Frequent updates to machine learning (ML) models are essential for maintaining accuracy in dynamic environments. However, updating models deployed on IoT devices with low-power microcontrollers remains challenging. Because the model, application, and system are all linked together into a monolithic program, model updates are typically accomplished through full device firmware updates (DFUs). Unfortunately, DFUs are disruptive, inflexible, energy-intensive, and inefficient. We present Minerva, an end-to-end ML model update system for microcontroller-based IoT devices. At the heart of Minerva is a new abstraction and mechanism called capsules. Capsules encapsulate ML models as pure, independent functions and decouple the model from the rest of the system. These properties enable updates that are lightweight, avoiding the time and energy costs of a DFU both in communication and flash reprogramming, and non-disruptive, eliminating the need for a full reboot. Because capsules work at the application level, they require no modifications to the underlying OS or firmware, and thus apply broadly to existing IoT systems. Minerva reduces model update time by up to 89× compared to a full DFU, and up to 73× compared to a state-of-the-art embedded OS, enabling the growing number of deeply embedded devices to receive regular model updates. Minerva is an open-source project available at https://github.com/ShishirPatil/minerva.

With the widespread deployment of Bluetooth Low Energy (BLE) enabled devices, we are moving towards a world full of distributed, resource-constrained devices. With this growth comes the challenge of maintaining and communicating with millions of intermittently connected low-power BLE devices. While sending data from these BLE devices back to central servers has successfully been realized at scale (e.g. Apple AirTags), downlink communication is more difficult. Given the common usage of BLE-only devices in public infrastructure and other safety-critical environments, this inability to communicate with devices in a timely manner poses a barrier to realizing safe and secure ubiquitous computing. We explore the design space of providing downlink connectivity to BLE-only devices, give an overview of the challenges, and propose a system to enable downlink BLE communication.

I2C is a popular interconnect bus for integrated circuits that employs an open-drain design to support multiple controllers. Originally designed for television motherboards with a static configuration of chips, I2C is now used in nearly every corner of electronics, from wearables to phones to satellites. Some emerging applications, however, are more dynamic and distributed than I2C can support today, limiting its utility. In particular, one typically static, design-time parameter---the value of a pull-up resistor---is chosen to roughly match the total distributed bus and gate capacitance, ensuring a pull-up rise time that satisfies the I2C spec. But with greater dynamism---when many I2C nodes are added or removed at run-time---a static pull-up resistor can lead to a bus that is at best inefficient and at worst inoperable. In this paper, we present a system that can measure the rise time in situ and at I2C data rates, allowing us to optimally adjust the pull-up resistor at runtime. With this rise time measurement capability in hand, we show how it can be used for other useful and novel functions, including detecting when I2C nodes are added or removed, enabling an efficient inband-interrupt signaling scheme that eliminates the need for interrupt polling, and even full-duplex reverse data transfer encoded into clock edges. We implement our design, called I4C, using commercial microcontrollers and discrete electronics, and evaluate it on a testbed of several nodes, demonstrating its viability and efficiency.

Originally designed by Philips to interconnect static configurations of chips on TV motherboards, I2C is now used in nearly every corner of electronics, from computers to phones to satellites. Some emerging applications, however, are more dynamic and distributed than I2C can support today, limiting its utility. This need has become even more pressing with the rise of modular systems, which lead to I2C component systems being frequently added and removed from the bus, necessitating a more dynamic I2C operation.

While recent work explores novel tools to make electronics and device design easier and more accessible, these tend to be either highly automated (great for novices, but limiting for more advanced users) or highly manual (suitable for experts, but imposes a higher skill barrier to entry). In this work, we examine a middle ground: user-guided design space exploration to bridge an intuitive-but-ambiguous high-level representation to a fully-specified, fabrication-ready circuit. Our system helps users understand and make design choices by sweeping the design space of alternatives for electronics parts (e.g., choice of microcontroller), marking invalid options, and plotting points to visualize trade-offs (e.g., for power and size). We discuss the overall system and its structure, report on the results of a small but in-depth user study with participants from a wide range of electronics backgrounds, and draw insights on future directions for improving electronics design for everyone.

