2010 CENS Annual Research Review
October 20, 2010
Poster Abstract Listing

A Course on Sensor Networks (PASI) at the La Selva Towers MRI Installation, CENS 01
Michael Allen, Eric Graham, Thomas Harmon, Erin Riordan, Philip Rundel

A workshop, titled "Expanding the Frontier in Tropical Ecology through Embedded Sensors", was sponsored by a Pan-American Advanced Studies Institute (PASI) grant from the National Science Foundation. In this two-week training course, 15 scientists lead 30 ecologists in exploring the new applications of instrumentation and technologies that are available for applications in tropical ecology, and discussed how the field can be reshaped as ecologists uncover new aspects of tropical forests. The Institute took place at the La Selva Biological Station in Costa Rica, where the Organization for Tropical Studies has established, in collaboration with the Center for Embedded Networked Sensing (CENS), a Rainforest Ecological Research Portal supported by a cluster of instrumented towers and a canopy walkway outfitted with the necessary cyber-infrastructure for demonstrations and field projects.

Design, Production, Deployment and Data Collection of the First Production Version Automated MiniRhizotrons (AMR), CENS 02
Michael Allen, Michael Taggart, Kuni Kitajima, Thomas Unwin, Rebecca Hernandez

The production version of the Automated Mini-Rhizotron (AMR) is the culmination of five-years work to develop the next generation imaging system for soil ecosystem study. It combines not only a high precision microscope and motor control system, but also a sophisticated, web-based, software component for visualizing the resulting images. This presentation describes some of the challenges associated with the design of both HW and SW systems as well presents some of the new data we were able to capture using them. The challenges we had to overcome were daunting. With images measuring 3mm X 2.25mm, we wanted to build a product with great repeatability and reasonable depth of field at
100X magnification on a shoestring budget. The other half of the problem was that unlike existing, lower resolution, manually operated systems, we had to keep track of orders of magnitude more images, and develop a software system that could position motors within a few microns and allow a researcher to easily see and manipulate the results generated. When construction of the prototype was completed initial sample soil images were collected as part of testing and refining the unit for production. Unfortunately, from an engineering standpoint, the images were so good that the prototype was pressed into use almost immediately, which meant long periods in the field with short trips to the lab for
repair work and debugging. During this period, many improvements were made in the mechanical design and the RootView software continued to evolve to better meet researcher needs. Before bringing the production units online earlier this year, the prototype unit collected over .5M images. With the addition of five new production units at the James Reserve and the La Selva Biological field station, we believe that over the past three years, the AMR has now collected more images than all manual systems combined. The AMR has allowed unprecedented detail in soil ecology imaging, including capturing time series photos of previously unobserved phenomena. Combining this with the webbased interface, RootView, it is changing how research is done and bringing into question some of the models used to study these systems.

Evaluating Freeway Underpasses as Wildlife Corridor Linkages, CENS 03
Michelle Murphy and Michael Allen

Free from anthropogenic disturbance, natural landscapes consist of a mosaic of interconnected habitats. A consequence of human development is often the loss of this original connectivity. With the expansion of urbanization, agriculture and alternative energy resource development, desert environments are becoming fragmented at an increasing rate, making knowledge of the impacts on wildlife in these areas especially important. Highway underpasses may function as wildlife corridors, providing both a potential constraint as well as means for wildlife to make safe crossings between suitable habitats in areas where man-made barriers, such as railways and high-speed highways, may be impeding wildlife
movement. However, there are few studies demonstrating the degree to which such corridor structures are actually used by wildlife in desert environments. Six pre-existing freeway underpass structures, located along the western portion of the I-10 freeway in the Coachella Valley, Riverside County, California, will be evaluated using camera traps for one year to determine whether they function as a linkage in facilitating wildlife movement between the Peninsular and Transverse Mountain Ranges. Additional objectives are to determine whether variables such as the size and surrounding disturbance levels of these structures influence wildlife use, and whether underpasses may be functioning as prey traps for predatory species. Results of this study will be used as pilot data for planning and implementing a more expansive study to identify generalized principles for effective corridor design in arid environments.

From Landscape to Leaf: digital cameras zooming in on plant phenology, CENS 04
Eric Graham, Erin Riordan, Eric Yuen

Plant phenology, the timing of periodic phases of plant development (bud-burst, flowering, leaf senescence), is one of the most responsive and easily observed traits in nature impacted by climate change. Cameras placed in different ecosystems allow us to scale our observations spatially, observing how different areas respond to climate and climate change. Not only green vegetation, but flowers and senescing leaves can be readily detected. We are working on methods to take advantage of the pan-tilt-zoom capabilities of the internet-connected cameras available to us to better describe plant phenological events and correlate them with ecosystem drivers.

High Frequency Monitoring of Near-shore Marine Ecosystem Yields Insight into Phytoplankton Bloom Dynamics, CENS 05
Beth Stauffer, Alyssa Gellene, Carl Oberg, Diane Rico, Gaurav Sukhatme, and David Caron

Harmful microalgal blooms pose significant problems to coastal communities throughout the United States and worldwide. While the effects of blooms are often highly visible, the underlying causes of such blooms remain poorly understood. Coastal ocean observing allows for better documentation of environmental conditions throughout the timecourse of blooms, but the analyses of such datasets have relied primarily on approaches that are inappropriate for such time-dependent, non-stationary natural phenomena. King Harbor is a near-shore marine system in the greater Los Angeles region with a history of red tides that has been outfitted with sensors since 2008. Using time series analysis methods, including wavelet approaches, data from King Harbor was analyzed for meaningful, temporally-localized relationships between physical forces and more biologically-relevant constituents (chlorophyll, dissolved oxygen, turbidity).

Measuring and Modeling Soil Salinity under Flood-Irrigation with Dairy Wastewater, CENS 06
Patrick Barnes, Heidi Dietrich, Alexander Rat'ko, Christopher Butler, Sandra Villamizar-Amaya, Henry Pai, Thomas Harmon

Accumulation of salts in soils and groundwater is a pervasive issue in agricultural soils in semi-arid environments. The objective of this study was to evaluate the effectiveness of current sensor technologies and modeling techniques in assessing water and solute fluxes in dairy farms irrigated with wastewater. Environmental sensors measuring soil moisture, temperature and salinity were deployed at the field site over one growing season. The recorded data were then used to inverse model the hydraulic and solute transport parameters of the field site using HYDRUS 1-D. The model was shown to successfully describe field measurements of the water content (R^2 = 0.78) and overall salinity (R^2 = 0.75) at the dairy site. Long-term projections were also possible using historical climate data that showed accumulation and movement of solutes in the soil profile geometry. Additional work can be done with higher resolution and longer term data sets to observe variation in the optimized parameters. Moreover, while the model can effectively describe conditions in the root zone, coupling these efforts with groundwater information will be important in evaluating the risk of salt accumulation in groundwater resources.

