Foundation models tested for clinical practice using human-designed metrics may mask fundamental differences in information processing. We investigated this using the clock drawing test (CDT), a cognitive screening tool. Three foundation models achieved 94% accuracy on conventional metrics, matching experts. However, upon decomposing the CDT into 24 questions across five cognitive domains, results diverged significantly. In cases with unanimous model agreement, they still disagreed with human raters in 22% cases. Performance varied drastically with 88% alignment with humans on rule-based executive questions but only 46% on context-dependent anticipatory thinking questions. We observed that models abstained three times more than humans, primarily owing to poor data quality. These findings show standard clinical evaluation metrics fail to capture how foundation models process information. High aggregate accuracy obscures component-level failures. We contribute a systematic evaluation of frontier models’ healthcare capabilities, demonstrate theory-driven task decomposition, and discuss design implications for better human-AI collaborative systems.

Low-frequency antenna arrays enable obstacle-penetrating communications, but their large physical footprint—a 4 × 4 array at 150 MHz requires 16 m2—prevents practical deployment. DeformRF enables portable, deformable arrays that compress to backpack-size yet maintain beamforming performance when deployed on flexible substrates. Our key insight is that data-driven methods can predict complex electromagnetic behavior under deformation without real-time simulations. DeformRF combines: (1) a 260,000-sample synthetic dataset mapping deformations to EM characteristics, (2) physics-inspired ML models achieving >94% prediction accuracy on average, and (3) smartphone-based 3D reconstruction requiring no infrastructure. In real-world experiments, DeformRF maintains beamforming gains within 1 dB of optimal despite severe deformation, while baselines degrade by 4–10 dB. For emergency response scenarios, our 4 × 4 flexible array achieves ±5° direction-finding accuracy when tracking signals through buildings, enabling practical indoor deployment of large low-frequency arrays.

This paper introduces WISP, a printable graphene-based wearable system that enables truly imperceptible micro-gesture recognition with sub-10mm finger movements—up to 4 × smaller than existing approaches. While conventional gesture systems require conspicuous hand motions that disrupt social interactions, WISP detects subtle pinch-like micro-gestures that are virtually invisible to observers yet produce distinct force signatures at the wrist through tendon and skin deformation. Our co-designed hardware-software pipeline leverages Laser-Induced Graphene (LIG) technology to create ultra-thin, skin-conformal strain sensor arrays that can be fabricated on-demand using standard laser cutters and achieves 100μVpp noise floor while consuming only 10mW power. Evaluation with 19 participants demonstrates 99.2% hotword detection, 90.2% user-specific recognition, and 82.9% cross-user generalization. WISP maintains high accuracy across real-world conditions including walking and various clothing scenarios, scales to 12-gesture vocabularies, and extends to context-aware object interactions with 93.8% accuracy.

In this paper, we present VYRE, a ring-based oscillometric wearable designed for low-burden and robust frequent blood pressure monitoring. VYRE revisits the clinically established oscillometric method — widely accepted in arm and wrist form factors because of its high accuracy — and extends it to a compact ring form factor, currently realized as a proof-of-concept prototype. The key innovation is the ability to derive oscillometric signals on the finger by inducing controlled inflation and deflation to capture arterial oscillations in response to circumferential tension. This enables accurate estimation of blood flow dynamics within the digital arteries for blood pressure inference. VYRE leverages a lightweight model to estimate systolic and diastolic pressures from the measured oscillations observed from a vibration sensor integrated into the ring. Compared to PPG-based methods, this approach offers significant improvement in robustness to signal drift, ambient light variations, and skin tone differences. Each measurement requires a user-initiated ∼ 40-second quiet hold, after which the system returns systolic and diastolic readings without per-user calibration. In an IRB-approved study involving 71 participants, VYRE achieves mean absolute errors of 6.36 mmHg for systolic and 4.96 mmHg for diastolic pressure relative to a reference cuff, with biases of − 1.27mmHg and 0.0mmHg, respectively. The standard deviations of 7.87 mmHg (SBP) and 6.22 mmHg (DBP) meet the AAMI requirements (≤ 8mmHg), and 95% limits of agreement fall within [ − 16.69, 14.15] mmHg for systolic and [ − 12.2, 12.19] mmHg for diastolic pressure. Correlation with the reference is strong (Pearson r = 0.75 for SBP, r = 0.67 for DBP), confirming consistent tracking. The system maintained accuracy across finger sizes and hand postures. User feedback indicated that 84% of participants rated the device as comfortable or very comfortable, and 70% expressed willingness to use it daily. These results confirm VYRE’s practicality, accuracy, and potential as a compact, on-demand blood pressure monitoring solution.

Resource-constrained IoT devices increasingly rely on deep learning models, however, these models experience significant accuracy drops due to domain shifts when encountering variations in lighting, weather, and seasonal conditions. While cloud-based retraining can address this issue, many IoT deployments operate with limited connectivity and energy constraints, making traditional fine-tuning approaches impractical. We explore this challenge through the lens of wildlife ecology, where camera traps must maintain accurate species classification across changing seasons, weather, and habitats without reliable connectivity. We introduce WildFiT, an autonomous in-situ adaptation framework that leverages the key insight that background scenes change more frequently than the visual characteristics of monitored species. WildFiT combines background-aware synthesis to generate training samples on-device with drift-aware fine-tuning that triggers model updates only when necessary to conserve resources. Our background-aware synthesis surpasses efficient baselines by 7.3% and diffusion models by 3.0% while being orders of magnitude faster, our drift-aware fine-tuning achieves Pareto optimality with 50% fewer updates and 1.5% higher accuracy, and the end-to-end system outperforms domain adaptation approaches by 20-35% while consuming only 11.2 Wh over 37 days—enabling battery-powered deployment. An open-source implementation of WildFiT is available at https://github.com/mmehdirk/WildFit.

Various wireless technologies have been utilized for sensing. Although promising, these wireless sensing technologies have inherent limitations. Prior research mainly focuses on overcoming the limitations of an individual wireless sensing technology, and little attention has been paid to the potential benefits of sensing with more than one wireless technology. In this paper, we introduce the concept of cross-technology sensing for the first time, and propose LoFiSen to enable LoRa-to-WiFi sensing. LoFiSen leverages the strengths of both LoRa and WiFi—combining LoRa's long-range capability with WiFi's pervasiveness. The chirp characteristic of LoRa signal significantly improves the sensing range of WiFi, and the widespread availability of WiFi devices makes LoRa sensing more pervasive. LoFiSen is fully compatible with LoRa and WiFi protocols, and can work on commodity LoRa and WiFi hardware. The key component of our design is enabling the WiFi receiver to capture fine-grained LoRa signal variations for sensing. Real-world experiments demonstrate that LoFiSen improves the WiFi sensing range for respiration monitoring from 8 m to 41 m, and pushes the walking sensing range from 16 m to 73.5 m. Through-wall passive respiration monitoring, previously infeasible with state-of-the-art WiFi sensing, is now possible with LoFiSen.

LoRa-based contact-free wireless sensing has attracted a lot of attention owing to its long sensing range, enabling wide-area sensing for the first time. While promising, existing LoRa sensing assumes there is no communication going on which is usually not true. We observe a severe degradation of sensing performance in real-world settings in the presence of communication. This issue hinders LoRa sensing from being adopted in real life and being integrated into the already established LoRa networking infrastructure. In this paper, we propose LSencom which takes the first step toward making LoRa-based wireless sensing work in the presence of communication. The key design is to employ the reversed chirp, i.e., downchirp, for sensing while keeping the original upchirp for communication. This design smartly leverages the orthogonality between downchirp and upchirp to mitigate the interference between communication and sensing. While the upchirp-downchirp design can remove most of the interference, we further adopt a novel chirp rotation method to deal with the remaining power leakage interference from upchirp to downchirp, enhancing the sensing performance. We implement LSencom on commodity LoRa nodes. Real-world experiments demonstrate that LSencom can reduce the communication-induced interference on sensing by 22.3 dB, and enable LoRa sensing even in the presence of multiple communication links.

Pediatric mental health is a growing concern around the world, affecting children’s social-emotional development and increasing the risk of poor behavioral outcomes later in life. However, obtaining a behavioral diagnosis in early childhood is challenging due to lack of access to resources, low parental mental health literacy, and children’s dependence on several stakeholders to coordinate care for them. While app-based, at-home screening tools could offer a scalable and convenient diagnostic solution for families, stakeholder perspectives on their utility and usability remain to be examined. This work reports on a survey of child mental health practitioners and interviews with parents to illustrate existing barriers to care that stakeholders encounter, the perceived benefits of app-based screening tools in meeting their needs, and the challenges in scaling up these tools. We identify where stakeholders agree or disagree, delineate key design tensions, and provide recommendations for the development of future screening technologies.