Reinforcement learning (RL) methods for social robot navigation show great success navigating robots through large crowds of people, but the performance of these learning-based methods tends to degrade in particularly challenging or unfamiliar situations due to the models’ dependency on representative training data. To ensure human safety and comfort, it is critical that these algorithms handle uncommon cases appropriately, but the low frequency and wide diversity of such situations present a significant challenge for these data-driven methods. To overcome this challenge, we propose modifications to the learning process that encourage these RL policies to maintain additional caution in unfamiliar situations. Specifically, we improve the Socially Attentive Reinforcement Learning (SARL) policy by (1) modifying the training process to systematically introduce deviations into a pedestrian model, (2) updating the value network to estimate and utilize pedestrian-unpredictability features, and (3) implementing a reward function to learn an effective response to pedestrian unpredictability. Compared to the original SARL policy, our modified policy maintains similar navigation times and path lengths, while reducing the number of collisions by 82% and reducing the proportion of time spent in the pedestrians’ personal space by up to 19 percentage points for the most difficult cases. We also describe how to apply these modifications to other RL policies and demonstrate that some key high-level behaviors of our approach transfer to a physical robot.

Embedded systems are deeply integrated into critical applications but, despite their importance, lack an effective means to apply over-the-air software patches without significant downtime. Standard mechanisms for firmware updates require device reboots that wipe important in-memory state. Prior efforts have proposed "live" updates to address this problem, applying patches to an embedded application without a reset, but they tackle a limited set of applications or propose a clean-slate design. In this paper, we present Retcon, a live update toolchain for embedded systems that supports a familiar event-driven programming model and does not require application code changes. Retcon leverages static analysis at compile time to determine when it will be safe to update a device. To find safe update points in the presence of complex asynchronous behavior, we define a novel system state, asynchronous quiescence, in which an update can be applied. We evaluate Retcon on a set of embedded event-driven applications – a dual-chamber pacemaker model, a programmable logic controller runtime, an artificial pancreas system, and a sensing node – and demonstrate Retcon’s ability to make low-overhead updates in less than one millisecond.

Imagine being able to deploy a small, battery- powered device nearly anywhere on earth that humans frequent and having it be able to send data to the cloud without needing to provision a network—without buying a physical gateway, setting up WiFi credentials, or acquiring a cellular SIM. Such a capability would address one of the greatest bottlenecks to deploying the long-tail of small, embedded, and power-constrained IoT devices in nearly any setting. Unfortunately, decoupling the device deployment from the network configuration needed to transmit, or backhaul, sensor data to the cloud remains a tricky challenge, but the success of Tile and AirTag offers hope. They have shown that mobile phones can crowd-source worldwide local network coverage to find lost items, yet expanding these systems to enable general-purpose backhaul raises privacy concerns for network participants. In this work, we present Nebula, a privacy-focused architecture for global, intermittent, and low-rate data backhaul to enable nearly any thing to eventually connect to the cloud while (i) preserving the privacy of the mobile network participants from the platform provider by decentralizing data flow through the system, (ii) incentivizing participation through micropayments, and (iii) preventing system abuse.

Fine-tuning models on edge devices like mobile phones would enable privacy-preserving personalization over sensitive data. However, edge training has historically been limited to relatively small models with simple architectures because training is both memory and energy intensive. We present POET, an algorithm to enable training large neural networks on memory-scarce battery-operated edge devices. POET jointly optimizes the integrated search search spaces of rematerialization and paging, two algorithms to reduce the memory consumption of backpropagation. Given a memory budget and a run-time constraint, we formulate a mixed-integer linear program (MILP) for energy-optimal training. Our approach enables training significantly larger models on embedded devices while reducing energy consumption while not modifying mathematical correctness of backpropagation. We demonstrate that it is possible to fine-tune both ResNet-18 and BERT within the memory constraints of a Cortex-M class embedded device while outperforming current edge training methods in energy efficiency. POET is an open-source project available at https://github.com/ShishirPatil/poet

The operation of wireless devices is often affected by their physical locations and spatial relationships to other objects and phenomena in the environment. As devices become increasing small and mobile, these relationships can change rapidly. We need tools that can provide an at-a-glance operating picture of each moment and highlight spatial patterns. Such insights are poorly supported by existing tools that abstract away the physical reality of the underlying network. Fortunately, recent advances in localization technologies and mixed reality platforms provide an opportunity for a new approach: mixed reality network introspection. Mixed reality allows users to see network traffic situated directly in the real world, taking advantage of human spatial intuition and visual processing to quickly identify behavior of interest. We propose an architecture for mixed reality network introspection, and implement a prototype called XRShark, a network visualizer with augmented reality and virtual reality modes. Initial evaluations of XRShark show promise in enabling users to identify behaviors of local networks that would be difficult to detect using traditional tools.