Spatial River Survey System (SpaRSS): an efficient protocol for mapping water quality and riparian zone features, CENS 07
Henry Pai, Christopher Butler, Sandra Villamizar Amaya, Patrick Barnes, Thomas Harmon

Stream flow and water quality often vary with seasonal conditions and demand fluctuations. Typically, monitoring stations are sparsely situated in space, but capture relatively high resolution temporal data (e.g, hourly) to gauge flow and key water quality parameters such as temperature and salinity. Along the Merced River, spatiotemporally distributed fluxes from groundwater (GW) and drainage and diversion canals remain ungauged or difficult to quantify, rendering accurate forecasting water quality conditions challenging. This work demonstrates the spatial river survey system (SpaRSS), presenting a set of synoptic water quality traces collected on the Merced River. SpaRSS synchronizes a high frequency GPS coordinates on a kayak equipped with multiparameter water quality sensors. High-density spatial data collection under different flow and seasonal conditions can help identify key river segments for with respect to landscape-river interactions.

Temporal Variation of Stream Metabolism in Response to Disturbance within a Managed River System, CENS 08
Sandra Villamizar, Henry Pai, Christopher Butler, Patrick Barnes, Thomas Harmon

Metabolism estimates (gross primary production, GPP and community respiration, CR) obtained through the continuous monitoring of physicochemical properties in managed rivers may be used to evaluate the effects of various disturbances on ecosystem function. This work highlights the development of a GPP/CR observational network on the human-dominated Lower Merced River, currently the southern-most extent of Chinook salmon habitat in the Central Valley of California. Our investigations include spatial (both longitudinal and transverse gradients) and temporal (daily, seasonal and interannual) variation of these metabolism estimates as we are interested in relating responses of this type of lotic system to disturbances such as short- or long-term reservoir operational changes for drought management, flood control, fish habitat enhancement, or alleviation of salinity and nutrient discharges due to land management practices. The observational network will be described in terms of: (1) design and installation of a reproducible infrastructure of GPP/CR monitoring stations, (2) analysis aimed at linking the spatio-temporal metabolic trends to natural factors such as the seasonal radiation availability or nutrient input from leaf decay, and (3) separating natural effects from the ones triggered by human disturbances in order to better inform water resources management decisions. Observations over the 2009-10 water year, demonstrate that the Lower Merced River behaves as a heterotrophic system, with large temporal changes in metabolism clearly observable by the monitoring network. For example, the GPP/CR ratio decreased from 0.6 to 0.2 as a consequence of a large flow disturbance associated with short-term reservoir releases mandated biannually to support salmon migration. This and other examples set at different temporal and spatial scales will be presented and discussed in terms of management implications.

Testing of Nitrate and Ammonium Sensors in a Saturated Soil Column, CENS 09
Joseph Ferrer, Sarika Doshi, Jessy Avelar and Jose A. Saez

Testing of ion-selective nitrate and ammonium sensors were performed in the laboratory using a saturated soil column. The column was packed with soil from an agricultural site in Palmdale, which uses a center pivot system to irrigate fodder crops with recycled water. Management of nitrogen is essential to protect the underlying groundwater at the site. The ammonium and nitrate sensors were calibrated before and after the experiment by immersing the sensors in concentration standards. A peristaltic pump and reservoir system were used to convey deionized water through the column at steady flow rates. Instantaneous spikes of nitrate and ammonium (5 to 50 mg-N) were injected just upstream of the column at various times to observe the sensors’ readings over a period of approximately two weeks. The column was mostly tested under saturated and steady flow conditions, but unsaturated conditions were also explored by stopping the flow and observing the sensors’ responses in the drying soil column. Sensor data from the experiments were analyzed to determine mass recovery, modal time and other relevant transport parameters that helped assess the sensors’ responses. The nitrate sensors responded relatively well to the spikes, as had been the case in previous efforts. The ammonium sensors were less responsive, which may be due to rapid sorption to the soil matrix. Refinements in experimental design, new sensors and the use of deionized water improved results over previous efforts, but challenges remain in sensor reliability due to the challenging field conditions encountered in the subsurface environment.

GeoNet: a platform for rapid distributed geophysical sensing, CENS 10
Igor Stubailo, Dustin McIntire, Martin Lukac, Paul Davis, William Kaiser, John Wallace, Deborah Estrin

In the GeoNet experiment, the science objective is to use a rapidly installable wirelessly linked seismic network to make near-real time unaliased observations in aftershock or volcanic zones. To accomplish this, we collaborate with Reftek to construct a new generation digital acquisition system (DAS) based on the UCLA-developed LEAP (low-power energy aware processing) system and a newly developed low-power A/D converter from Texas Instruments. The instrument will have two parts: DAS and a seismic sensor. The DAS would have a solar panel attached on top, battery inside (with external power plug), internal GPS antenna (with a possibility of attaching an external one), external N-type onnector for an antenna. Field installation would involve attaching the box to a post and bury the seismometer. It would then become a node in a wireless network of neighbors, e.g. along a dirt road that could bring event data out in real time, or, in the low power mode, on a duty cycle, e.g., 5 min every hour. The radios would also deliver network time, a backup to GPS. In a building, the instrument would not need the solar panels, although after an earthquake power may be unavailable. An instrument plus Episensor (or other) would be installed on various floors. Radio connectivity through the floors would be used to provide network time and transport event data. Where available (e.g., near windows) GPS time would calibrate the network time. With no solar energy available, the battery with an active sensor would last 4 days. Longer deployments would require a larger external battery. At the moment, we have two prototypes that have been tested in short-time field deployments in Palmdale and near the Salton Sea.

PERUSE: Peru seismic experiment locating earthquakes and seismic tomography, CENS 11
Emily Foote, Luis Dominguez, Igor Stubailo, Steven Skinner, Kristin Phillips, Paul Cox, Richard Guy, Victor Aguilar, Hernando Tavera, Audin Lurence, Paul Davis, Deborah Estrin, Martin Lukac, Robert Clayton

This work describes preliminary tomography results from the Peru Seismic Experiment (PERUSE) a 100 station broadband seismic network installed in Peru. The network consists of a linear array of broadband seismic stations that was installed mid-2008 that ran from the Peruvian coast near Mollendo to Lake Titicaca. A second line was added in late 2009 between Lake Titicaca and Cusco. Teleseismic and local earthquake travel time residuals are being combined in the tomographic inversions. The crust under the Andes is found to be 70-80 km thick decreasing to 30 km near the coast. The morphology of the Moho is consistent with the receiver function images (Phillips et al., Fall AGU 2010) and also gravity. Ray tracing through the heterogeneous structure is used to locate earthquakes. However the rapid spatial variation in crustal thickness, possibly some of the most rapid in the world, generates shadow zones when using conventional ray tracing for the tomography. We use asymptotic ray theory that approximates effects from finite frequency kernels to model diffracted waves in these regions. The observation of thickened crust suggests that models that attribute the recent acceleration of the Altiplano uplift to crustal delamination are less likely than those that attribute it to crustal compression.

Metadata Tensions: library principles vs. everyday scientific data practices, CENS 12
Matthew Mayernik, Jillian Wallis, Christine Borgman

Data sharing requirements and mandates are becoming more common, and many institutions are investigating data curation methods. Metadata are a critical component to any institutional scientific data curation initiative. At CENS, we have identified a number of metadata challenges related to data management and use, specifically: 1) the ambiguous responsibility for metadata creation between information professionals, working scientists, and hardware/software tools, 2) the tension between the highly principled library metadata approach and the ad hoc everyday practices of working researchers, 3) the ways that metadata creation and knowledge are distributed socially in research settings, and 4) the role of metadata at different stages of the data life cycle. In this poster, we describe a study of the metadata practices of CENS researchers and illustrate how these challenges impact our work in developing a registry of CENS data.