While existing strategies to execute deep learning-based classification on low-power platforms assume the models are trained on all classes of interest, this paper posits that adopting context-awareness i.e. narrowing down a classification task to the current deployment context consisting of only recent inference queries can substantially enhance performance in resource-constrained environments. We propose a new paradigm, CACTUS, for scalable and efficient context-aware classification where a micro-classifier recognizes a small set of classes relevant to the current context and, when context change happens (e.g., a new class comes into the scene), rapidly switches to another suitable micro-classifier. CACTUS features several innovations, including optimizing the training cost of context-aware classifiers, enabling on-the-fly context-aware switching between classifiers, and balancing context switching costs and performance gains via simple yet effective switching policies. We show that CACTUS achieves significant benefits in accuracy, latency, and compute budget across a range of datasets and IoT platforms.

While a number of recent efforts have explored the use of "cloud offload" to enable deep learning on IoT devices, these have not assumed the use of duty-cycled radios like BLE. We argue that radio duty-cycling significantly diminishes the performance of existing cloud-offload methods. We tackle this problem by leveraging a previously unexplored opportunity to use early-exit offload enhanced with prioritized communication, dynamic pooling, and dynamic fusion of features. We show that our system, FLEET, achieves significant benefits in accuracy, latency, and compute budget compared to state-of-art local early exit, remote processing, and model partitioning schemes across a range of DNN models, datasets, and IoT platforms.

Wireless sensing is capable of capturing rich information of human target without requiring sensors attached to the target. Although promising, two critical issues still exist, i.e., (i) limited sensing range, and (ii) severe interference in real-world settings. Recently, LoRa is employed to improve the sensing range. Although LoRa sensing is able to achieve a longer sensing range than WiFi and acoustic sensing, it is still limited to tens of meters. In this paper, we propose ChirpSen, which fully exploits the property of chirp signal to increase the sensing range. ChirpSen adopts a chirp concentration scheme to concentrate the power of all signal samples in a LoRa chirp at one timestamp, improving the signal power and accordingly boosting the sensing range. With a longer sensing range, the interference issue also becomes more severe. We propose a novel scheme to flexibly control the sensing coverage by tuning the LoRa chirp length in software. Real-world experiments show that ChirpSen is able to increase the detection range of a small size drone (12 cm × 10 cm × 8 cm) from 18 m to 160 m. ChirpSen is capable of monitoring a human's respiration rate at 138 m and tracking a human target's walking trajectory 210 m away.

Recent trends in flexible antennas and printed circuit boards present an opportunity to leverage deformable substrates such as textiles to deploy large UHF, VHF and ISM band antenna arrays in smart homes. Low-frequency large antenna arrays are rarely deployed in indoor settings due to their large size which makes them bulky and difficult to deploy. By embedding these arrays on existing surfaces such as curtains, we can improve through-wall sensing, beamforming for IoT devices equipped with low-power radios and indoor localization of Bluetooth tags.

Visual object tracking has seen significant progress in recent years. However, the vast majority of this work focuses on tracking objects within the image plane of a single camera and ignores the uncertainty associated with predicted object locations. In this work, we focus on the geospatial object tracking problem using data from a distributed camera network. The goal is to predict an object’s track in geospatial coordinates along with uncertainty over the object’s location while respecting communication constraints that prohibit centralizing raw image data. We present a novel single-object geospatial tracking data set that includes high-accuracy ground truth object locations and video data from a network of four cameras. We present a modeling framework for addressing this task including a novel backbone model and explore how uncertainty calibration and fine-tuning through a differentiable tracker affect performance.

The diagnosis of psychological disorders in early childhood is of utmost importance given their severe impact on children's academic and social skills as well as general adaptive functioning. Wearable and video-based systems have the potential to collect important diagnostic information in the form of neurophysiological and behavioral signals. However, accurate prediction of psychological disorder status from multimodal data streams necessitates their combination into meaningful features for classification models. In this work, we present a multitask supervised contrastive learning approach to learn useful multimodal embeddings from functional Near-Infrared Spectroscopy, galvanic skin response, and facial video data collected during a frustration-inducing task. The generated embeddings are able to accurately infer emotion regulation-related psychological disorders with an F1 score of 0.91, having significant implications for early-childhood mental health diagnoses.

Recent advances in fabric-based sensors have made it possible to densely instrument textile surfaces on smart toys without changing their look and feel. While such surfaces can be instrumented with traditional sensors, rigid elements change the nature of interaction and diminish the appeal of plush toys.

Sensing and data collection platforms are the crucial components of high-quality datasets that can fuel advancements in research. However, such platforms usually are ad-hoc designs and are limited in sensor modalities. In this paper, we discuss our experience designing and deploying a multi-modal multi-node sensor data collection platform that can be utilized for various data collection tasks. The main goal of this platform is to create a modality-rich data collection platform suitable for Internet of Things (IoT) applications with easy reproducibility and deployment, which can accelerate data collection and downstream research tasks.

Soil moisture sensing is a basic function required by applications like precision irrigation. Recently, RF based soil moisture sensing solutions [10, 43] have been proposed, which, however, can hardly support large scale deployment in challenging outdoor environments, since they must have dedicated signal emitters and also require power supply for either the signal emitters (WiFi or RFID reader) or both the transceivers (WiFi AP and client). LTE signal provides a unique opportunity for soil moisture sensing as the ubiquitously deployed base stations are naturally always-on signal emitters, eliminating the need for deploying extra hardware. In this paper, we implement a low-cost LTE based soil moisture sensor using commercial off-the-shelf hardware. We also realize duty-cycled soil sensing by automatically self-calibrating the phase offset after powering on the devices, significantly reducing the overall power consumption of the sensor. Extensive experiments show that our low-cost sensor ($55) achieves a high accuracy (3.15%) which is comparable to high-end soil moisture sensors ($850), wide coverage (2.4 km from the base station) and low power consumption (lasting 16 months using batteries).

Wireless sensing is an emerging technology that can obtain rich context information of human targets in a contact-free manner. Though promising, a missing component of current wireless sensing is sensing under device motions. In this work, we propose to integrate wireless sensing with the mobility of a robot. This is non-trivial because we find that device motions can severely degrade the sensing performance and even completely fail existing wireless sensing systems. In this paper, we propose novel signal processing schemes to address the impact of device motions to enable sensing with device mobility. For the first time, we integrate the robot's mobility with LoRa sensing to enlarge the sensing coverage. Comprehensive experiments demonstrate the effectiveness of the proposed system. We employ two representative sensing applications, i.e., fine-grained respiration monitoring and coarse-grained human walking sensing, to showcase the performance of our system. The proposed system is able to achieve accurate sensing in the presence of device motions, moving wireless sensing one step forward towards truly ubiquitous sensing for real-life adoption.

Clinical-grade wearable sleep monitoring is a challenging problem since it requires concurrently monitoring brain activity, eye movement, muscle activity, cardio-respiratory features, and gross body movements. This requires multiple sensors to be worn at different locations as well as uncomfortable adhesives and discrete electronic components to be placed on the head. As a result, existing wearables either compromise comfort or compromise accuracy in tracking sleep variables. We propose PhyMask, an all-textile sleep monitoring solution that is practical and comfortable for continuous use and that acquires all signals of interest to sleep solely using comfortable textile sensors placed on the head. We show that PhyMask can be used to accurately measure all the signals required for precise sleep stage tracking and to extract advanced sleep markers such as spindles and K-complexes robustly in the real-world setting. We validate PhyMask against polysomnography (PSG) and show that it significantly outperforms two commercially-available sleep tracking wearables—Fitbit and Oura Ring.

A long-standing challenge in radios for wearables is to design ultra-low power, yet high performance receivers with good sensitivity and spectral efficiency while being compatible with WiFi. The vanilla envelope detector used in standard UHF RFID is the most popular receivers on backscatter tags since they are passive but suffer from poor sensitivity and cannot decode complex modulations, which makes them a poor choice for directly decoding data from WiFi packets.

While multi-modal deep learning is useful in distributed sensing tasks like human tracking, activity recognition, and audio and video analysis, deploying state-of-the-art multi-modal models in a wirelessly networked sensor system poses unique challenges. The data sizes for different modalities can be highly asymmetric (e.g., video vs. audio), and these differences can lead to significant delays between streams in the presence of wireless dynamics. Therefore, a slow stream can significantly slow down a multi-modal inference system in the cloud, leading to either increased latency (when blocked by the slow stream) or degradation in inference accuracy (if inference proceeds without waiting). In this paper, we introduce speculative inference on multi-modal data streams to adapt to these asymmetries across modalities. Rather than blocking inference until all sensor streams have arrived and been temporally aligned, we impute any missing, corrupt, or partially-available sensor data, then generate a speculative inference using the learned models and imputed data. A rollback module looks at the class output of speculative inference and determines whether the class is sufficiently robust to incomplete data to accept the result; if not, we roll back the inference and update the model's output. We implement the system in three multi-modal application scenarios using public datasets. The experimental results show that our system achieves 7 -- 128× latency speedup with the same accuracy as six state-of-the-art methods.