The radio transmitter is the most power-consuming component of a wireless embedded system. We present Judo, a radio transmitter that enables power balance between the wireless transmission, sensing, and processing tasks of a wireless embedded system. Judo transmitters leverage the fact that modern radio transceivers offer high receive sensitivity at low power. Therefore, even if the radio transmitter emits a weak signal, the link budget and transmission range will often remain high. With this key insight, we revisit the radio transmitter architecture by dramatically reducing the radiated power and hence the overall power draws. Specifically, Judo transmitter uses a tunnel diode oscillator to integrate the stages of a radio transmitter into a single energy-efficient step. In this step, baseband signals are generated and mixed using peak power below 100 μW. However, we sacrifice stability of tunnel diode oscillator for low-power consumption. We use the injection-locking phenomenon to stabilise the tunnel diode oscillator with an external carrier signal. Based on this novel architecture, we implement a transmitter that supports frequency-shift keying as a modulation scheme. Judo transmits over distances greater than 100 m at a bit rate of 100 kbps. It does so with an emitter device providing the carrier signal, and located at a distance of more than 100 m from Judo transmitter. In terms of critical link metrics, it outperforms the radio transmitters commonly used in wireless embedded systems.

Cloud services are vital tools for storing, analyzing, and responding to data collected by sensors in the Internet of Things (IoT). With ThingSpeak, a popular web platform that provides these services, users can easily create cloud "channels" to receive, host, and visualize sensor data. In this study, we scrape public channels from 6,511 users to construct a comprehensive picture of both the ThingSpeak developer community and their applications. We release this data to support future work. From this data, we examine 37,989 visualizations and uncover relationships between application domains and visualization techniques utilized. Further, we investigate how ThingSpeak's interface impacts user design choices. To learn which channels most successfully disseminate information, we explore design patterns on channels "liked" on Facebook or discussed in ThingSpeak forums. Finally, we briefly comment on how services like ThingSpeak can better support users' needs moving forward.

In much of the world, electricity grids are not instrumented at the customer level, limiting insights into the power quality experienced by utility customers. Moreover, to understand grid performance, regulators and investors must depend on utilities to self-report reliability data. To address these challenges, we introduce PowerWatch, an agile methodology to directly measure customer experience and aggregated grid performance without relying on the utility for deployment or management. PowerWatch employs a system of distributed sensors coupled with cloud-based analytics. We evaluate the PowerWatch methodology by deploying 462 sensors in homes and businesses in Accra, Ghana for over a year, yielding the largest open-source data set on electricity reliability at the customer-level in the region. We describe the architecture, design, and performance of PowerWatch, as well as the data that are collected, explaining how we determine the accuracy and coverage of our methodology without ground truth. Finally, we report on grid performance issues, finding nearly twice as many outages as the utility observed, suggesting a need for better grid performance monitoring.

In many engineering disciplines such as circuit board, chip, and mechanical design, a hardware description language (HDL) approach provides important benefits over direct manipulation interfaces by supporting concepts like abstraction and generator meta-programming. While several such HDLs have emerged recently and promised power and flexibility, they also present challenges – especially to designers familiar with current graphical workflows. In this work, we investigate an IDE approach to provide a graphical editor for a board-level circuit design HDL. Unlike GUI builders which convert an entire diagram to code, we instead propose generating equivalent HDL from individual graphical edit actions. By keeping code as the primary design input, we preserve the full power of the underlying HDL, while remaining useful even to advanced users. We discuss our concept, design considerations such as performance, system implementation, and report on the results of an exploratory remote user study with four experienced hardware designers.