Who is Responsible for CENS data?, CENS 13
Jillian Wallis, Christine Borgman, Matthew Mayernik

The curation of data is predicated on the availability of data. If data are available then someone has made them available. This person is presumably responsible for the data. Responsibility is a term we do not use lightly, and implies everything from authorship to long-term maintenance. We have assumed that researchers are willing to be responsible for their data, but it is unclear whether we can make this assumption. As part of on-going data practices research at CENS we have collected two rounds of interviews, including questions about data ‘responsibility’. To get at the underlying meaning, or talk around the notions of responsibility without just asking who is responsible for data, we broke this down into a series of questions that captured different aspects of responsibility, including authorship, ownership, and a colloquial definition of responsibility. During both rounds we saw a variety of interpretations of terminology and an even wider variety of responses. This poster will present some of this variety and the implications for data curation.

Outreach Programs: ELAC Summer Environmental Academy & Future Tech Now! , CENS 14
Wes Uehara, Karen Kim, Melissa Burt, Diana Dalbotten, Barbara Gibson, Eric Graham, Vanessa Green, Marco Molinaro, Alisa Lee, Keith Oden, Sigolene Ortega, Armando Rivera-Figueroa, Marina Rueda, Ruby Vargas-Lainez, April Wilkinson

As part of the Center outreach efforts CENS has partnered with East Los Angeles College (ELAC) to host a Summer Science Academy designed to pipeline underrepresented youth from lower socio-economic areas of East Los Angeles toward higher-education in STEM. Participants engaged in a What’s Invasive citizen science campaign exposing them to CENS mobile phone technology, data collection and visualization and environmental awareness and stewardship. Students spent a second day with UCLA’s Center for Excellence in Engineering & Diversity (CEED) visiting labs and hearing from current UCLA students who successfully transferred from ELAC. In addition to our local outreach efforts CENS also partnered with the Society for Advancement of Chicanos and Native Americans in Science (SACNAS), the National Science Foundation and six other Science and Technology Centers (STCs) to expose a national audience of participants in the 2010 SACNAS annual meetings to hands-on demonstrations and “science-play” of technology and research conducted at our Centers. CENS highlighted our newly launched Android based What’s Invasive citizen scientist campaign. Participants were able to make simulated
observations on site and learn about access to existing campaigns as well as gaining access to the application and setting up parks in their local area.

Summer@CENS Evaluation, CENS 15
Tiffani Riggers and Cynthia Milstein

CENS High School and Undergraduate Scholars Program directly involves high school and undergraduate students in CENS research through a comprehensive summer internship experience. The program is the core of CENS educational pipeline and is an excellent example of aligned Center research and education activities. The program brings together talented undergraduates from around the country and local high school students to engage in Center research for 8-10 weeks over the summer. This poster highlights the summative and formative evaluation findings from the 2010 summer and also shows demographics and a brief yearly history of the program.

ENGAGE: UCLA residence plaza electricity monitoring experiment, CENS 16
Victor Chen, Neil Lessem, Yeung Lam, Jackson Ding, Manda Paul, Stoytcho Stoytchev, Joe Tsai, Robert Gilbert, Magali Delmas, William Kaiser

Electricity generation accounts for over 40 percent of the carbon dioxide emitted by the United States, with residential and commercial buildings collectively accounting for over two-thirds of electricity usage. This is not surprising considering that people in the United States spend more than 90 percent of their lives in buildings. This means that we, as individuals, are responsible for a significant percentage of our greenhouse gas emissions as a result of our everyday actions. Our hypothesis is that with higher frequency, higher granularity information that is presented in the proper context, individuals will be better equipped to make decisions about their electricity usage behavior that can lead to reductions. Various incentives can also play a large role in encouraging people to modify their behaviors. In the residence plazas on the UCLA campus, students do not pay electricity bills so they do not have a built-in financial incentive to conserve electricity. This resulting analysis can focus on the isolated effects of private and private information, competition, and other incentives independent of financial states.

Green Edge Networks, CENS 17
Shuai Hao, Nilesh Mishra, Ramesh Govindan

The carbon footprint attributable to the edge of the Internet (laptops, desktops, notebooks, and handhelds) will, within a decade, likely surpass that attributable to the Internet infrastructure. The energy consumed by devices at the edge of the Internet will add considerable load on the energy production and distribution network. An approach to tackling this looming problem is make these devices energy-optimal and energy-proportional: energy optimality requires the devices to use the right amount of energy needed to perform the job, while energy proportionality requires the device to use energy proportional to the utilisation of the whole device or of its sub-systems. The goal of our work, Green Edge
Networks, is to come up with policies and mechanisms which enable edge-devices to be energy-optimal and energyproportional. In this proposal, we explore methods for coordinated adaptation to the energy bottleneck in the processing chain in order to achieve energy-optimality and energy-proportionality. In order to achieve coordinated adaptation, we envision an architecture consisting of two components. First, we need an energy profiling subsystem which can provide accurate energy consumption of each component with contextual information about the application on whose behalf the component is being run. Second, we need control algorithms that achieve coordinated energy management in the presence of diverse workloads.

Indoor Energy and Occupancy Monitoring, CENS 18
Han Zhao, Chenni Qian, Yan Wang, Younghun Kim, Jonathan Friedman, Mani Srivastava

Imminent energy resource crisis and ubiquitous energy waste lead to higher demands for people to utilize resources sensibly and efficiently. To monitor indoor resource consumption and archive personal consumption profile, and thus give user feedback to aware them of energy conservation, we design a three-level indoor energy and occupancy monitoring hierarchy in this poster, as described below. 1. Hardware We design and implement Qin Node, a low power non-intrusive indoor energy consumption monitoring platform that allows indirect sensing of power, water, and light usage. (1) Sensors Qin Node employs ADXL330 accelerometer, HMC1001 magnetic sensor and light sensor to nonintrusively sense water, electricity and light, respectively. The underlying mechanism is that (a) appliance power consumption is related to the magnetic field it generated around its wire, we use this electromagnetic characteristic to indicate power consumption; (2) the amount of water flow in a water pipe is related to the pipe vibration, thus we use this feature to indicate water flow from acceleration of the water pipe. (2) MCU and Radio The microcontroller and radio we use are based on UC Berkeley’s EPIC Core. It has a MSP430 microprocessor and CC2420 radio chip. (3) Features Qin node can be initialized and reset by simply bringing a magnet close to the reed switch. It avoids physical contact with the board, as opposed to other traditional sensor board that use buttons to reset the circuit, since it goes in a sealed case. A highly sensitive vibra-tab is used to trigger the whole circuit. The board is powered by two AAA batteries. Preliminary testing has shown that Qin Node achieves desired low power and functionality. 2. Energy Monitoring (1) Common electrical equipments. We build a fine-grained power consumption profile for networked equipments in our lab, including printer and firewall server. It turns out that their power consuming behaviors are significantly different from each other, though they are shared by each one in the lab. Specifically, printer consumes much more energy in working condition than that in idle state, which means there are non-periodic pulses in its energy profile. On the contrary, to keep connectivity anytime, firewall server has to be running all time, which cause their nearly constant power consumption. (2) Individual power consumption The purpose of this is to provide a detailed dashboard showing energy consumed by everyone in our lab. It is an individual level profile of power consumption. The real- time results will be displayed online so that everyone is able to access to it easily, making them aware how much energy they are consuming. 3. Occupancy Monitoring. We have employed door sensors mounted on the front and back door side frames to detect door open events. Whenever there is a door open event captured by one of the door sensors, the corresponding sonar sensor supervising the same door will be triggered, measuring the distance of the closest obstacle to itself. In this context, the sonar sensor will indirectly measure the height of the person who is entering or leaving the lab. In out experiment, we have 11 subjects involved in the height measurement. They participate in the experiment with natural walking state. By collecting the readings of the sonar sensors in a great amount of events, we can get a set of training data whose input is a specific subject’s height and whose output is the subject’s identity. By using all the training data, we can build the normal distribution model for every participant’s measured height. Then when we get a random height value, we can input it into every subject’s mathematical model and assign this value to a single individual after comparing every output. In order to improve the accuracy of the height-based identification, we also resort to side information like network traffic in our experiment. We assume that when someone is inside the lab, there will be a busy status in that person’s network traffic. So when our system decides someone is entering lab, we expect an increase in that subject’s network packages after a short timestamp. To sum up, our research is based on personal power and occupancy
profile to make people aware of and further, improve energy efficiency. In the future, there are several things needed to be investigated deeply. First of all, we plan to replace our current commercial power meter with Qin Nodes that have wireless communication to ease our implementation. Besides, we try to develop our web server in a further step, in order to make our webpage more-friendly to end users. Finally, we will explore the relation of multiple data streams collected by various sensors from Qin Node and occupancy analysis. Hopefully, we will combine and analyze as much useful information as possible to build a energy-aware system.