Advances in flexible conductive substrates such as conductive wallpaper and paint present new opportunities for optimizing the performance of IoT nodes in smart homes and buildings. In this paper, we explore an unconventional use of such substrates for pulling frequencies of oscillators across IoT devices and wireless front-ends connected to the substrate. We show that by using this technique, we can replace precise crystal oscillators by lower precision and lower cost ceramic oscillators without compromising their ability to be used for tasks that require precise frequencies such as frequency-synchronized multi-static backscatter and synchronized sampling. We present an end-to-end design including a) analysis of conditions under which frequency pulling of oscillators across conductive substrates can work, b) a new technique to detect frequency locking across oscillators without requiring explicit communication, and c) an adaptive method that can be used to synchronize oscillators at minimum power consumption. We then show that these elements can be composed to design a high-performance multi-static backscatter system that performs as well as one that uses a shared high-precision clock but at an order of magnitude less monetary cost. We show that our system can scale and operate at very low power, while having low complexity since it requires no explicit interaction among devices attached to the substrate.

Besides the communication function, wireless signals are recently exploited for sensing purposes, enabling diverse applications. However, designing a wireless sensing system that provides truly pervasive coverage at city or even national scale and at the same time does not affect ongoing data communication is still challenging. In this work, we propose to involve the pervasive LTE signals into the ecosystem of wireless sensing. Although LTE sensing solves the coverage issue and does not compromise the communication function, it brings unique challenges. Due to the long distance between LTE base stations and terminals, the LTE signal interacts with diverse objects during the propagation process which causes severe interference in sensing. We enable LTE sensing by designing delicate signal processing schemes to combat against the severe interference. We demonstrate the advantages of LTE sensing using two typical applications, indoor respiration sensing and outdoor traffic monitoring. Extensive experiments show that the proposed system can achieve highly accurate respiration sensing with the blind spot and orientation-sensitive issues greatly mitigated. For traffic monitoring, the error of car speed estimation is lower than 2 mph, as good as commercial devices on the market.

Recent years have seen dramatic advances in low-power neural accelerators that aim to bring deep learning analytics to IoT devices; simultaneously, there have been considerable advances in the design of low-power radios to enable efficient compute offload from IoT devices to the cloud. Neither is a panacea --- deep learning models are often too large for low-power accelerators and bandwidth needs are often too high for low-power radios. While there has been considerable work on deep learning for smartphone-class devices, these methods do not work well for small battery-powered IoT devices that are considerably more resource-constrained.

The recent innovation of frequency-shifted (FS) backscatter allows for backscattering with commodity devices, which are inherently half-duplex. However, their reliance on oscillators for generating the frequency-shifting signal on the tag, forces them to incur the transient phase of the oscillator before steady-state operation. We show how the oscillator's transient phase can pose a fundamental limitation for battery-less tags, resulting in significantly low bandwidth efficiencies, thereby limiting their practical usage.

Recent years have seen exciting developments in the use of RFID tags as sensors to enable a range of applications including home automation, health and wellness, and augmented reality. However, widespread use of RFIDs as sensors requires significant instrumentation to deploy tethered readers, which limits usability in mobile settings. Our solution is WearID, a low-power wrist-worn backscatter reader that bridges this gap and allows ubiquitous sensing of interaction with tagged objects. Our end-to-end design includes innovations in hardware architecture to reduce power consumption and deal with wrist attenuation and blockage, as well as signal processing architecture to reliably detect grasping, touching, and other hand-based interactions. We show via exhaustive characterization that WearID is roughly 6× more power efficient than state-of-art commercial readers, provides 3D coverage of 30 to 50 cm around the wrist despite body blockage, and can be used to reliably detect hand-based interactions. We also open source the design of WearID with the hope that this can enable a range of new and unexplored applications of wearables.

In mobile health (mHealth) and human activity recognition (HAR), collecting labeled data often comes at a significantly higher cost or level of user burden than collecting unlabeled data. This motivates the idea of attempting to optimize the collection of labeled data to minimize cost or burden. In this paper, we develop active learning methods that are tailored to the mHealth and HAR domains to address the problems of labeled data scarcity and the cost of labeled data collection. Specifically, we leverage between-user similarity to propose a novel hierarchical active learning framework that personalizes models for each user while sharing the labeled data collection burden across a group, thereby reducing the labeling effort required by any individual user. We evaluate our framework on a publicly available human activity recognition dataset. Our hierarchical active learning framework on average achieves between a 20% and 70% reduction in labeling effort when compared to standard active learning methods.

The ability to unobtrusively and continuously monitor one's facial expressions has implications for a variety of application domains ranging from affective computing to health-care and the entertainment industry The standard Facial Action Coding System (FACS) along with camera based methods have been shown to provide objective indicators of facial expressions; however, these approaches can also be fairly limited for mobile applications due to privacy concerns and awkward positioning of the camera. To bridge this gap, W!NCE re-purposes a commercially available Electrooculography-based eyeglass (J!NS MEME) for continuously and unobtrusively sensing of upper facial action units with high fidelity. W!NCE detects facial gestures using a two-stage processing pipeline involving motion artifact removal and facial action detection. We validate our system's applicability through extensive evaluation on data from 17 users under stationary and ambulatory settings.

The ability to monitor eye closures and blink patterns has long been known to enable accurate assessment of fatigue and drowsiness in individuals. Many measures of the eye are known to be correlated with fatigue including coarse-grained measures like the rate of blinks as well as fine-grained measures like the duration of blinks and the extent of eye closures. Despite a plethora of research validating these measures, we lack wearable devices that can continually and reliably monitor them in the natural environment. In this work, we present a low-power system, iLid, that can continually sense fine-grained measures such as blink duration and Percentage of Eye Closures (PERCLOS) at high frame rates of 100fps. We present a complete solution including design of the sensing, signal processing, and machine learning pipeline and implementation on a prototype computational eyeglass platform.

An upcoming frontier for distributed computing might literally save lives in future military operations. In civilian scenarios, significant efficiencies were gained from interconnecting devices into networked services and applications that automate much of everyday life from smart homes to intelligent transportation. The ecosystem of such applications and services is collectively called the Internet of Things (IoT). Can similar benefits be gained in a military context by developing an IoT for the battlefield? This paper describes unique challenges in such a context as well as potential risks, mitigation strategies, and benefits.

Smart apparel with embedded sensors have the potential to revolutionize human behavior sensing by leveraging everyday clothing as the sensing substrate. However, existing textile-based sensing techniques rely on tight-fitting garments to obtain sufficient signal to noise, making it uncomfortable to wear and limiting the technology to niche applications like athletic performance monitoring.

Duty-cycling has emerged as the predominant method for optimizing power consumption of low-power radios, particularly for sensors that transmit sporadically in small bursts. But duty-cycling is a poor fit for applications involving high-rate sensor data from wearable sensors such as IMUs, microphones, and imagers that need to stream data to the cloud to execute sophisticated machine learning models.

We consider the problem of continuous computer-vision based analysis of video streams from mobile cameras over extended periods. Given high computational demands, general visual processing must currently be offloaded to the cloud. To reduce mobile battery and bandwidth consumption, recent proposals offload only "interesting" video frames, discarding the rest. However, determining what to discard is itself typically a power-hungry computer vision calculation, very often well beyond what most mobile devices can afford on a continuous basis. We present the Glimpse system, a re-design of the conventional mobile video processing pipeline to support such "early discard" flexibly, efficiently and accurately. Glimpse is a novel architecture that gates wearable vision using low-power vision modalities. Our proposed architecture adds novel sensing, processing, algorithmic and programming-system components to the camera pipeline to this end. We present a complete implementation and evaluation of our design. In common settings, Glimpse reduces mobile power and data usage by more than one order of magnitude relative to earlier designs, and moves continuous vision on lightweight wearables to the realm of the practical.

The human eye offers a fascinating window into an individual's health, cognitive attention, and decision making, but we lack the ability to continually measure these parameters in the natural environment. We demonstrate CIDER, a system that operates in a highly optimized low-power mode under indoor settings by using a fast Search-Refine controller to track the eye, but detects when the environment switches to more challenging outdoor sunlight and switches models to operate robustly under this condition. Our design is holistic and tackles a) power consumption in digitizing pixels, estimating pupillary parameters, and illuminating the eye via near-infrared and b) error in estimating pupil center and pupil dilation. We demonstrate that CIDER can estimate pupil center with error less than two pixels (0.6°), and pupil diameter with error of one pixel (0.22mm). Our end-to-end results show that we can operate at power levels of roughly 7mW at a 4Hz eye tracking rate, or roughly 32mW at rates upwards of 250Hz.

Mobile health research on illicit drug use detection typically involves a two-stage study design where data to learn detectors is first collected in lab-based trials, followed by a deployment to subjects in a free-living environment to assess detector performance. While recent work has demonstrated the feasibility of wearable sensors for illicit drug use detection in the lab setting, several key problems can limit lab-to-field generalization performance. For example, lab-based data collection often has low ecological validity, the ground-truth event labels collected in the lab may not be available at the same level of temporal granularity in the field, and there can be significant variability between subjects. In this paper, we present domain adaptation methods for assessing and mitigating potential sources of performance loss in lab-to-field generalization and apply them to the problem of cocaine use detection from wearable electrocardiogram sensor data.