One of the major challenges faced by Augmented Reality (AR) systems is linking virtual content accurately on physical objects and locations. This problem is amplified for applications like mobile payment, device control or secure pairing that requires authentication. In this paper, we present an active LED tag system called GLITTER that uses a combination of Bluetooth Low-Energy (BLE) and modulated LEDs to anchor AR content with no a priori training or labeling of an environment. Unlike traditional optical markers that encode data spatially, each active optical marker encodes a tag’s identifier by blinking over time, improving both the tag density and range compared to AR tags and QR codes.We show that with a low-power BLE-enabled micro-controller and a single 5 mm LED, we are able to accurately link AR content from potentially hundreds of tags simultaneously on a standard mobile phone from as far as 30 meters. Expanding upon this, using active optical markers as a primitive, we show how a constellation of active optical markers can be used for full 3D pose estimation, which is required for many AR applications, using either a single LED on a planar surface or two or more arbitrarily positioned LEDs. Our design supports 108 unique codes in a single field of view with a detection latency of less than 400 ms even when held by hand.

Social scientists, psychologists, and epidemiologists use empirical human interaction data to research human behaviour, social bonding, and disease spread. Historically, systems measuring interactions have been forced to choose between deployability and measurement fidelity---they operate only in instrumented spaces, under line-of-sight conditions, or provide coarse-grained proximity data. We introduce SociTrack, a platform for autonomous social interaction tracking via wireless distance measurements. Deployments require no supporting infrastructure and provide sub-second, decimeter-accurate ranging information over multiple days. The key insight that enables both deployability and fidelity in one system is to decouple node mobility and network management from range measurement, which results in a novel dual-radio architecture. SociTrack leverages an energy-efficient and scalable ranging protocol that is accurate to 14.8 cm (99th percentile) in complex indoor environments and allows our prototype to operate for 12 days on a 2000 mAh battery. Finally, to validate its deployability and efficacy, SociTrack is used by early childhood development researchers to capture caregiver-infant interactions.

Mainstream board-level circuit design tools work at the lowest level of design --- schematics and individual components. While novel tools experiment with higher levels of design, abstraction often comes at the expense of the fine-grained control afforded by low-level tools. In this work, we propose a hardware description language (HDL) approach that supports users at multiple levels of abstraction from broad system architecture to subcircuits and component selection. We extend the familiar hierarchical block diagram with polymorphism to include abstract-typed blocks (e.g., generic resistor supertype) and electronics modeling (i.e., currents and voltages). Such an approach brings the advantages of reusability and encapsulation from object-oriented programming, while addressing the unique needs of electronics designers such as physical correctness verification. We discuss the system design, including fundamental abstractions, the block diagram construction HDL, and user interfaces to inspect and fine-tune the design; demonstrate example designs built with our system; and present feedback from intermediate-level engineers who have worked with our system.

Printed Circuit Board (PCB) design tools are critical in helping users build non-trivial electronics devices. While recent work recognizes deficiencies with current tools and explores novel methods, little has been done to understand modern designers and their needs. To gain better insight into their practices, we interview fifteen electronics designers of a variety of backgrounds. Our open-ended, semi-structured interviews examine both overarching design flows and details of individual steps. One major finding was that most creative engineering work happens during system architecture, yet current tools operate at lower abstraction levels and create significant tedious work for designers. From that insight, we conceptualize abstractions and primitives for higher-level tools and elicit feedback from our participants on clickthrough mockups of design flows through an example project. We close with our observation on opportunities for improving board design tools and discuss generalizability of our findings beyond the electronics domain.

The vision of sensor systems that collect critical and previously ungathered information about the world is often only realized when sensors, students, and subjects move outside the academic laboratory. However, deployments at even the smallest scales introduce complexities and risks that can be difficult for a research team to anticipate. Over the past year, our interdisciplinary team of engineers and economists has been designing, deploying, and operating a large sensor network in Accra, Ghana that measures power outages and quality at households and firms. This network consists of 457 custom sensors, over 3,000 mobile app instances, thousands of participant surveys, and custom user incentive and deployment management systems. In part, this deployment supports an evaluation of the impacts of investments in the grid on reliability and the subsequent effects of improvements in reliability on socioeconomic well-being. We report our experiences as we move from performing small pilot deployments to our current scale, attempting to identify the pain points at each stage of the deployment. Finally, we extract high-level observations and lessons learned from our deployment activities, which we wish we had originally known when forecasting budgets, human resources, and project timelines. These insights will be critical as we look toward scaling our deployment to the entire city of Accra and beyond, and we hope that they will encourage and support other researchers looking to measure highly granular information about our world's critical systems.