CENS 19 (currently unavailable)


Energy Expenditure of Treadmill Walking using On-Body Accelerometers and Gyroscopes, CENS 20
Harshvardhan Vathsangam, Adar Emken, E Schroeder, Donna Spruijt-Metz, Gaurav Sukhatme

Walking is the most common activity among people who are physically active. Standard practice physical activity characterization from body-mounted inertial sensors uses accelerometer-generated counts. There are two problems with this - imprecision (due to usage of proprietary counts) and incompleteness (due to incomplete description of motion). We address both these problems by directly predicting energy expenditure during steady-state treadmill walking from a hip-mounted inertial sensor comprised of a tri-axial accelerometer and a tri-axial gyroscope. We use Bayesian Linear Regression to predict energy expenditure based on modeling joint probabilities of streaming data. The prediction is significantly better with data from a 6 axis sensor as compared with streaming data from only 2 linear accelerations as is common in current practice. We also show how counts from a commercially available accelerometer can be reproduced from raw streaming acceleration data (up to a linear transformation) with high correlation (.9787 +/- .0089 for the Xaxis and .9141 +/-.0460 for the Y-axis acceleration streams). The paper emphasizes the role of probabilistic techniques in conjunction with joint modeling of tri-axial accelerations and rotational rates to improve energy expenditure prediction for steady-state treadmill walking.

Physiological Data Collection on Mobile Devices, CENS 21
Yi Han, Kasturi Rangan Raghavan, Zainul Charbiwala, Haksoo Choi, Mani Srivastava

With the availability of higher performance computing in low power regimes, it is imperative to draw the line between local and remote processing very carefully. We characterize and study the impact of implementing fast algorithms for ECG analysis on the sensor platform. We use a custom board designed around off-the-shelf modules to demonstrate our experiments and analysis. In particular, we show that it is possible to extract QRS, the RR interval from ECG data on the sensor platform itself. We show that computing the Lomb Periodogram is feasible. The power and computation time for each algorithm and the cost of communication is characterized.

SensorSafe: privacy-preserving sharing of sensory information for medical studies and healthcare, CENS 22
Haksoo Choi, Max Greenblatt, Zainul Charbiwala, Supriyo Chakraborty, Mani Srivastava

In medical studies and mobile healthcare services, sharing of personal sensory information is inevitable. While we need to share such information, it is important to preserve privacy of individuals. The key challenge is balancing individual privacy and information utility. That is achieving a certain utility of the information while restricting or hiding sensitive information from the personal data. We propose a privacy-preserving data sharing architecture to solve the problem. Our architecture provides several mechanisms: fine-grained access control, remote data stores, data obfuscation, and utility assessment. With our privacy mechanisms, individual can obtain control over who can access what information, choose where their data are physically stored, and restrict inferences that can be drawn using their data. In addition, using our utility assessment tool, medical study organizers can find study subjects who provide desired quality of information.

Exploiting AML Algorithm for Multiple Acoustic Source 2D and 3D Estimations, CENS 23
Juo-Yu Lee, Ralph E. Hudson, Kung Yao

Direction-of-arrival (DOA) estimation in array signal processing has been central to a multitude of areas such as acoustic tracking, sonar, radar and mobile communication systems. Over the past few decades, a wide variety of high resolution algorithms have been proposed for source DOA estimation. Among the DOA estimation techniques using sensor arrays, the maximum-likelihood (ML) estimator has been shown to have superior performance under many challenging environments. For example, an ML estimator can be tailored to account for the large bandwidth exhibited by wideband signals. We derive the optimal parametric ML solution to estimate DOA of wideband sources in the far field. The
wideband data is transformed to the frequency domain, and the signal spectrum can be represented by the narrowband model for each frequency bin. This allows a direct optimization for the source DOA under the assumption of i.i.d. noise instead of the two-step optimization that involves relative time-delay estimation. We apply the DFT to data of finite length and derive the solution, which we refer to as the approximated ML (AML) solution. Later, we will demonstrate through simulations the AML algorithm solves the DOA estimation problem for multiple acoustic sources. In practice, the number of sources must be determined independent of any DOA estimation algorithm, but here, we assume it is known for the purpose of this paper. For the single-source case, one can show the AML formulation is equivalent to maximizing the sum of the weighted cross correlation functions between time shifted sensor data. The optimization using all sensor pairs mitigates the ambiguity problem that often arises in the relative time-delay estimation between two widely separated sensors for the two step LS methods. In the case of multiple sources, we apply an efficient alternating projection procedure, which avoids the multidimensional search by sequentially estimating the location of one source while fixing the estimates of other source locations from the previous iteration. Besides the development of the AML
method, we also derive the theoretical Cram´er-Rao bound (CRB) for both performance comparison and basic understanding purposes. The CRB shows that the DOA variance bound can be broken down into two separate parts: one that depends on the signal characteristics and one that depends on the array geometry. The signal dependent part shows that theoretically, the source DOA RMS error is linearly proportional to the noise level and the speed of propagation and inversely proportional to the source spectrum and frequency. Thus, better source DOA estimates can be obtained for high-frequency signals than low-frequency signals.