This demo presents our experiences in developing a joystick for videogames that uses RFID backscattering for battery-free operation. Specifically, we develop a system to gather data from a wireless and battery-less joystick, named JoyTag, while it interacts with a videogame console. Our system enables consumers to use JoyTag at every moment without caring about charging.

While many radio technologies are available for mobile devices, none of them are designed to deal with asymmetric available energy. Battery capacities of mobile devices vary by up to three orders of magnitude between laptops and wearables, and our inability to deal with such asymmetry has limited the lifetime of constrained portable devices.

In this paper, we look at making backscatter practical for ultra-low power on-body sensors by leveraging radios on existing smartphones and wearables (e.g. WiFi and Bluetooth). The difficulty lies in the fact that in order to extract the weak backscattered signal, the system needs to deal with self-interference from the wireless carrier (WiFi or Bluetooth) without relying on built-in capability to cancel or reject the carrier interference.

The human eye offers a fascinating window into an individual's health, cognitive attention, and decision making, but we lack the ability to continually measure these parameters in the natural environment. The challenges lie in: a) handling the complexity of continuous high-rate sensing from a camera and processing the image stream to estimate eye parameters, and b) dealing with the wide variability in illumination conditions in the natural environment. This paper explores the power-robustness tradeoffs inherent in the design of a wearable eye tracker, and proposes a novel staged architecture that enables graceful adaptation across the spectrum of real-world illumination. We propose CIDER, a system that operates in a highly optimized low-power mode under indoor settings by using a fast Search-Refine controller to track the eye, but detects when the environment switches to more challenging outdoor sunlight and switches models to operate robustly under this condition. Our design is holistic and tackles a) power consumption in digitizing pixels, estimating pupillary parameters, and illuminating the eye via near-infrared, b) error in estimating pupil center and pupil dilation, and c) model training procedures that involve zero effort from a user. We demonstrate that CIDER can estimate pupil center with error less than two pixels (0.6°), and pupil diameter with error of one pixel (0.22mm). Our end-to-end results show that we can operate at power levels of roughly 7mW at a 4Hz eye tracking rate, or roughly 32mW at rates upwards of 250Hz.

Backscatter provides dual-benefits of energy harvesting and low-power communication, making it attractive to a broad class of wireless sensors. But the design of a protocol that enables extremely power-efficient radios for harvesting-based sensors as well as high-rate data transfer for data-rich sensors presents a conundrum. In this paper, we present a new {\em fully asymmetric} backscatter communication protocol where nodes blindly transmit data as and when they sense. This model enables fully flexible node designs, from extraordinarily power-efficient backscatter radios that consume barely a few micro-watts to high-throughput radios that can stream at hundreds of Kbps while consuming a paltry tens of micro-watts. The challenge, however, lies in decoding concurrent streams at the reader, which we achieve using a novel combination of time-domain separation of interleaved signal edges, and phase-domain separation of colliding transmissions. We provide an implementation of our protocol, LF-Backscatter, and show that it can achieve an order of magnitude or more improvement in throughput, latency and power over state-of-art alternatives.

Mobile sensor data-to-knowledge (MD2K) was chosen as one of 11 Big Data Centers of Excellence by the National Institutes of Health, as part of its Big Data-to-Knowledge initiative. MD2K is developing innovative tools to streamline the collection, integration, management, visualization, analysis, and interpretation of health data generated by mobile and wearable sensors. The goal of the big data solutions being developed by MD2K is to reliably quantify physical, biological, behavioral, social, and environmental factors that contribute to health and disease risk. The research conducted by MD2K is targeted at improving health through early detection of adverse health events and by facilitating prevention. MD2K will make its tools, software, and training materials widely available and will also organize workshops and seminars to encourage their use by researchers and clinicians.

Continuous, real-time tracking of eye gaze is valuable in a variety of scenarios including hands-free interaction with the physical world, detection of unsafe behaviors, leveraging visual context for advertising, life logging, and others. While eye tracking is commonly used in clinical trials and user studies, it has not bridged the gap to everyday consumer use. The challenge is that a real-time eye tracker is a power-hungry and computation-intensive device which requires continuous sensing of the eye using an imager running at many tens of frames per second, and continuous processing of the image stream using sophisticated gaze estimation algorithms. Our key contribution is the design of an eye tracker that dramatically reduces the sensing and computation needs for eye tracking, thereby achieving orders of magnitude reductions in power consumption and form-factor. The key idea is that eye images are extremely redundant, therefore we can estimate gaze by using a small subset of carefully chosen pixels per frame. We use a sparse pixel-based gaze estimation algorithm that is a multi-layer neural network learned using a state-of-the-art sparsity-inducing regularization function which minimizes the gaze prediction error while simultaneously minimizing the number of pixels used. Our results show that we can operate at roughly 70mW of power, while continuously estimating eye gaze at the rate of 30 Hz with errors of roughly 4 degrees.

This paper argues for a clean-slate redesign of wireless sensor systems to take advantage of the extremely low power consumption of backscatter communication and emerging ultra-low power sensor modalities. We make the case that existing sensing architectures incur substantial overhead for a variety of computational blocks between the sensor and RF front end - while these overheads were negligible on platforms where communication was expensive, they become the bottleneck on backscatter-based systems and increase power consumption while limiting throughput. We present a radically new design that is minimalist, yet efficient, and designed to operate end-to-end at tens of μWs while enabling high-data rate backscatter at rates upwards of many hundreds of Kbps. In addition, we demonstrate a complex reader-driven MAC layer that jointly considers energy, channel conditions, data utility, and platform constraints to enable network-wide throughput optimizations. We instantiate this architecture on a custom FPGA-based platform connected to microphones, and show that the platform consumes 73x lower power and has 12.5x higher throughput than existing backscatter-based sensing platforms.

Continuous, real-time tracking of eye gaze is valuable in a variety of scenarios including hands-free interaction with the physical world, detection of unsafe behaviors, leveraging visual context for advertising, life logging, and others. While eye tracking is commonly used in clinical trials and user studies, it has not bridged the gap to everyday consumer use. The challenge is that a real-time eye tracker is a power-hungry and computation-intensive device which requires continuous sensing of the eye using an imager running at many tens of frames per second, and continuous processing of the image stream using sophisticated gaze estimation algorithms. Our key contribution is the design of an eye tracker that dramatically reduces the sensing and computation needs for eye tracking, thereby achieving orders of magnitude reductions in power consumption and form-factor. The key idea is that eye images are extremely redundant, therefore we can estimate gaze by using a small subset of carefully chosen pixels per frame. We instantiate this idea in a prototype hardware platform equipped with a low-power image sensor that provides random access to pixel values, a low-power ARM Cortex M3 microcontroller, and a bluetooth radio to communicate with a mobile phone. The sparse pixel-based gaze estimation algorithm is a multi-layer neural network learned using a state-of-the-art sparsity-inducing regularization function that minimizes the gaze prediction error while simultaneously minimizing the number of pixels used. Our results show that we can operate at roughly 70mW of power, while continuously estimating eye gaze at the rate of 30 Hz with errors of roughly 3 degrees.

, a mobile solution that leverages a wristband containing a 9-axis inertial measurement unit to capture changes in the orientation of a person's arm, and a machine learning pipeline that processes this data to accurately detect smoking gestures and sessions in real-time. Our key innovations are fourfold: a) an arm trajectory-based method that extracts candidate hand-to-mouth gestures, b) a set of trajectory-based features to distinguish smoking gestures from confounding gestures including eating and drinking, c) a probabilistic model that analyzes sequences of hand-to-mouth gestures and infers which gestures are part of individual smoking sessions, and d) a method that leverages multiple IMUs placed on a person's body together with 3D animation of a person's arm to reduce burden of self-reports for labeled data collection. Our experiments show that our gesture recognition algorithm can detect smoking gestures with high accuracy (95.7%), precision (91%) and recall (81%). We also report a user study that demonstrates that we can accurately detect the number of smoking sessions with very few false positives over the period of a day, and that we can reliably extract the beginning and end of smoking session periods.