Incentives are a key facet of human studies research, yet the state-of-the-art often designs and implements incentive systems in an ad-hoc, on-demand manner. We introduce the first vocabulary for formally describing incentive systems and develop a software infrastructure that enables UI-based graphical generation of complex, auditable, reliable, and reproducible incentive systems. We call this infrastructure the Open INcentive Kit (OINK). A review of recent literature from several communities finds that of the one hundred and twenty-one publications that incorporate incentives, only thirty-one describe their incentive system in detail, and all of these could be implemented using OINK. We evaluate OINK in practice by using it for an active energy monitoring deployment in Ghana and find that OINK successfully facilitates thousands of individual incentive payments. Finally, we describe our efforts to generalize OINK for different research communities, specifically focusing on architectural decisions around extensibility to support unanticipated use cases. OINK is free and open-source software.

We present TotTernary, the first infrastructure-free, long-term research platform which enables domain scientists to measure human interactions with previously unprecedented accuracy and flexibility. The platform can be tailored to varying scenarios by adjusting the ranging fidelity and the update rate and demonstrates dynamic adaptation to its environment. As a mobile, accurate, and reliable system for ranging measurements, it allows users to gather both distance and position information with decimeter accuracy. Measuring only 61 X 35 mm and weighing 7.7 g, it integrates two radios to achieve both low-power neighbour discovery and direct user interactions using Bluetooth Low Energy as well as precise ultra-wideband ranging measurements. The system introduces a novel energy-efficient ranging protocol with linear message complexity to achieve life times of up to 39 days. Furthermore, leveraging both antenna and frequency diversity, we demonstrate that the median ranging error can be reduced by up to 86% through an efficient aggregation of measurements over multiple channels. Our system is capable of highly reliable and consistent measurements with as little as 14.8 cm of ranging error in the 99th percentile and a 90% confidence interval of 11.3 cm.

Today, most sensors that harvest energy from indoor solar, ambient RF, or thermal gradients buffer small amounts of energy in capacitors as they intermittently work through a sensing task. While the utilization of capacitors for energy storage affords these systems indefinite lifetimes, their low energy capacity necessitates complex intermittent programming models for state retention and energy management. However, recent advances in battery technology lead us to reevaluate the impact that increased energy storage capacity may have on the necessity of these programming models and the reliability of energy harvesting sensors.

Today, most energy harvesting sensors rely on capacitors to buffer small amounts of energy as they intermittently work through a sensing task. While the utilization of capacitors for energy storage affords these systems indefinite lifetimes, their low energy capacity necessitates complex intermittent programming models for state retention and energy management. Recent advances in battery technology lead us to reevaluate the impact that increased energy storage capacity may have on the necessity of these programming models and the reliability and lifetime of energy harvesting sensors.

Low-power wide-area networks (LPWANs) are a compelling answer to the networking challenges faced by many Internet of Things devices. Their combination of low power, long range, and deployment ease has motivated a flurry of research, including exciting results on backscatter and interference cancellation that further lower power budgets and increase capacity. But despite the interest, we argue that unlicensed LPWAN technologies can only serve a narrow class of Internet of Things applications due to two principal challenges: capacity and coexistence. We propose a metric, bit flux, to describe networks and applications in terms of throughput over a coverage area. Using bit flux, we find that the combination of low bit rate and long range restricts the use case of LPWANs to sparse sensing applications. Furthermore, this lack of capacity leads networks to use as much available bandwidth as possible, and a lack of coexistence mechanisms causes poor performance in the presence of multiple, independently-administered networks. We discuss a variety of techniques and approaches that could be used to address these two challenges and enable LPWANs to achieve the promise of ubiquitous connectivity.

City-scale sensing holds the promise of enabling deeper insight into how our urban environments function. Applications such as observing air quality and measuring traffic flows can have powerful impacts, allowing city planners and citizen scientists alike to understand and improve their world. However, the path from conceiving applications to implementing them is fraught with difficulty. A successful city-scale deployment requires physical installation, power management, and communications-all challenging tasks standing between a good idea and a realized one. The Signpost platform, presented at IPSN 2018, has been created to address these challenges. Signpost enables easy deployment by relying on harvested, solar energy and wireless networking rather than their wired counterparts. To further lower the bar to deploying applications, the platform provides the key resources necessary to support its pluggable sensor modules in their distributed sensing tasks. In this demo, we present the Signpost hardware and several applications running on a deployment of Signposts on UC Berkeley's campus, including distributed, energy-adaptive traffic monitoring and fine grained weather reporting. Additionally we show the cloud infrastructure supporting the Signpost deployment, specifically the ability to push new applications and parameters down to existing sensors, with the goal of demonstrating that the existing deployment can serve as a future testbed.