Imagers as Sensors: building categories from video streams, CENS 24
Teresa Ko, Eric Graham, Angelo Cenedese, Stefano Soatto, Deborah Estrin

We explore automated object detection and categorization in image sequences within the context of natural environments. Inherent in these environments are significant challenges to be modeled -- for example, complex texture, background motion, and object mimicry. We present a general background model that is applicable to natural scenes. Our approach models the underlying warping of pixel locations arising from background motion. The background is modeled as a set of warping layers where, at any given time, different layers may be visible due to the motion of an occluding layer. Foreground regions are thus defined as those that cannot be modeled by some composition of some
warping of these background layers. We show how changes in intensity/color histograms of pixel neighborhoods can be used to discriminate foreground and background regions. We find that this approach compares favorably with the state of the art, while requiring less computation. We have designed and implemented a system for cataloging putative objects of interest into viewable clusters from an image sequence and user input. We represent an object as a barcode that indicates whether or not a feature is present across all views. The approach is unbiased towards redundant views -- that is, it does not matter how many times an object appears from the same viewpoint. At the same time, the approach does not penalize for missing views -- so that successful object categorization does not require capturing all viewpoints. We use this representation to cluster objects into viewable clusters that users can label according to the categories of their interest. We then feed these labels back into the system to automatically label new objects that appear in the image sequence. We find that the system significantly reduces the amount of time users would spend looking at uninformative images.

Detection and Localization of Fires in Videos, CENS 25
Avinash Ravichandran & Stefano Soatto

Indoor and outdoor fires have the potential to cause vast amounts of damage to both property and lives. While indoor fires can be monitored using existing fire and smoke detectors, such methods cannot be employed outdoors. In this work, we address the issue of detecting fires from video sequences with applications to both indoor and outdoor fire detection. Our algorithm explicitly models the temporal evolution of such video sequences using linear dynamical systems and classifies them with concepts from object categorization for detecting and localizing fires. We believe by employing such an algorithm to videos, we can not only detect the onset of fires, but also provide first responders with additional information such as the location of the fire and possibly, the type of fire. This would enable us to enhance existing detection capabilities indoors and provide a method to monitor outdoor locations with the ability of fire detection.

Decision Support System for Large Scale Oceanographic Experiments, CENS 26
Jnaneshwar Das, Hörður Heiðarsson, Arvind Pereira, Beth Stauffer, Ryan Smith, David Caron, Burt Jones, Gaurav Sukhatme

Growing complexity of multi-disciplinary experiments requires integration of situational awareness, data synthesis, and planning. To address this issue, the Monterey Bay Aquarium Research Institute (MBARI) has proposed a Decision Support System (DSS) that will bring together existing MBARI technology within a framework that is current and scalable for technology transfer. It is a new effort for bringing together separate elements within MBARI, while also leveraging contributions from participating researchers outside MBARI. This work addresses two of the three pieces of such a system: situational awareness and planning. In the summer of 2010, a prototype decision support system was
developed with the intention of requirements gathering and trial usage during an oncoming field experiment. The prototype system is now live and is being used in the ongoing October 2010 BIOSPACE experiment that brings together multiple institutions and universities. The DSS prototype is being used alongside an ONR funded tool developed at MBARI called Collaborative Science (CoSci). This poster showcases the DSS prototype and describes how it is being used for this large-scale field experiment.

Lab-on-Chip Aquatic Microorganism Analysis System, CENS 27
Wendian Shi, Han-Chieh (Jay) Chang, Mike Liu, Leyla Sabet, Beth Stauffer, Astrid Schnetzer, David Caron, Chih-Ming Ho, Yu-Chong Tai

This is a project that aims to expedite research in marine biology using chip-based and state-of-the-art detection technology. The project is a joint effort that will incorporate the expertise of three different groups, Dr. Chih-Ming Ho at UCLA, Dr. David Caron at USC and Dr. Yu-Chong Tai at Caltech. One main focus of the project is to develop labon-a-chip devices that reduce total sample volume and detection time. Also, the chips can be fabricated in large quantities with minimal cost so many experiments can be run in parallel. Here at Caltech, a portable cytometer for counting and identifying different types of algae and a chip to culture a small number of algae and screen for factors inducing toxin production will be developed. The cytometer can allow researchers to quickly know the algae concentration to monitor algae growth. Algal bloom and toxins produced by different algae have always caused problems to the environment and marine ecology. Pseudo-nitzschia is one type of algae that produces a neural toxin called domoic acid. However, during Pseudo-nitzschia bloom, domoic acid is not always produced. Studies done by other groups have suggested that many factors (such as trace metal, macronutrient, or ionic concentration) might induce
or suppress algae to produce toxin. Yet, exact causes are unclear. To completely elucidate the causes of toxin production, many potential compounds will have to be screened. This leads to an enormous amount of experiments to be performed and large quantity of reagents and cells to be used. To speed up the process of screening for possible factors inducing toxin production, we would like to make a chip to culture Pseudo-nitzschia under different growing conditions. At the same time, an Ultra Sensitive Electrochemical Sensor will be developed for detection of domoic acid at Dr. Chih-Ming Ho’s lab at UCLA. The current state-of-the-art detection technology indicates that per cell toxin load may range over 2 or 3 orders of magnitude but its sensitivity is limited since a sample size of at least 100 cells/mL is required. The new
sensor will be able to push the sensitivity to 10 cells/mL or to even single molecules of domoic acid. This sensor will not only enable the detection of domoic acid produced by algae cells inside the culture chip, such sensor will also have the broad application of detecting domoic acid from field samples.

Obstacle Detection and Avoidance for an Autonomous Surface Vehicle using a Profiling Sonar, CENS 28
Hordur Heidarsson and Gaurav Sukhatme

We present an experimental study of a mechanically scanned profiling sonar for Autonomous Surface Vehicle (ASV) obstacle detection and avoidance. We extract potential obstacles from echo returns and suggest a scanning strategy for sonar in this application. We demonstrate with simulations (driven by data collected in the field) the potential for an ASV to rely solely on sonar data to navigate and avoid obstacles in a lake and harbor environment.

Path Planning for Environmental Monitoring with AUVs, CENS 29
Jonathan Binney and Garuav Sukhatme

Autonomous underwater vehicles (AUVs) are widely used to monitor and study environmental phenomena. Because of the difficulty in communicating with underwater vehicles, improving their ability to operate autonomously has the potential to greatly increase their usefulness. Better autonomy can also make it easier for small groups of people to control large numbers of AUVs at a time. One of the key capabilities needed for robot autonomy is that of path planning; in this work, we give an overview of the path planning research we have done for environmental monitoring using AUVs.

Persistent Ocean Monitoring with Underwater Gliders: towards accurate reconstruction of dynamic ocean processes, CENS 30
Ryan Smith, Mac Schwager, Stephen Smith, Daniela Rus, Gaurav Sukhatme

This poster presents a path planning algorithm and a velocity control algorithm for underwater gliders to persistently monitor a patch of ocean. The algorithms address a pressing need among ocean scientists to collect high-value data for studying ocean events of scientific and environmental interest, such as the occurrence of harmful algal blooms. The path planner optimizes a cost function that blends two competing factors: it maximizes the information value of the path, while minimizing the deviation from the path due to ocean currents. The speed control algorithm then optimizes the speed along the planned path so that higher resolution samples are collected in areas of higher information value. The resulting paths are closed circuits that can be repeatedly traversed to collect long term ocean data in dynamic environments. The algorithms were tested during sea trials on an underwater glider operating off the coast of southern California over the course of several weeks. The results show significant improvements in data resolution and path reliability compared to a sampling path that is typically used in the region.