The ubiquity of smartphones has led to the emergence of mobile crowdsourcing markets, where smartphone users participate to perform tasks in the physical world. Mobile crowdsourcing markets are uniquely different from their online counterparts in that they require spatial mobility, and are therefore impacted by geographic factors and constraints that are not present in the online case. Despite the emergence and importance of such mobile marketplaces, little to none is known about the labor dynamics and mobility patterns of agents. This paper provides an in-depth exploration of labor dynamics in mobile task markets based on a year-long dataset from a leading mobile crowdsourcing platform. We find that a small core group of workers (< 10%) account for a disproportionately large proportion of activity (> 80%) generated in the market. We find that these super agents are more efficient than other agents across several dimensions: a) they are willing to move longer distances to perform tasks, yet they amortize travel across more tasks, b) they work and search for tasks more efficiently, c) they have higher data quality in terms of accepted submissions, and d) they improve in almost all of these efficiency measures over time. We find that super agent efficiency stems from two simple optimizations --- they are 3x more likely than other agents to chain tasks and they pick fewer lower priced tasks than other agents. We compare mobile and online micro-task markets, and discuss differences in demographics, data quality, and time of use, as well as similarities in super agent behavior. We conclude with a discussion of how a mobile micro-task market might leverage some of our results to improve performance.

Paper displays have a number of attractive properties, in particular the ability to present visual information perpetually with no power source. However, they are not digitally updatable or re-usable. Bistable display materials, such as e-paper, promise to enable displays with the best properties of both paper and electronic displays. However, rewriting a pixelated bistable display requires substantial energy, both for communication and for setting the pixel states.

Ubiquitous physiological sensing has the potential to profoundly improve our understanding of human behavior, leading to more targeted treatments for a variety of disorders. The long term goal of this work is development of novel computational tools to support the study of addiction in the context of cocaine use. The current paper takes the first step in this important direction by posing a simple, but crucial question: Can cocaine use be reliably detected using wearable electrocardiogram (ECG) sensors? The main contributions in this paper include the presentation of a novel clinical study of cocaine use, the development of a computational pipeline for inferring morphological features from noisy ECG waveforms, and the evaluation of feature sets for cocaine use detection. Our results show that 32mg/70kg doses of cocaine can be detected with the area under the receiver operating characteristic curve levels above 0.9 both within and between-subjects.

Mobile phones have evolved from communication devices to indispensable accessories with access to real-time content. The increasing reliance on dynamic content comes at the cost of increased latency to pull the content from the Internet before the user can start using it. While prior work has explored parts of this problem, they ignore the bandwidth costs of prefetching, incur significant training overhead, need several sensors to be turned on, and do not consider practical systems issues that arise from the limited background processing capability supported by mobile operating systems. In this paper, we make app prefetch practical on mobile phones. Our contributions are two-fold. First, we design an app prediction algorithm, APPM, that requires no prior training, adapts to usage dynamics, predicts not only which app will be used next but also when it will be used, and provides high accuracy without requiring additional sensor context. Second, we perform parallel prefetch on screen unlock, a mechanism that leverages the benefits of prediction while operating within the constraints of mobile operating systems. Our experiments are conducted on long-term traces, live deployments on the Android Play Market, and user studies, and show that we outperform prior approaches to predicting app usage, while also providing practical ways to prefetch application content on mobile phones.

Near Field Communication (NFC) on mobile phones presents new opportunities and threats. While NFC is radically changing how we pay for merchandise, it opens a pandora's box of ways in which it may be misused by unscrupulous individuals. This could include malicious NFC tags that seek to compromise a mobile phone, malicious readers that try to generate fake mobile payment transactions or steal valuable financial information, and others. In this work, we look at how to protect mobile phones from these threats while not being vulnerable to them. We design a small form-factor "patch", EnGarde, that can be stuck on the back of a phone to provide the capability to jam malicious interactions. EnGarde is entirely passive and harvests power through the same NFC source that it guards, which makes our hardware design minimalist, and facilitates eventual integration with a phone. We tackle key technical challenges in this design including operating across a range of NFC protocols, jamming at extremely low power, harvesting sufficient power for perpetual operation while having minimal impact on the phone's battery, designing an intelligent jammer that blocks only when specific blacklisted behavior is detected, and importantly, the ability to do all this without compromising user experience when the phone interacts with a legitimate external NFC device.

Near Field Communication (NFC) on mobile phones presents new opportunities and threats. While NFC is radically changing how we pay for merchandise, it opens a pandora's box of ways in which it may be misused by unscrupulous individuals. This could include malicious NFC tags that seek to compromise a mobile phone, malicious readers that try to generate fake mobile payment transactions or steal valuable financial information, and others. In this demo, we show the capabilities of our hardware solution, EnGarde, that protects a mobile phone from these vulnerabilities.

The proliferation of sensors on mobile phones and wearables has led to a plethora of context classifiers designed to sense the individual's context. We argue that a key missing piece in mobile inference is a layer that fuses the outputs of several classifiers to learn deeper insights into an individual's habitual patterns and associated correlations between contexts, thereby enabling new systems optimizations and opportunities. In this paper, we design CQue, a dynamic bayesian network that operates over classifiers for individual contexts, observes relations across these outputs across time, and identifies opportunities for improving energy-efficiency and accuracy by taking advantage of relations. In addition, such a layer provides insights into privacy leakage that might occur when seemingly innocuous user context revealed to different applications on a phone may be combined to reveal more information than originally intended. In terms of system architecture, our key contribution is a clean separation between the detection layer and the fusion layer, enabling classifiers to solely focus on detecting the context, and leverage temporal smoothing and fusion mechanisms to further boost performance by just connecting to our higher-level inference engine. To applications and users, CQue provides a query interface, allowing a) applications to obtain more accurate context results while remaining agnostic of what classifiers/sensors are used and when, and b) users to specify what contexts they wish to keep private, and only allow information that has low leakage with the private context to be revealed. We implemented CQue in Android, and our results show that CQue can i) improve activity classification accuracy up to 42%, ii) reduce energy consumption in classifying social, location and activity contexts with high accuracy(>90%) by reducing the number of required classifiers by at least 33%, and iii) effectively detect and suppress contexts that reveal private information.

RFID-scale sensors present a new frontier for distributed sensing. In contrast to existing sensor deployments that rely on battery-powered sensors, RFID-scale sensors rely solely on harvested energy. These devices sense and store data when not in contact with a reader, and use backscatter communication to upload data when a reader is in range. Unlike conventional RFID tags that only transmit identifiers, RFID sensors need to transfer potentially large amounts of data to a reader during each contact event. In this paper, we propose several optimizations to the RFID network stack to support efficient bulk transfer while remaining compatible with existing Gen 2 readers. Our key contribution is the design of a coordinated bulk transfer protocol for RFID-scale sensors that maximizes channel utilization and minimizes energy lost due to idle listening while also minimizing collisions. We present an implementation of the protocol for the Intel WISP, and describe several parameters that are tuned using empirical measurements that characterize the wireless channel. Our results show that the burst protocol improves goodput in comparison to vanilla EPC Gen 2 tags, improves energy-efficiency, allows multiple RFID sensors to share the channel, and also coexists with passive, non-sensor tags.

Near Field Communications (NFC) is an emerging technology that conveniently establishes radio communication by bringing two entities in close proximity of one another. Many use cases for devices equipped with this technology have been proposed ranging from payment systems to convenient data exchange.

As mobile apps become more closely integrated into our everyday lives, mobile app interactions ought to be rapid and responsive. Unfortunately, even the basic primitive of launching a mobile app is sorrowfully sluggish: 20 seconds of delay is not uncommon even for very popular apps.

Backscatter communication offers an ultra-low power alternative to active radios in urban sensing deployments - communication is powered by a reader, thereby making it virtually "free". While backscatter communication has largely been used for extremely small amounts of data transfer (e.g. a 12 byte EPC identifier from an RFID tag), sensors need to use backscatter for continuous and high-volume sensor data transfer. To address this need, we describe a novel link layer that exploits unique characteristics of backscatter communication to optimize throughput. Our system offers several optimizations including 1) understanding of multi-path self-interference characteristics and link metrics that capture these characteristics, 2) design of novel mobility-aware probing techniques that use backscatter link signatures to determine when to probe the channel, 3) bitrate selection algorithms that use link metrics to determine the optimal bitrate, and 4) channel selection mechanism that optimize throughput while remaining compliant within FCC regulations. Our results show upto 3x increase in goodput over other mechanisms across a wide range of channel conditions, scales, and mobility scenarios.

We present BlockRate, a wireless bitrate adaptation algorithm designed for blocks, or large contiguous units of transmitted data, as opposed to small packets. Our work is motivated by the observation that recent research results suggest significant overhead amortization benefits of blocks. Yet state-of-the-art bitrate algorithms are optimized for adaptation on a per-packet basis, so they can either have the amortization benefits of blocks or high responsiveness to underlying channel conditions of packets, but not both.

Micro-incentives represent a new but little-studied trend in participant compensation for user studies. In this paper, we use a combination of statistical analysis and models from labor economics to evaluate three canonical micro-payment schemes in the context of high-burden user studies, where participants wear sensors for extended durations. We look at how these strategies affect compliance, data quality, and retention, and show that when used carefully, micro-payments can be highly beneficial. We find that data quality is different across the micro-incentive schemes we experimented with, and therefore the incentive strategy should be chosen with care. We think that adaptive micro-payment based incentives can be used to successfully incentivize future studies at much lower cost to the study designer, while ensuring high compliance, good data quality, and lower retention issues.

