerable number of studies.
There are papers related to context information modeling (Chen, 2011) that propose a hybrid approach to resolve context inconsistencies.The authors Degeler and Lazovik (2011) introduced Context Consistency Diagrams, a data structure for reasoning about situations with incomplete knowledge of the environment and context conflicts that cannot be unambiguously resolved.
Context-aware applications use sensing and inference to attempt to determine users' contexts and to take appropriate actions.However, they are prone to uncertainty, and this may compromise the trust users have in them (Lim and Dey, 2011).We also found studies that are focused on providing intelligibility about how context-aware applications work to improve users' impressions of them.The authors Lim and Dey (2011) investigated the impact of intelligibility on understanding and user impression of applications with varying certainty.Lim and Dey (2010) proposed a toolkit to obtain eight types of explanations from the most popular decision models of context-aware applications.
Some papers were classified in two subcategories: A-Recognition: This category groups studies that derive high level context information from raw data (provided by sensors).This action is called recognition, interpretation or inference.We classified many studies in this category, including proposals that focus on emotional/ psychological state recognition.Healey et al. (2010) performed a study on emotion recognition.Rachuri et al. (2010) presented a mobile sensing platform capable of recognizing some user's psychological states, for social psychology studies based on mobile phones.There are also studies related to voice recognition.Amft (2010) proposed an unsupervised speaker identification system for personal annotations of conversations and meetings.Studies related to situation recognition such as Lovett et al. (2010) proposed a real-time data fusion method that combines the calendar with location and social network data to improve the representation of reality, which is not given by the calendar alone.Another example is the work of Zhuang et al. (2011).Many other papers focus on recognition.Steinhoff and Schiele (2010) showed the promise of dead reckoning using solely a device placed freely in a trousers' front pocket.Larson et al. (2011) presented an algorithm to detect coughs from the audio stream of a mobile phone.Some of the studies of this category were classified in a specific subcategory: A.1-Activity: By "activity" we mean physical human activity.Activity recognition is receiving considerable attention by researchers.(2011) presented a service that allows videos to be saved in user-specified real-world locations, and then serendipitously discover the location of a story as people approach them.Some papers were classified in two other subcategories: B.1-Tracking: Tracking of mobile entities (living being or objects) is a fundamental building block for location-based services.For instance, to deliver a warning message to a car about an accident on the road, the system needs to keep track of the cars on that road to check which car should receive the message.This category is composed of proposals that focus mainly on algorithms or mechanisms to track mobile entities.Becker et al. (2011) showed how to use hand-off patterns from cellular phone networks to identify which routes people take through a city.Weinschrott et al. (2011) introduced mechanisms using mobile devices such as smartphones to track mobile objects.The main contribution of Scellato et al. (2011) was a method to perform location prediction based on nonlinear time series analysis of the arrival and residence times of users in relevant places.Higuchi et al. (2011) proposed a cooperative algorithm to localize mobile nodes.
B.2-Navigation: This subcategory is composed of studies that focus on methods and techniques to help user's navigation.Pielot and Boll (2010) investigated the concept of tactile waypoint navigation to overcome existing limitations in commercial pedestrian navigation systems.Minamimoto et al. (2010) proposed a local map generation algorithm that could be used in the recognition of an accident site in an emergency situation.Hoseinitabatabaei et al. (2011) presented an approach to determine a user's facing direction with a mobile phone, regardless of the mobile phone's orientation.