Toward Risk Aware Mission Planning for Autonomous Underwater Vehicles, CENS 31
Arvind de Menezes Pereira, Burton Jones, Ryan Smith, Jonathan Binney, Hordur Heidarsson, Matthew Ragan, Geoff Hollinger, David Caron, Gaurav Sukhatme

Given historical Automated Identification System (AIS) data for several months, we propose a mission planner for AUVs that builds risk maps for this data and utilizes it to select waypoints for the AUVs which put it below a certain risk threshold while also picking points which reduce the cost of communication while traversing between a specified start and end location. Our plan also takes into account bathymetric data to plan for paths that avoid the coast and other features such as kelp forests. Plans are found by sampling the search space and waypoints picked are evaluated to minimize the risk of collision with ships, the coastline as well as to ensure the risk of collision is kept low despite deadreckoning
errors which might force the AUV into a risky area. The planner selects the best path as the one with the minimum cost of communication among all the feasible paths that have risk lower than a user-specified amount. The poster describes two proposed techniques to solve this problem. One uses Probabilistic Roadmaps followed by a search for the path between the start and end locations with the lowest communication cost. The other algorithm is based on Anytime RRTs and quickly finds a sub-optimal solution satisfying the risk constraints, but correspondingly looks for better solutions providing lower communication costs.

Macro-Programming Support for Optimization in Wireless Sensor-Actuator Networks, CENS 32
Rahul Balani and Mani Srivastava

Wireless sensor/actuator nodes collect an enormous amount of data over space and time to estimate state of the environment and control actuators to modify it. In numerous applications such as intelligent light control, agricultural irrigation, target tracking etc., these estimation and control problems can be expressed as optimization of a convex cost function involving data from all the sensor nodes. Consequently, selecting an optimization algorithm, as well as its implementation, is closely tied to network characteristics of the deployment and application constraints. Thus, programming these networks is not only hard due to complex inter-node coordination and resource management, but the resulting code cannot be reused across different deployments for the same application. The aim of this research is to develop a programming framework that automatically finds the most efficient implementation, either centralized or distributed, of an optimization algorithm for a given problem, deployment characteristics and application constraints. This is popularly known as Macroprogramming. MacroLab [Hnat et al., ACM Sensys 2008] is one such framework for WSNs that hides complexity of sensor network programming from the users and enables code reuse through automatic deployment-specific decomposition of programs written using Matlab-like operations. However, we believe that using
MacroLab's naive low-level decomposition and cost analysis can result in inefficient implementations of optimization algorithms. In this project, we extend MacroLab to support a popular iterative optimization algorithm called Subgradient Descent that allows both centralized and distributed implementations. The user encodes the cost function and subgradient descent algorithm in MacroLab for the entire network. The extended compiler utilizes static and runtime cost analysis to automatically decompose this program and produce a centralized or distributed implementation for a specific deployment and application constraints. For instance, a centralized implementation is chosen for a star topology,
or a distributed implementation with sequential iterations is selected for a ring topology. However, for ad-hoc mesh networks, the extended framework exploits domain-specific knowledge - that actuators have limited but overlapping regions of influence - encoded in the program to automatically distribute and parallelize its iterations across the sensor nodes. This poster currently focuses on the theory behind parallel subgradient iterations and shows some preliminary simulation results to demonstrate improvements from application of physical properties of actuators to the intelligent light control problem.

Activity Classification using Accelerometers, CENS 33
Jay Chien, Greg Pottie, William Kaiser, Natali Ruchansky, Claire Lochner, Elizabeth Do, Tremaine Rawls

In this project, we design a physical activity classification system using a body sensor network (BSN) consisting of costsensitive tri-axial accelerometers. We focus on two tasks: general everyday activities (walking, running, sitting, etc.), and workspace activities (different motions and sitting postures). We use a naive Bayes classifier, and show that we can use a systematic way to train the system. For each task, we find a set of features that separate the corresponding activities. We obtained a classification accuracy of 84% for the general activities task, and a classification accuracy of 94% for the workspace activities task.

Energy Delay Tradeoffs in Smartphone Applications, CENS 34
Moo-Ryong Ra, Jeongyeup Paek, Abhishek Sharma, Ramesh Govindan, Martin Krieger, Michael Neely

Many applications are enabled by the ability to capture videos on a smartphone and to have these videos uploaded to an Internetconnected server. This capability requires the transfer of large volumes of data from the phone to the infrastructure. Smartphones have multiple wireless interfaces – 3G/EDGE and WiFi – for data transfer, but there is considerable variability in the availability and achievable data transfer rate for these networks. Moreover, the energy costs for transmitting a given amount of data on these wireless interfaces can differ by an order of magnitude. On the other hand, many of these applications are often naturally delay-tolerant, so that it is possible to delay data transfers until a
lower-energy WiFi connection becomes available. In this paper, we present a principled approach for designing an optimal online algorithm for this energy-delay tradeoff using the Lyapunov optimization framework. Our algorithm, called SALSA, can automatically adapt to channel conditions and requires only local information to decide whether and when to defer a transmission. We evaluate SALSA using real world traces as well as experiments using a prototype implementation on a modern smartphone. Our results show that SALSA can be tuned to achieve a broad spectrum of energy-delay tradeoffs, is closer to an empirically-determined optimal than any of the alternatives we compare it to, and,
can save 10-40% of battery capacity for some workloads.

RAPS: rate adaptive positioning systems for energy efficient localization on smartphones, CENS 35
Jeongyeup Paek, Joongheon Kim, Ramesh Govindan

Many emerging smartphone applications require position information to provide location-based or context-aware services. In these applications, GPS is often preferred over its alternatives such as GSM/WiFi based positioning systems because it is known to be more accurate. However, GPS is extremely power hungry. Hence a common approach is to periodically duty-cycle GPS. However, GPS duty-cycling trades-off positioning accuracy for lower energy. A key requirement for such applications, then, is a positioning system that provides accurate position information while spending minimal energy. In this paper, we present RAPS, rate-adaptive positioning system for smartphone applications.
It is based on the observation that GPS is generally less accurate in urban areas, so it suffices to turn on GPS only as often as necessary to achieve this accuracy. RAPS uses a collection of techniques to cleverly determine when to turn on GPS. It uses the location-time history of the user to estimate user velocity and adaptively turn on GPS only if the estimated uncertainty in position exceeds the accuracy threshold. It also efficiently estimates user movement using a duty-cycled accelerometer, and utilizes Bluetooth communication to reduce position uncertainty among neighboring devices. Finally, it employs cell tower-RSS blacklisting to detect GPS unavailability (e.g., indoors) and avoid turning on
GPS in these cases. We evaluate RAPS through real-world experiments using a prototype implementation on a modern smartphone and show that it can increase phone lifetimes by more than a factor of 3.8 over an approach where GPS is always on.

Feedback System for Mobile Participatory Sensing, CENS 36
John Jenkins, Jinha Kang, Deborah Estrin, Eric Graham

We have designed a system for participatory sensing to aggregate data and generate visualizations for that data. It will connect to several existing campaign servers as an initial pilot. It will use this data to generate rankings among other users if applicable and comparing users against their previous recordings. It will deliver this information to end users as a web object that can be rendered initially on mobile devices.