Mobile sensing is difficult without power. Emerging Computational RFIDs (CRFIDs) provide both sensing and general-purpose computation without batteries--instead relying on small capacitors charged by energy harvesting. CRFIDs have small form factors and consume less energy than traditional sensor motes. However, CRFIDs have yet to see widespread use because of limited autonomy and the propensity for frequent power loss as a result of the necessarily small capacitors that serve as a microcontroller's power supply. Our results show that hybrid harvesting CRFIDs, which use an ambient energy micro-harvester, can complete a variety of useful workloads--even in an environment with little ambient energy available.

Mobile phones are becoming increasingly sophisticated with a rich set of on-board sensors and ubiquitous wireless connectivity. However, the ability to fully exploit the sensing capabilities on mobile phones is stymied by limitations in multimedia processing techniques. For example, search using cellphone images often encounters high error rate due to low image quality.

Although memory is an important constraint in embedded sensor nodes, existing embedded applications and systems are typically designed to work under the memory constraints of a single platform and do not consider the interplay between memory and flash storage. In this paper, we present the design of a memory-adaptive flash-based embedded sensor system that allows an application to exploit the presence of flash and adapt to different amounts of RAM on the embedded device. We describe how such a system can be exploited by data-centric sensor applications. Our design involves several novel features: flash and memory-efficient storage and indexing, techniques for efficient storage reclamation, and intelligent buffer management to maximize write coalescing. Our results show that our system is highly energy-efficient under different workloads, and can be configured for embedded sensor platforms with memory constraints ranging from a few kilobytes to hundreds of kilobytes.

An important challenge in mobile sensor networks is to enable energy-efficient communication over a diversity of distanceYC while being robust to wireless effects caused by node mobility. In this paper, we argue that the pairing of two complementary radios with heterogeneous range characteristics enables greater range and interference diversity at lower energy cost than a single radio. We make three contributions towards the design of such multi-radio mobile sensor systems. First, we present the design of a novel reinforcement learning-based link layer algorithm that continually learns channel characteristics and dynamically decides when to switch between radios. Second, we describe a simple protocol that translates the benefits of the adaptive link layer into practice in an energy-efficient manner. Third, we present the design of Arthropod, a mote-class sensor platform that combines two such heterogneous radios (XE1205 and CC2420) and our implementation of the Q-learning based switching protocol in TinyOS 2.0. Using experiments conducted in a variety of urban and forested environments, we show that our system achieves up to 52% energy gains over a single radio system while handling node mobility. Our results also show that our system can handle short, medium and long-term wireless interference in such environments.

An important challenge in mobile sensor networks is to enable energy-efficient communication over a diversity of distances while being robust to wireless effects caused by node mobility. In this paper, we argue that the pairing of two complementary radios with heterogeneous range characteristics enables greater range diversity at lower energy cost than a single radio. We make three contributions towards the design of such multi-radio mobile sensor systems. First, we present the design of a novel reinforcement learning-based link layer algorithm that continually learns channel characteristics and dynamically decides when to switch between radios. Second, we describe a simple protocol that translates the benefits of the adaptive link layer into practice in an energy-efficient manner. Third, we present the design of Arthropod, a mote-class sensor platform that combines two such heterogeneous radios (XE1205 and CC2420) and our implementation of the Q-learning based switching protocol in TinyOS 2.0. Using experiments conducted in a variety of urban and forested environments, we show that our system achieves up to 52% energy gains over a single radio system.

RFID sensor networks comprising batteryless devices that are passively powered by RFID readers present exciting possibilities for ubiquitous computing applications. They require minimal maintenance, are cheap to manufacture and have small form factor. However, their lack of autonomy due to the need for constant power from an RFID reader hinders their deployment. We demonstrate that RFIDs augmented with ambient energy-harvesting capabilities may be used as a first-class sensor platform, allowing them to operate untethered from reader infrastructure. Specifically, we show that a CRFID-based accelerometer sensor can provide both real-time and delayed access to time-stamped sensor data when provided with a small amount of solar energy. The data is collected using a standard RFID reader and displayed in a graphical interface.

Crowdsourcing is a new paradigm for utilizing the power of "crowds" of people to facilitate large scale tasks that are costly or time consuming with traditional methods. Crowdsourcing has enormous potential that can be truly unleashed when extended to sensor-rich mobile devices, such as smart phones. In this paper, we demonstrate mCrowd, an iPhone based mobile crowdsourcing platform that enables mobile users to post and work on sensor-related crowdsourcing tasks. mCrowd enables mobile users to fully utilize the rich sensors equipped with iPhone to participate and accomplish crowdsourcing tasks at fingertips, including geolocation-aware image collection, image tagging, road traffic monitoring, and others.

Flash memories are in ubiquitous use for storage on sensor nodes, mobile devices, and enterprise servers. However, they present significant challenges in designing tree indexes due to their fundamentally different read and write characteristics in comparison to magnetic disks. In this paper, we present the Lazy-Adaptive Tree (LA-Tree), a novel index structure that is designed to improve performance by minimizing accesses to flash. The LA-tree has three key features: 1) it amortizes the cost of node reads and writes by performing update operations in a lazy manner using cascaded buffers, 2) it dynamically adapts buffer sizes to workload using an online algorithm, which we prove to be optimal under the cost model for raw NAND flashes, and 3) it optimizes index parameters, memory management, and storage reclamation to address flash constraints. Our performance results on raw NAND flashes show that the LA-Tree achieves 2x to 12x gains over the best of alternate schemes across a range of workloads and memory constraints. Initial results on SSDs are also promising, with 3x to 6x gains in most cases.

Local storage is required in many sensor network applications, both for archival of detailed event information, as well as to overcome sensor platform memory constraints. Recent gains in energy efficiency of new-generation NAND flash storage have strengthened the case for in-network storage by data-centric sensor network applications. We argue that current storage solutions offering a simple file system abstraction are inadequate for sensor applications to exploit storage. Instead, we propose Capsule—a rich, flexible and portable object storage abstraction that offers stream, file, array, queue and index storage objects for data storage and retrieval. Further, Capsule supports checkpointing and rollback of object state for fault tolerance. Our experiments demonstrate that Capsule provides platform independence, greater functionality and greater energy efficiency than existing storage solutions.

This paper presents the design and implementation of a dual-camera sensor network that can be used as a memory assistant tool for assisted living. Our system performs energy-efficient object detection and recognition of commonly misplaced objects. The novelty in our approach is the ability to tradeoff between recognition accuracy and computational efficiency by employing a combination of low complexity but less precise color histogram-based image recognition together with more complex image recognition using SIFT descriptors. In addition, our system can seamlessly integrate feedback from the user to improve the robustness of object recognition. Experimental results reveal that our system is computation-efficient and adaptive to slow changes of environmental conditions.

Recent advances in sensor networks permit the use of a large number of relatively inexpensive distributed computational nodes with camera sensors linked in a network and possibly linked to one or more central servers. We argue that the full potential of such a distributed system can be realized if it is designed as a distributed search engine where images from different sensors can be captured, stored, searched and queried. However, unlike traditional image search engines that are focused on resource-rich situations, the resource limitations of camera sensor networks in terms of energy, bandwidth, computational power, and memory capacity present significant challenges. In this paper, we describe the design and implementation of a distributed search system over a camera sensor network where each node is a search engine that senses, stores and searches information. Our work involves innovation at many levels including local storage, local search, and distributed search, all of which are designed to be efficient under the resource constraints of sensor networks. We present an implementation of the search engine on a network of iMote2 sensor nodes equipped with low-power cameras and extended flash storage. We evaluate our system for a dataset comprising book images, and demonstrate more than two orders of magnitude reduction in the amount of data communicated and up to 5x reduction in overall energy consumption over alternate techniques.

This paper presents the design, implementation and evaluation of a distributed network of smart cameras whose function is to detect and localize falls, an important application in elderly living environments. A network of overlapping smart cameras uses a decentralized procedure for computing inter-image homographies that allows the location of a fall to be reported in 2D world coordinates by calibrating only one camera. Also, we propose a joint routing and homography transformation scheme for multi-hop localization that yields localization errors of less than 2 feet using very low resolution images. Our goal is to demonstrate that such a distributed low-power system can perform adequately in this and related applications. A prototype implementation is given for low-power Agilent/UCLA Cyclops cameras running on the Crossbow MICAz platform. We demonstrate the effectiveness of the fall detection as well as the precision of the localization using a simulation of our sample implementation.

The ease of deployment of battery-powered and mobile systems is pushing the network edge far from powered infrastructures. A primary challenge in building untethered systems is offering powerful aggregation points and gateways between heterogeneous end-points---a role traditionally played by powered servers. Microservers are battery-powered in-network nodes that play a number of roles: processing data fromclients, aggregating data, providing responses to queries, and actingas a network gateway. Providing QoS guarantees for theseservices can be extremely energy intensive. Since increasedenergy consumption translates to a shorter lifetime, there is a need for a new way to provide these QoS guarantees at minimal energy consumption.