City-scale sensing holds the promise of enabling a deeper understanding of our urban environments. However, a city-scale deployment requires physical installation, power management, and communications all challenging tasks standing between a good idea and a realized one. This indicates the need for a platform that enables easy deployment and experimentation for applications operating at city scale. To address these challenges, we present Signpost, a modular, energy-harvesting platform for city-scale sensing. Signpost simplifies deployment by eliminating the need for connection to wired infrastructure and instead harvesting energy from an integrated solar panel. The platform furnishes the key resources necessary to support multiple, pluggable sensor modules while providing fair, safe, and reliable sharing in the face of dynamic energy constraints. We deploy Signpost with several sensor modules, showing the viability of an energy-harvesting, multi-tenant, sensing system, and evaluate its ability to support sensing applications. We believe Signpost reduces the difficulty inherent in city-scale deployments, enables new experimentation, and provides improved insights into urban health.

Ultra wideband technology has shown great promise for providing high-quality location estimation, even in complex indoor multipath environments, but existing ultra wideband systems require tens to hundreds of milliwatts during operation. Backscatter communication has demonstrated the viability of astonishingly low-power tags, but has thus far been restricted to narrowband systems with low localization resolution. The challenge to combining these complimentary technologies is that they share a compounding limitation, constrained transmit power. Regulations limit ultra wideband transmissions to just -41.3 dBm/MHz, and a backscatter device can only reflect the power it receives. The solution is long-term integration of this limited power, lifting the initially imperceptible signal out of the noise. This integration only works while the target is stationary. However, stationary describes the vast majority of objects, especially lost ones. With this insight, we design Slocalization, a sub-microwatt, decimeter-accurate localization system that opens a new tradeoff space in localization systems and realizes an energy, size, and cost point that invites the localization of every thing. To evaluate this concept, we implement an energy-harvesting Slocalization tag and find that Slocalization can recover ultra wideband backscatter in under fifteen minutes across thirty meters of space and localize tags with a mean 3D Euclidean error of only 30 cm.

Network connectivity is often one of the most challenging aspects of deploying sensors. In many countries, cellular networks provide the most reliable, highest bandwidth, and greatest coverage option for internet access. Repurposing smartphones as gateways could extract value from hundreds of millions of devices currently considered to be e-waste. While these factors make smartphones a seemingly ideal platform to serve as a gateway between sensors and the cloud, we find that a device designed for multi-tenant operation and frequent human interaction becomes unreliable when tasked to continuously run a single application with no human interaction, a somewhat counter-intuitive result. Further, we find that economy phones cannot physically withstand continuous operation, resulting in a surprisingly high rate of permanent device failures in the field. If these observations hold more broadly, they would make mobile phones poorly suited to a range of sensing applications for which they have been rumored to hold great promise.

Network connectivity is often one of the most challenging aspects of deploying sensors. In many countries, cellular networks provide the most reliable, highest bandwidth, and greatest coverage option for Internet access. While this makes smartphones a seemingly ideal platform to serve as a gateway between sensors and the cloud, we find that a device designed for multi-tenant operation and frequent human interaction becomes unreliable when tasked to continuously run a single application with no human interaction, a seemingly counter-intuitive result. Further, we find that economy phones cannot physically withstand continuous operation, resulting in a surprisingly high rate of permanent device failures in the field. If these observations hold more broadly, they would make mobile phones poorly suited to a range of sensing applications for which they have been rumored to hold great promise.