HCI Evaluation
This category of studies presents investigations of the use of new or existing ubicomp technologies to help the design and deployment of future ubicomp applications, devices or systems.This is a relevant and very active field in the field of ubicomp, as we can see from the amount of related papers.Dey et al. (2011) presented an investigation about the users' proximity to their mobile phones.Boesen et al. (2010) evaluated the ways in which Location-Based Services are changing domestic relationships.Truong et al. (2010) evaluated a proposed task-centered battery interface (TCBI) for mobile devices.The goal of TCBI is to help users to better understand how different applications affect battery usage.Dillahunt et al. (2010) proposed a set of open research questions and design recommendations for technologies that may affect and be affected by the conflict between stakeholders around energy use.Betz (2010) studied how maintenance workers deal with repair-related information, how this information is related to objects and places and to which extent those objects can be seen as boundary-objects for the involved groups of actors.MacMillan et al. (2010) performed a large scale trial of a game and also presented some evaluation techniques to be applied in that case.
We categorized some papers in three subcategories: A-Behavioral Change: This subcategory is composed of studies that propose and/or evaluate mechanisms to provide user behavior change (e.g., make healthier meal choices).Rogers et al. (2010) presented an evaluation of the potential of embedded displays (designed to give information in subtle and playful ways) to nudge people to change their behavior.Grimes et al. (2010) presented a field study evaluation of a mobile game in which players learn how to make healthier meal choices.In the line of e-health, Oliveira et al. (2010) presented a mobile social game that engages users to be more compliant in taking their daily medication.
B-Design Recommendation: This subcategory is composed of proposals that focus on providing guidance and recommendations on designing or deploying pervasive technologies.Poole et al. (2011) provided guidance for introducing ubiquitous computing technologies in institutions with established norms and rules.Clinch et al. (2011) reflected on how a digital signage system is used in detail (content and use related practices) and provided a set of lessons for designers of future ubicomp public display systems.
C-Novel Interactions: This subcategory groups research projects that propose or evaluate new types of interaction (e.g., touch interaction).Related to touch interaction, Bonner et al. (2010) presented an eyes-free gesture-based text entry system for multi-touch devices.Related to haptic interaction, Linden et al. (2011) designed an immersive theater experience to raise awareness and question perceptions of 'blindness'.A multimodal experience was created, comprising ambient sounds and narratives and an assortment of themed tactile objects, intended to be felt.They also designed a novel haptic device to enhance their discovery of a pitch-black space.As an example of a new proposal, Bulling and Roggen (2011) investigated the feasibility of using eye movement analysis to recognize visual memory recall processes of people looking at familiar pictures or unfamiliar pictures.They also introduced cognition-awareness and cognitive context as new paradigms in ubiquitous computing.Some studies were classified in two other subcategories: C.1-Interactive Surfaces: This subcategory groups studies related to interactive surfaces (tangible user interfaces).They merge input and output into a single space, enabling designers to go beyond simple touch interactions, to integrate physical objects as interactive tools on the surface.Costanza and Shelley (2010) evaluated the impact of Tangible User Interfaces on users' everyday environments through two tangible interfaces.Hincapié- Ramos et al. (2011) explored a n alternative for integrating everyday objects and tabletops called mediated tabletop interaction.Åkerman et al. (2010) also proposed a solution that belongs in this category.
C.2-Collaborative: This subcategory is composed of papers that enable interaction of people to create and share context on pervasive devices.Willis and Poupyrev (2010) presented a system that allows people to create and share drawings directly from a regular mobile phone.Perring et al. (2010) evaluated a collaborative workspace table prototype.

Invisible Computing
Traditional computers consume much of our attention and separate us from what is happening around us.Weiser believed that in the ubicomp world computation could be integrated with common objects used for everyday work practices, rather than forcing computation to be a separate activity (Krumm, 2009).Invisible computing is essential to the ubicomp visions and is related to technologies that require minimal interaction from users.This section is composed of studies dedicated to this challenge.
We classified 23 papers in this category.Bidot (2011), for example, presented an approach to aggregate and use devices that support the everyday life of human users in ambient intelligence environments.Some of them were classified in a specific subcategory, named: Behavioral Patterns.
A-Behavioral patterns: To achieve invisible computing, a fundamental step is the understanding of people and system behavioral patterns.For example, location patterns can be used to automatically program home thermostats.The papers in this subcategory focus mainly on user behavioral pattern identification, modeling, and prediction.Lathia and Capra (2011) demonstrated how automated fare collection records reveal the hidden behaviors of individuals and their behaviors responses to travel incentives.Zhang et al. (2011) aimed to discover anomalous driving patterns from taxis' GPS traces, targeting applications such as the automatic detection of taxi driving frauds or road network change.Madan et al. (2010) used mobile phones to identify behavior changes reflected in mobile phone sensors, when individuals suffer from common colds, influenza, fever, stress and mild depression.Choujaa (2010) performed a study about the prediction of mobile phone users' activity.