In this paper, we focus on a network of rich sensors that are geographically distributed and argue that the design of such networks poses very different challenges from traditional mote-class sensor network design. We identify the need to handle the diverse requirements of multiple users to be a major design challenge, and propose a utility-driven approach to maximize data sharing across users while judiciously using limited network and computational resources. Our utility-driven architecture addresses three key challenges for such rich multi-user sensor networks: how to define utility functions for networks with data sharing among end-users, how to compress and prioritize data transmissions according to its importance to end-users, and how to gracefully degrade end-user utility in the presence of bandwidth fluctuations. We instantiate this architecture in the context of geographically distributed wireless radar sensor networks for weather, and present results from an implementation of our system on a multi-hop wireless mesh network that uses real radar data with real end-user applications. Our results demonstrate that our progressive compression and transmission approach achieves an order of magnitude improvement in application utility over existing utility-agnostic non-progressive approaches, while also scaling better with the number of nodes in the network.

The emerging of rich sensor networks poses very different design challenges from traditional "mote-class" sensor networks. One important challenge is that these networks are designed to handle the diverse requirements of multiple users. In this work, we demonstrate how multiple end user needs are handled in rich sensor networks using a utility-driven architecture. We instantiate this architecture in the context of geographically distributed wireless radar sensor networks for weather, and demonstrate the real-time operation of the prototype on a radar testbed in Okalahoma.

A camera sensor network is a wireless network of cameras designed for ad-hoc deployment. The camera sensors in such a network need to be properly calibrated by determining their location, orientation, and range. This paper presents Snapshot, an automated calibration protocol that is explicitly designed and optimized for camera sensor networks. Snapshot uses the inherent imaging abilities of the cameras themselves for calibration and can determine the location and orientation of a camera sensor using only four reference points. Our techniques draw upon principles from computer vision, optics, and geometry and are designed to work with low-fidelity, low-power camera sensors that are typical in sensor networks. An experimental evaluation of our prototype implementation shows that Snapshot yields an error of 1-2.5 degrees when determining the camera orientation and 5-10cm when determining the camera location. We show that this is a tolerable error in practice since a Snapshot-calibrated sensor network can track moving objects to within 11cm of their actual locations. Finally, our measurements indicate that Snapshot can calibrate a camera sensor within 20 seconds, enabling it to calibrate a sensor network containing tens of cameras within minutes.

Local storage is required in many sensor network applications, both for archival of detailed event information, as well as to overcome sensor platform memory constraints. While extensive measurement studies have been performed to highlight the trade-off between computation and communication in sensor networks, the role of storage has received little attention. The storage subsystems on currently available sensor platforms have not exploited technology trends, and consequently the energy cost of storage on these platforms is as high as that of communication. Current flash memories, however, offer a low-priced, high-capacity and extremely energy-efficient storage solution.In this paper, we perform a comprehensive evaluation of the active and sleep-mode energy consumption of available flash-based storage options for sensor platforms. Our results demonstrate more than a 100-fold decrease in per-byte energy consumption for surface-mount parallel NAND flash in comparison with the MicaZ on-board serial flash. In addition, this dramatically reduces storage energy costs relative to communication, introducing a new dimension in traditional computation vs communication trade-offs. Our results have significant ramifications on the design of sensor platforms as well as on the energy consumption of sensing applications. We quantify the potential energy gains for two commonly used sensor network services: communication and in-network data aggregation. Our measurements show significant improvements in each service: 50-fold and up to 10-fold reductions in energy for communication and data aggregation respectively.

Improved energy-efficiency and storage capacity of new-generation NAND flash memory makes a compelling case for storage-centric sensor networks. Such a storage-centric sensor network emphasizes the use of platforms with larger storage and more extensive use of the storage capacities on sensors. We demonstrate the feasibility of storage-centric sensor networks using an instance of a storage-centric camera sensor network that is more energy-efficient in comparison to a traditional camera sensor network. We demonstrate multiple camera sensors, each consisting of a Cyclops camera attached to a MicaZ mote, using motion-triggered image capturing. The captured images are archived locally on flash storage and summaries of detected events are transmitted to the base-station. The base-station picks the events of interest from the summaries and requests the original captured image from the sensor as required. The use of high-capacity energy-efficient flash storage at the sensor allows us to trade-off expensive radio communication for cheaper local storage, improving the life-time of the battery and consequently, the life of the storage-centric camera sensor network.

Recent gains in energy-efficiency of new-generation NAND flash storage have strengthened the case for in-network storage by data-centric sensor network applications. This paper argues that a simple file system abstraction is inadequate for realizing the full benefits of high-capacity lowpower NAND flash storage in data-centric applications. Instead we advocate a rich object storage abstraction to support flexible use of the storage system for a variety of application needs and one that is specifically optimized for memory and energy-constrained sensor platforms. We propose Capsule, an energy-optimized log-structured object storage system for flash memories that enables sensor applications to exploit storage resources in a multitude of ways. Capsule employs a hardware abstraction layer that hides the vagaries of flash memories for the application and supports energy-optimized implementations of commonly used storage objects such as streams, files, arrays, queues and lists. Further, Capsule supports checkpointing and rollback of object states to tolerate software faults in sensor applications running on inexpensive, unreliable hardware. Our experiments demonstrate that Capsule provides platform-independence, greater functionality, more tunability, and greater energy-efficiency than existing sensor storage solutions, while operating even within the memory constraints of the Mica2 Mote. Our experiments not only demonstrate the energy and memory-efficiency of I/O operations in Capsule but also shows that Capsule consumes less than 15% of the total energy cost in a typical sensor application.

We consider the joint optimization of sensor placement and transmission structure for data gathering, where a given number of nodes need to be placed in a field such that the sensed data can be reconstructed at a sink within specified distortion bounds while minimizing the energy consumed for communication. We assume that the nodes use either joint entropy coding based on explicit communication between sensor nodes, where coding is done when side information is available, or Slepian-Wolf coding where nodes have knowledge of network correlation statistics. We consider both maximum and average distortion bounds. We prove that this optimization is NP-complete since it involves an interplay between the spaces of possible transmission structures given radio reachability limitations, and feasible placements satisfying distortion bounds.We address this problem by first looking at the simplified problem of optimal placement in the one-dimensional case. An analytical solution is derived for the case when there is a simple aggregation scheme, and numerical results are provided for the cases when joint entropy encoding is used. We use the insight from our 1-D analysis to extend our results to the 2-D case and compare it to typical uniform random placement and shortest-path tree. Our algorithm for two-dimensional placement and transmission structure provides two to three fold reduction in total power consumption and between one to two orders of magnitude reduction in bottleneck power consumption. We perform an exhaustive performance analysis of our scheme under varying correlation models and model parameters and demonstrate that the performance improvement is typical over a range of data correlation models and parameters. We also study the impact of performing computationally-efficient data conditioning over a local scope rather than the entire network. Finally, we extend our explicit placement results to a randomized placement scheme and show that such a scheme can be effective when deployment does not permit exact node placement.

We consider data gathering by a network with a sink node and a tree communication structure, where the goal is to minimize the total transmission cost of transporting the information, collected by the nodes, to the sink node. This problem requires a joint optimization of the data representation at the nodes and of the transmission structure. First, we study the case when the measured data are correlated random variables, both in the lossless scenario with Slepian-Wolf coding, and in the high-resolution lossy scenario with optimal rate-distortion allocation. We show that the optimal transmission structure is the shortest path tree, and we find, in closed-form, the rate and distortion allocation. Second, we study the case when the measured data are deterministic piecewise constant signals, and data is described with adaptive level wavelet-based multiresolution representation. We show experimentally that, when computation is decentralized, there is an optimal network division into node groups of adaptive size. Finally, we also analyze the node positioning problem where, given a correlation structure and an available number of sensors, the goal is to place the nodes optimally in terms of minimizing the transmission cost; our results show that important gains can be obtained compared to a uniformly distributed sensor positioning.

This paper argues that a camera sensor network containing heterogeneous elements provides numerous benefits over traditional homogeneous sensor networks. We present the design and implementation of senseye---a multi-tier network of heterogeneous wireless nodes and cameras. To demonstrate its benefits, we implement a surveillance application using senseye comprising three tasks: object detection, recognition and tracking. We propose novel mechanisms for low-power low-latency detection, low-latency wakeups, efficient recognition and tracking. Our techniques show that a multi-tier sensor network can reconcile the traditionally conflicting systems goals of latency and energy-efficiency. An experimental evaluation of our prototype shows that, when compared to a single-tier prototype, our multi-tier senseye can achieve an order of magnitude reduction in energy usage while providing comparable surveillance accuracy.