Map based Compression for Location Traces, CENS 37
Zainul Charbiwala, Magahet Mendiola, Younghun Kim, Nabil Hajj Chehade, Deborah Estrin, Mani Srivastava

Maps have become a commodity on mobile devices, as have GPS receivers. While many applications employ GPS for acquiring localization data, a handful pay attention to its sampling rate. An unnecessarily high rate not only means higher energy consumption for GPS receivers, but also leads to reduced efficiency in downstream handling of location data. In this paper, we present a compression algorithm for pre-acquired location traces and argue that a key component of realizing this compression is the use of available map information. We describe a system that exploits this street-level knowledge to implement a lossy compression algorithm for GPS trajectories. The algorithm is online, allowing streaming
compression, and users can configure the level of temporal, spatial or speed fidelity needed for their application. We illustrate the fidelity vs. compression performance of the algorithm on a diverse set of real-world location traces collected over a two year span.

MobiProg: an adaptive programming system for cloud-enabled smartphone applications, CENS 38
Luis Pedrosa, Nilesh Mishra, Nupur Kothari, Ramesh Govindan, Michael Gray, Dero Gharibian, Taehee Lee, Todd Millstein, Jeff Vaughan

With the advent of smartphones as an emerging class of personal Internet capable devices, mobile applications, or apps, are rapidly becoming the corner stone of what defines the user's experience. While many of these apps will perform trivial or otherwise meaningless tasks, a new class of cloud-enabled applications is becoming more popular. These apps augment the smartphone's capabilities, allowing them to leverage the large datasets and the computational power that can be harnessed in large-scale computing infrastructures known as server clouds. Many of these cloud-enabled mobile applications delegate all or most of the business logic to the cloud, reducing the smartphone to a thin client. However,
there is a vast universe of unexplored possibilities where the smartphone collaborates with the cloud to enhance the user's experience. The MobiProg project aims to create a new framework for the development of new cloud-enabled smartphone applications.

Privacy and Data Control: building the personal data vault, CENS 39
Min Mun, Katie Shilton, Shuai Hao, Nilesh Mishra, Jeffrey Burke, Deborah Estrin, Ramesh Govindan, Mark Hansen, Jerry Kang

Participatory sensing data streams may be used for self-reflection and improvement, examining the environment, or as a tool for clinicians to judge the efficacy of their treatment. But current architectures give participants very little control over how and when to share this personal information. We are working on a new privacy architecture for participatory sensing designed to give users control over their raw data, provide high-level tools and guidance and encourage continued user engagement.

Participation & Innovation: values in participatory sensing, CENS 40
Katie Shilton, Jeffrey Burke, Deborah Estrin, Mark Hansen, Jim Waldo

Participatory sensing raises diverse ethical challenges. What values are important in design of participatory sensing systems? How does participatory sensing differ from surveillance? And can participatory sensing be empowering? This poster explores how ethics affect new technology development and innovation in the CENS participatory sensing laboratory.

Sensing Everyday Places and Paths using Less Energy, CENS 41
Donnie Kim, Younghun Kim, Deborah Estrin, Mani Srivastava

Continuously understanding a user’s location context in colloquial terms and the paths that connect the locations unlocks many opportunities for emerging applications. While extensive research effort has been made on efficiently tracking a user’s raw coordinates, few attempts have been made to efficiently provide everyday contextual information about these locations as places and paths. We introduce SensLoc, a practical location service to provide such contextual information, abstracting location as place visits and path travels from sensor signals. SensLoc comprises of a robust place detection algorithm, a sensitive movement detector, and an on-demand path tracker. Based on a user’s mobility,
SensLoc proactively controls active cycle of a GPS receiver, a Wi-Fi scanner, and an accelerometer. Pilot studies show that SensLoc can correctly detect 94% of the place visits, track 95% of the total travel distance, and still only consume 13% of energy than algorithms that periodically collect coordinates to provide the same information.

Understanding Smartphone Usage, CENS 42
Hossein Falaki, Ratul Mahajan, Srikanth Kandula, Dimitrios Lymberopoulos, Ramesh Govindan, Deborah Estrin

Smartphones are being adopted at a phenomenal pace but little is known publicly about how people use these devices. To understand smartphone usage we conducted a comprehensive study using detailed traces from 255 users. We characterized intentional user activities - interactions with the device and the applications used - and the impact of those activities on network and energy usage. In this poster I will present our findings regarding the immense diversity among users. Along all aspects that we studied, users differ by one or more orders of magnitude. This level of diversity suggests that mechanisms to improve user experience or energy consumption will be more effective if they learn and adapt to user behavior. We found that qualitative similarities exist among users that facilitate the task of learning user behavior. I
will present models that can be used to explain users interactions with their smartphones. For instance, the relative application popularity can be modeled using an exponential distribution, with different distribution parameters for different users.

Using Semi-Supervised Learning in Self-Improving Activity Classifiers, CENS 43
Brent Longstaff, Sasank Reddy, Deborah Estrin

Mobile phones’ increasing ubiquity has created many opportunities for personal context sensing. Personal activity is an important part of a user’s context, and automatically recognizing it is vital for health and fitness monitoring applications. Recording a stream of activity data enables monitoring patients with chronic conditions affecting ambulation and motion, as well as those undergoing rehabilitation treatments. Modern mobile phones are powerful enough to perform activity classification in real time, but they typically use a static classifier that is trained in advance or require the user to manually add training data after the application is on his/her device. Our work investigates ways of automatically
augmenting activity classifiers after they are deployed in an application. It compares active learning and three different semi-supervised learning methods, self-learning, En-Co-Training, and democratic co-learning, to determine which show promise for this purpose. The results show that active learning, En-Co-Training, and democratic co-learning perform well when the initial classifier’s accuracy is low (75-80%). When the initial accuracy is already high (90%), these methods are no longer effective, but they do not hurt the accuracy either. Overall, active learning gave the highest improvement, but democratic co-learning was almost as good and does not require user interaction. Thus, democratic co-learning
would be the best choice for most applications, since it would significantly increase the accuracy for initial classifiers that performed poorly.

AndWellness: an open mobile system for activity and experience sampling, CENS 44
Nithya Ramanathan, John Hicks, Donnie Kim, Josh Selsky, Mohammad Monibi, Deborah Estrin

We are extending the current capacity of Experience Sampling by developing and validating three innovative functions on smartphones: 1) Automated activity trace: Provides continuous, unobtrusive logging of a participant's activity (still, run, walk, bike, drive) computed using location and acceleration traces. 2) Contextual probes: Reminds participants to record an assessment based on time, location, or other contextual information. 3) Real-time feedback: Helps track and encourage progress towards behavior change goals. These functions aim to improve self-monitoring health behaviors by increasing participant compliance, cost-effectiveness, reliability, and validity. We are building the AndWellness system based on these three self-monitoring functions for young overweight mothers to monitor diet, exercise, and stress. In
three focus groups, 24 ethnically diverse, new mothers aged 18-35 identified the optimal design strategy for software that is attractive, accessible, and acceptable. Preliminary qualitative analysis of their feedback suggests that our new system could be a powerful tool for behavior change by: (1) Encouraging 'Observation in the moment', (2) Prioritizing accountability (over accuracy), and (3) Setting goals. Based on mothers' preferences, we have adjusted our system priorities: Since the measures are approximate, validity and reliability tests, while important, have become secondary to keeping people engaged in data capture. We will evaluate different engagement mechanisms in a study of 60 young
mothers self-monitoring their diet, stress, and exercise. We are extending the software platform to support three additional studies with diverse patient cohorts (100 middle aged cancer survivors; 30 South Asian immigrants; 30 HIV+ persons) to self-monitor a broader range of behaviors (diet, exercise, sleep, mood, medication adherence, sexual behaviors, and drug use.)