Innovative Technology
We would like to emphasize that every work listed in this taxonomy presents innovation in its field.The category "Innovative Technology" is composed of studies that focus mainly on original devices or visionary conceptual ideas for ubiquitous computing.All the studies identified in this category (22 papers) were categorized into two groups: A-New Devices: This group presents the design of novel devices.Usually the new device is presented through a prototype.Wei et al. (2011) presented a dining table embedded with interactive subsystems that augment and transport the experience of communal family dining to create a sense of coexistence among remote family members.Cohn et al. (2010) presented a sensing solution for automatically identifying gas use.Cheng et al. (2010) described a new on-body capacitive sensing approach to derive activity-related information.
B-New Ideas: This category is composed of papers that present new ideas that do not demand the design of new devices.In other words, they rely on existing devices (e.g., popular devices) to offer new services.For example, the study by Ho et al. (2011) proposed a thermal-based power meter system that uses thermal imaging to track disaggregated appliance usage.Gupta et al. (2011) described a sensing approach that turns ordinary compact fluorescent light bulbs into sensors of human proximity.Du et al. (2011) utilized consumer depth cameras (Kinect style) to allow nonexpert users to scan their personal spaces into 3D models.Kuznetsov et al. (2011) argued for expanding the current landscape of sensing to include living organisms, such as plants and animals, along with traditional tools and digital devices.

Resource Limitation
This category is composed of studies that focus mainly on common resource limitations present in many ubicomp devices, like smartphones.Katmon and Ryan (2011) proposed a replication scheme that dynamically creates and releases replicas on client nodes during runtime, with the goal of improving response time performance.Qin et al. (2010) proposed a VANET-WSN system to overcome the inherent limitations of a pure VANET-based system.
This category is further classified in three subcategories: A-Energy: The studies in this subcategory present new techniques to improve energy savings and analyses of energy-related issues.Musolesi et al. (2010) presented the design, implementation and evaluation of several techniques to optimize the information uploading process for continuous sensing on mobile phones.Pyles et al. (2011) exploited, during runtime, the silence periods of a VoIP call to save smart phone Wi-Fi energy.Ramos et al. (2011) proposed a Low Energy Assisted Positioning solution that can save up to 80% of GPS energy consumption in typical trajectory-based service scenarios.Xu et al. (2010) presented an energy-efficient RFID inventory algorithm.Rice and Hay (2010) examined the power consumption of two handsets and showed in detail the energy used by particular actions.
B-Connectivity: In certain situations, the service may lack continuous network connectivity, such as in a mobile environment.This group of studies deals mainly with this issue.Yang and Wu (2011) studied the problem of intermittent connected passive RFID networks.Razafindralamboo et al. (2010) presented a distributed algorithm for adaptive infrastructure deployment in two-tiered intermittent connected networks.And finally Pitkanen and Karkkainen (2010) explored mobile web access using asynchronous messaging via WLAN hotspots.
C-Sensing: A given event may be more difficult to observe by using just one device.On the other hand, when other people collaborate with their devices (e.g., sensor-enabled cell phones) this can be simpler.In this category there are studies related to participatory sensing (Burke et al., 2006;Silva et al., 2012), but this same category could be appropriated to group studies related to opportunistic sensing (Campbell et al., 2006).Reddy et al. (2010a) and Lee (2010) explored user participation incentive in participatory sensing applications.Reddy et al. (2010b) discussed how to develop a recruitment framework to enable organizers to identify well-suited participants for data collections, based on geographic and temporal availability and behavior.

Privacy/Security
Today there are many possibilities to create, store and transmit several kinds of sensed data, including GPS location, videos, and photos.This brings new types of services and also many challenges related to the privacy and security of those personal data.This category (12 papers) is classified in two subcategories: A-Sensing/sharing data: The sensing of personal data is possible through the use, for example, of smartphones or smart environments, like a smart house (e.g. the tracking of energy consumption by a smart grid system).Increasing the amount of data stored and/or transmitted through the Internet, or another network, increases the need for securing the privacy of this data (Stajano, 2004).While people may benefit from services that make use of sensed data (e.g., location), it is important to consider the privacy risks of such services.This category is composed of studies that are mainly focused on those issues.Wiese et al. (2011)  B-Attacks/Defenses: This category is composed of studies that are more concerned with proposing or discussing methods for attacking and defending personal data.Kostakos et al. (2011) aimed to answer two questions related to targeted location-sharing privacy attacks.Peddinti and Saxena (2011) examined the level of privacy protection provided by the current query obfuscation techniques against adversarial location service providers.Finally, Saxena et al. (2011) presented a novel approach for authenticating multiple RFID tags.