In this position paper, we examine recent technology trends that have resulted in a broad spectrum of camera sensors, wireless radio technologies, and embedded sensor platforms with varying capabilities. We argue that future sensor applications will be hierarchical with multiple tiers, where each tier employs sensors with different characteristics. We argue that multi-tier networks are not only scalable, they offer a number of advantages over simpler, single-tier unimodal networks: lower cost, better coverage, higher functionality, and better reliability. However, the design of such mixed networks raises a number of new challenges that are not adequately addressed by current research. We discuss several of these challenges and illustrate how they can be addressed in the context of SensEye, a multi-tier video surveillance application that we are designing in our research group.

Archival storage of sensor data is necessary for applications that query, mine, and analyze such data for interesting features and trends. We argue that existing storage systems are designed primarily for flat hierarchies of homogeneous sensor nodes and do not fully exploit the multi-tier nature of emerging sensor networks, where an application can comprise tens of tethered proxies, each managing tens to hundreds of untethered sensors. We present TSAR, a fundamentally different storage architecture that envisions separation of data from metadata by employing local archiving at the sensors and distributed indexing at the proxies. At the proxy tier, TSAR employs a novel multi-resolution ordered distributed index structure, the Interval Skip Graph, for efficiently supporting spatio-temporal and value queries. At the sensor tier,TSAR supports energy-aware adaptive summarization that can trade off the cost of transmitting metadata to the proxies against the overhead of false hits resulting from querying a coarse-grain index. We implement TSAR in a two-tier sensor testbed comprising Stargate-based proxies and Mote-based sensors. Our experiments demonstrate the benefits and feasibility of using our energy-efficient storage architecture in multi-tier sensor networks.

Radio duty cycling has received significant attention in sensor networking literature, particularly in the form of protocols for medium access control and topology management. While many protocols have claimed to achieve significant duty-cycling benefits in theory and simulation, these benefits have often not translated to practice. The dominant factor that prevents the optimal usage of the radio in real deployment settings is time uncertainty between sensor nodes. This paper proposes an uncertainty-driven approach to duty-cycling where a model of long-term clock drift is used to minimize the duty-cycling overhead. First, we use long-term empirical measurements to evaluate and analyze in-depth the interplay between three key parameters that influence long-term synchronization - synchronization rate, history of past synchronization beacons and the estimation scheme. Second, we use this measurement-based study to design a rate-adaptive, energy-efficient long-term time synchronization algorithm that can adapt to changing clock drift and environmental conditions while achieving application-specific precision with very high probability. Finally, we integrate our uncertainty-driven time synchronization scheme with a MAC layer protocol, BMAC, and empirically demonstrate one to two orders of magnitude reduction in the transmit energy consumption at a node with negligible impact on the packet loss rate.

Time synchronization is critical to sensor networks at many layers of its design and enables better duty-cycling of the radio, accurate localization, beamforming and other collaborative signal processing. While there has been significant work in sensor network synchronization, measurement based studies have been restricted to very short-term (few minutes) datasets and have focused on obtaining accurate instantaneous synchronization. Long-term synchronization has typically been handled by periodic re-synchronization schemes with beacon intervals of a few minutes based on the assumption that long-term drift is too hard to model and predict. Thus, none of this work exploits the temporally correlated behavior of the clock drift. Yet, there are incredible energy gains to be achieved from better modeling and prediction of long-term drift that can provide bounds on long-term synchronization error across a sensor network. Better synchronization can lead to significantly lower duty-cycles of the radio, simplify signal processing and can enable an order of magnitude greater lifetime than current techniques.We measure, evaluate and analyze in-depth the long-term behavior of synchronization skew and drift on typical Mica sensor nodes and develop an efficient long-term time synchronization protocol. We use four real time data sets gathered over periods of 12-30 hours in different environmental conditions to study the interplay between three key parameters that influence long-term synchronization - synchronization rate, history of past synchronization beacons and the estimation scheme. We use this measurement-based study to design an online adaptive time-synchronization algorithm that can adapt to changing clock drift and environmental conditions while achieving application-specified precision with very high probability. We find that our algorithm achieves between one and two orders of magnitude improvement in energy efficiency over currently available time-synchronization approaches.

Wireless sensor networks enable dense sensing of the environment, offering unprecedented opportunities for observing the physical world. This article addresses two key challenges in wireless sensor networks: in-network storage and distributed search. The need for these techniques arises from the inability to provide persistent, centralized storage and querying in many sensor networks. Centralized storage requires multihop transmission of sensor data to Internet gateways which can quickly drain battery-operated nodes.Constructing a storage and search system that satisfies the requirements of data-rich scientific applications is a daunting task for many reasons: (a) the data requirements may be large compared to available storage and communication capacity of resource-constrained nodes, (b) user requirements are diverse and range from identification and collection of interesting event signatures to obtaining a deeper understanding of long-term trends and anomalies in the sensor events, and (c) many applications are in new domains where a priori information may not be available to reduce these requirements.This article describes a lossy, gracefully degrading storage model . We believe that such a model is necessary and sufficient for many scientific applications since it supports both progressive data collection for interesting events as well as long-term in-network storage for in-network querying and processing. Our system demonstrates the use of in-network wavelet-based summarization and progressive aging of summaries in support of long-term querying in storage and communication-constrained networks. We evaluate the performance of our linux implementation and show that it achieves: (a) low communication overhead for multiresolution summarization, (b) highly efficient drill-down search over such summaries, and (c) efficient use of network storage capacity through load-balancing and progressive aging of summaries.

We consider the joint optimization of sensor placement and transmission structure for data gathering, where a given number of nodes need to be placed in a field such that the sensed data can be reconstructed at a sink within specified distortion bounds while minimizing the energy consumed for communication. We assume that the nodes use joint entropy coding based on explicit communication between sensor nodes, and consider both maximum and average distortion bounds. The optimization is complex since it involves an interplay between the spaces of possible transmission structures given radio reachability limitations, and feasible placements satisfying distortion bounds. We address this problem by first looking at the simplified problem of optimal placement in the one-dimensional case. An analytical solution is derived for the case when there is a simple aggregation scheme, and numerical results are provided for the cases when joint entropy encoding is used. We use the insight from our 1-D analysis to extend our results to the 2-D case, and show that our algorithm for two-dimensional placement and transmission structure provides significant power benefit over a commonly used combination of uniformly random placement and shortest path trees.

Structural monitoring---the collection and analysis of structural response to ambient or forced excitation--is an important application of networked embedded sensing with significant commercial potential. The first generation of sensor networks for structural monitoring are likely to be data acquisition systems that collect data at a single node for centralized processing. In this paper, we discuss the design and evaluation of a wireless sensor network system (called Wisden for structural data acquisition. Wisden incorporates two novel mechanisms, reliable data transport using a hybrid of end-to-end and hop-by-hop recovery, and low-overhead data time-stamping that does not require global clock synchronization. We also study the applicability of wavelet-based compression techniques to overcome the bandwidth limitations imposed by low-power wireless radios. We describe our implementation of these mechanisms on the Mica-2 motes and evaluate the performance of our implementation. We also report experiences from deploying Wisden on a large structure.

Wireless sensor networks enable dense sensing of the environment, offering unprecedented opportunities for observing the physical world. Centralized data collection and analysis adversely impact sensor node lifetime. Previous sensor network research has, therefore, focused on in network aggregation and query processing, but has done so for applications where the features of interest are known a priori. When features are not known a priori, as is the case with many scientific applications in dense sensor arrays, efficient support for multi-resolution storage and iterative, drill-down queries is essential.Our system demonstrates the use of in-network wavelet-based summarization and progressive aging of summaries in support of long-term querying in storage and communication-constrained networks. We evaluate the performance of our linux implementation and show that it achieves: (a) low communication overhead for multi-resolution summarization, (b) highly efficient drill-down search over such summaries, and (c) efficient use of network storage capacity through load-balancing and progressive aging of summaries.

Previously proposed sensor network data dissemination schemes require periodic low-rate flooding of data in order to allow recovery from failure. We consider constructing two kinds of multipaths to enable energy efficient recovery from failure of the shortest path between source and sink. Disjoint multipath has been studied in the liteature. w propose a model braided multipath scheme, which results in several partially disjoint multipath schemes. We find that braided multipaths are a viable alternative for energy-efficient recovery from isolated and patterned failures

In most distributed systems, naming of nodes for low-level communication leverages topological location (such as node addresses) and is independent of any application. In this paper, we investigate an emerging class of distributed systems where low-level communication does not rely on network topological location. Rather, low-level communication is based on attributes that are external to the network topology and relevant to the application. When combined with dense deployment of nodes, this kind of named data enables in-network processing for data aggregation, collaborative signal processing, and similar problems. These approaches are essential for emerging applications such as sensor networks where resources such as bandwidth and energy are limited. This paper is the first description of the software architecture that supports named data and in-network processing in an operational, multi-application sensor-network. We show that approaches such as in-network aggregation and nested queries can significantly affect network traffic. In one experiment aggregation reduces traffic by up to 42% and nested queries reduce loss rates by 30%. Although aggregation has been previously studied in simulation, this paper demonstrates nested queries as another form of in-network processing, and it presents the first evaluation of these approaches over an operational testbed.