Cell phones as a Distributed Platform for Black Carbon Data Collection, CENS 45
Nithya Ramanathan, Martin Lukac, Muvva Ramana, Praveen Siva, Tanveer Ahmed, Abhishek Kar, Ibrahim Rehman, Veerabhadran Ramanathan, Deborah Estrin

Black carbon (BC), the visible component of soot that gives emissions such as diesel engine exhaust their dark color, has come to be recognized as a major contributor to global warming, and a frontline concern for climate change strategies (Ramanathan 2001, Jacobson 2010). We have developed a new low-cost instrument for gathering and measuring atmospheric BC concentrations that leverages cell phones to transmit data from an air filtration unit to a centralized database for analysis. Our new system relies on image processing techniques, as opposed to other more expensive optical methods, to interpret images of filters captured with a cell phone camera. As a result, the entire system costs less than $500 (and is orders of magnitude cheaper than an Aethalometer, the prevailing method for measuring atmospheric BC). We are working with community organizations in California and India who are interested in measuring black carbon in their specific communities. We will be recruiting participants in each area to maintain and record data from the sensors each day. We are working with The Energy Resources Institute, an international NGO based in India, to deploy this instrument with 60 people in conjunction with Project Surya, which aims to deploy clean cook-stoves and rigorously evaluate their impact on BC emissions. Field tests of this new instrument performed in California and India report an average error of less than 10% when compared with an Aethelometer and a thermal optical method for analysis. These excellent results hold the promise of making large-scale data collection of BC feasible and relatively easy to reproduce. The use of cell phones for data collection permits monitoring of BC to occur on a greater, more comprehensive scale not previously possible, and serves as a means of instituting more precise, variation-sensitive evaluations of emissions. By storing the data in a publicly available repository, our system will provide real-time access to mass-scale BC measurements to researchers and the public.

Scheduled Task Manager, CENS 46
Faisal Alquaddoomi, Sunil Menon, Martin Lukac, Nithya Ramanathan, Deborah Estrin

There has been significant effort invested by the CENS community as well as researchers at large in developing applications that use a combination of web and smartphone-based applications for performing data collection and analysis. While smartphones are the logical choice for most data collection tasks, especially automated data collection, there exists a problem of availability -- not everyone has access to smartphones, or even to web applications. This problem is especially apparent in disadvantaged communities, where resources and technical savvy are scarce. Fortunately, devices that support at least simple SMS messaging have reached sufficient penetration to allow these
communities to participate in data collection. That said, applications designed for SMS platforms must be developed under different assumptions than the typical smartphone/web application combination. There is an even greater emphasis in this area on keeping the interaction focused and functional, since texting requires greater effort on the part of the user. All value that the system provides to the end-user must be representable as text, and in a fairly small payload as well. The Scheduled Task Manager was conceived with just such concerns in mind. The population that we intend to serve (young adult survivors of childhood cancers) spans a broad range of socioeconomic circumstances; we cannot rely on our patients having access to smartphones or internet-enabled computers. Furthermore, the system must provide value to the patient that outweighs the difficulty of using the system. The end goal is to keep these individuals aware and involved in their continuing healthcare. To serve this goal, we intend to provide useful information about their healthcare as entered by clinicians overseeing the project, which will include appointment reminders and other such clinician-patient communication. Ultimately, we would like to support patient-driven data collection (i.e. "selfreporting"), which we hope will give patients a greater sense of agency in their care. The system itself is a study in designing a method by which user interactions can be compactly represented. We opted for a state machine model, in which each task is represented as a machine. The states represent the current state of the interaction with the user, and the transitions are possible expected user inputs. As the user interacts with the machine, their input drives it from state to state, which produces messages that are sent back to the client as well as various bits of information (appointment dates, feedback from appointments, etc.) which are reviewable by clinicians via the web interface. In the future, we hope to allow end-user authoring of state machines, which at the moment is a somewhat technical task.

Environmental Participatory Sensing: What's Invasive and BudBurst Mobile, CENS 47
Jameel Al-Aziz, Eric Graham, Kyung Han, Jinha Kang, Eric Yuen

CENS is developing mobile phone and web-based tools for formal and informal observation of ecosystems. We are collaborating with national environmental education campaigns, such as Project BudBurst, and with the National Park Service to increase participation in citizen scientist campaigns and to support park service personnel in day to day data gathering. Our experience with volunteers at UCLA and at the National Park Service has demonstrated that mobile phones are an efficient, effective and engaging method for collecting environmental and location data and hold great potential for both raising public awareness of environmental issues and collecting data that is valuable for both ecosystem management and research.

MOBILIZE: bringing participatory sensing into the Los Angeles Unified School District, CENS 48
Jean Ryoo, Deborah Estrin, Mark Hansen, Jane Margolis, Thomas Philip, Jody Priselac, Todd Ullah

Despite the fact that today's teenagers use mobile phones and computer technology on a daily basis for communicating, social networking, shopping, and more, high school students do not get to use such technology in meaningful ways at school. Furthermore, only the most privileged students receive quality science, technology, engineering, and math (STEM) education such that the majority of public high school students--especially students of color--do not have opportunities to learn the important computational and critical thinking skills necessary to be our future STEM leaders. In an effort to address issues in equity and access in public high schools by linking students' learning of computational
thinking with their technology use, a K-12/University partnership was developed in a project entitled "Mobilize." This partnership--between the UCLA Center for Embedded Networked Sensing (CENS), UCLA Center X in the Graduate School of Education and Information Studies, the Los Angeles Unified School District (LAUSD), and the Computer Science Teachers Association (CSTA)--will bring CENS Participatory Sensing systems (an innovative method of data collection and interpretation in which individuals use mobile phones to systematically collect and analyze data about their home communities) into LAUSD's Exploring Computer Science classrooms in an effort to prepare students to be
engaged, literate, and numerate participants of democracy. Through student-designed research projects using Participatory Sensing, Mobilize builds upon teenagers' fascination, engagement, and involvement with technology, fostering them to be creators instead of simple users of technology. Exploring Computer Science students and teachers of Mobilize will also work in collaboration with Math and Science classrooms in their research projects, therefore developing a multidisciplinary approach to computer science learning. Furthermore, Mobilize will offer an innovative model of professional development for current and future cadres of high school teachers by organizing educators into
multidisciplinary teams and learning communities that engage hands-on, inquiry-based teaching methods.

Participatory Sensing for Community Data Campaigns: Boyle Heights, CENS 49
Amelia Acker, Deborah Estrin, Martin Lukac

We developed a project with Boyle Heights Planning for Place group, using mobile phones to facilitate mapping, recording, and collecting data on circulation of community members to facilitate ‘home-grown’ strategies from a community resident perspective. In this poster we describe a community data campaign that made use of participatory sensing for environmental needs assessment and report on our fieldwork findings.

your.flowingdata – An Application for Personal Data Collection via Twitter, CENS 50
Nathan Yau and Mark Hansen

Millions of people use Twitter to update their friends and family on what they are doing. We add structure to this messaging system to create a basic syntax that lets people collect data about themselves, such as what they eat, where they go, or how many miles they drive. your.flowingdata is an application that lets people collect data via Twitter and then explore it through interactive visualization tools.