Software Design
This category is composed of studies that focus on pervasive software design (e.g., approaches and models).Harrington and Cahill Silva, Celes, Mota and Loureiro | A Picture of Present Ubicomp Research Exploring Publications from Important Events in the Field (2011) presented a model-driven approach to apply planning and optimization algorithms in pervasive computing applications.All other studies of this category were classified in two subcategories: A-Middleware: This subcategory is composed of studies that are focused on providing and discussing a middleware or framework to ease the development of pervasive systems.Ding et al. (2010) proposed a framework for adaptive software that supports the online reconfiguration of each concern in the adaptation loop.Chowdhury et al. (2010) presented a middleware for supporting energy-efficient, long-term remote health monitoring.Kang et al. (2010) presented an active resource orchestration framework for mobile context monitoring.
B-Systems/Apps: This subcategory contains studies concerned mainly with current pervasive applications or systems.Pham and Paluska (2010) proposed a tool that takes a step towards easing the development of ubicomp systems by reducing several debugging pain points.Lu et al. (2010) presented a mobile multiparty voice communication system that allows drivers to socialize with each other while driving.Nachman et al. (2010) proposed a system to manage diabetes that considers energy expenditure monitoring, diet-logging, and analysis of health data.Luyten (2010) presented an approach and tool that makes it possible to create an intuitive user interface for exploring and controlling the environment.

Ubicomp Research in Brazil
In this section we present the classification of studies published in SBCUP between 2010 and 2011 (Figure 10) using the proposed tax-onomy that was presented in details in previous sections.As shown in Figure 10, the most representative category is software design (52% of all papers), and the second one is context-aware (19% of all papers).In the classification performed for international studies, i.e., papers from abroad presented at SBCUP, the most representative category is context-aware (35% of all papers), and the second one is HCI evaluation (20% of all papers).Notice that we have not been able to classify studies in the HCI evaluation category considering papers from SBCUP.Another relevant observation is that in some categories, for instance invisible computing, we did not find enough papers to create subcategories that were previously identified.We believe that these observations may be interesting for two reasons.First, there are many opportunities for Brazilian researchers acting in related fields, such as HCI or security, to start addressing new issues brought by ubicomp, for instance development and evaluation of new forms of user interaction.Second, they may guide newcomers to start his/her research in less explored areas.

Conclusions
This work presented an overview and analysis of recent ubicomp research.Based on datasets of top conferences in ubicomp we showed relevant information, such as representativeness of authors and institutions and communities formed by authors.In the analysis of all papers published in top conferences in the ubiquitous computing area, we proposed a taxonomy of recent ubicomp research.Besides that, we also discussed the Brazilian ubicomp research based on the proposed taxonomy.We believe that this work can be useful to guide newcomers to this field, as well as for experienced researchers as it provides facts to enable discussion about the field.
In terms of future work, it is interesting to keep analyzing recent papers in the field to maintain the taxonomy updated.For this purpose the use of crowdsourcing tools may be a good option.
(a) Publications per year.(b) Authors per paper.(c) Publications per author.

Figure 3 .
Figure 3. Intersection of authors in the TOPint group.
Albinali et al. (2010) presented a technique to improve upon state-of-the-art energy expenditure estimation in physical activities.In Förster et al. (2010) the authors proposed and evaluated a dynamic hand gesture recognition system, based on wearable motion sensors.Maekawa et al. (2010) showed how to recognize activities of daily living (ADLs) with a designed sensor device, which is equipped with heterogeneous sensors such as camera, microphone, and accelerometer attached to a user's wrist.Helaoui et al. (2011) proposed a method for recognizing interleaved and concurrent activities.B-Location: Location is powerful context information, typically available in mobile devices with an embedded GPS.Location enables a wide range of new services and poses many research issues.Yang et al. (2010) provided a study of the accuracy of two existing methods for cell tower localization using wardriving data.Based on a cell-phone trace analysis,Calabrese et al. (2010) showed that the origins of people attending an event are strongly correlated to the type of event.Kranz and Schmidt (2010) discussed the Digital Enhanced Cordless Telecommunications and its potential for positioning and localization.Bentley et al.

Figure 9 :
Figure 9: Taxonomy of recent research on ubicomp -Top international conferences.
explored which features of interpersonal relationships influence sharing.Toch et al. (2010) presented a study about location sharing privacy preferences of users.De Cristofaro et al. (2011) presented an architecture geared to privacy-sensitive applications where personal information is shared among users.Brush et al. (2010) explored people's concerns about privacy related to personal data provided from location-based services.