We introduce Harmonium , a novel ultra wideband (UWB) RF localization architecture that achieves decimeter-scale accuracy indoors. Harmonium strikes a balance between tag simplicity and processing complexity to provide fast and accurate indoor location estimates. Harmonium uses only commodity components and consists of a small, inexpensive, lightweight, and FCC-compliant UWB transmitter or tag , fixed infrastructure anchors with known locations, and centralized processing that calculates the tag’s position. Anchors employ a new frequency-stepped narrowband receiver architecture that rejects narrowband interferers and extracts high-resolution timing information without the cost or complexity of traditional UWB approaches. In a complex indoor environment, 90% of position estimates obtained with Harmonium exhibit less than 31 cm of error with an average of 9 cm of inter-sample noise. In non-line-of-sight conditions (i.e., through-wall), 90% of position error is less than 42 cm. The tag draws 75 mW when actively transmitting, or 3.9 mJ per location fix at the 19 Hz update rate. Tags weigh 3 g and cost $4.50 USD at modest volumes. Furthermore, VLSI-based design concepts are identified for a simple, low-power realization of the Harmonium tag to offer a roadmap for the realization of Harmonium concepts in future integrated systems. Harmonium introduces a new design point for indoor localization and enables localization of small, fast objects such as micro quadrotors, devices previously restricted to expensive optical motion capture systems.

We present a method for calibration-free, infrastructure-free localization in sensor networks. Our strategy is to estimate node positions and noise distributions of all links in the network simultaneously - a strategy that has not been attempted thus far. In particular, we account for biased, non-line-of-sight (NLOS) range measurements from ultra-wideband (UWB) devices that lead to multi-modal noise distributions, for which few solutions exist to date. Our approach circumvents cumbersome a-priori calibration, allows for rapid deployment in unknown environments, and facilitates adaptation to changing conditions. Our first contribution is a generalization of the classical multidimensional scaling algorithm to account for measurements that have multi-modal error distributions. Our second contribution is an online approach that iterates between node localization and noise parameter estimation. We validate our method in 3-dimensional networks, (i) through simulation to test the sensitivity of the algorithm on its design parameters, and (ii) through physical experimentation in a NLOS environment. Our setup uses UWB devices that provide time-of-flight measurements, which can lead to positively biased distance measurements in NLOS conditions. We show that our algorithm converges to accurate position estimates, even when initial position estimates are very uncertain, initial error models are unknown, and a significant proportion of the network links are in NLOS.

City-scale sensing holds the promise of enabling deeper insight into how our urban environments function. Applications such as observing air quality and measuring sources of noise pollution can have powerful impacts, allowing city planners and citizen scientists alike to understand and improve their world. However, the path from conceiving applications to implementing them is fraught with many challenges. A successful city-scale deployment requires physical installation, power management, and communications---all challenging tasks standing between a good idea and a realized one, suggesting the need for a platform that enables easy deployment and experimentation of city-scale sensing applications. To address these basic challenges, we present Signpost, a modular platform for city-scale sensing. Signpost simplifies deployment and installation in cities by removing the need for connection to wired infrastructure and instead harvesting energy from an integrated solar panel. The platform provides the key resources necessary for its pluggable sensor modules to support city-scale applications. Signpost stores excess energy for later use, distributes energy between modules, and provides communication through multiple wireless protocols. It also offers local storage for sensor data and allows for additional processing in a duty-cycled Linux environment.

We describe opportunities and challenges with wireless robotic materials. Robotic materials are multi-functional composites that tightly integrate sensing, actuation, computation and communication to create smart composites that can sense their environment and change their physical properties in an arbitrary programmable manner. Computation and communication in such materials are based on miniature, possibly wireless, devices that are scattered in the material and interface with sensors and actuators inside the material. Whereas routing and processing of information within the material build upon results from the field of sensor networks, robotic materials are pushing the limits of sensor networks in both size (down to the order of microns) and numbers of devices (up to the order of millions). In order to solve the algorithmic and systems challenges of such an approach, which will involve not only computer scientists, but also roboticists, chemists and material scientists, the community requires a common platform --- much like the "Mote" that bootstrapped the widespread adoption of the field of sensor networks --- that is small, provides ample of computation, is equipped with basic networking functionalities, and preferably can be powered wirelessly.

Low-power microcontrollers lack some of the hardware features and most of the memory resources that usually enable multiprogrammable systems. Accordingly, operating system software for these platforms has not provided important features like memory isolation, dynamic memory allocation, and flexible concurrency. However, an emerging class of embedded applications are software platforms, rather than single purpose devices. Tock, a new operating system for low-power platforms, takes advantage of the limited hardware-protection mechanisms available on recent microcontrollers and the type-safety features of the Rust programming language to provide a multiprogramming environment that offers isolation of software faults, memory protection, and efficient memory management for dynamic application workloads written in any language while retaining the dependability requirements of long-running devices.

Low-power microcontrollers lack some of the hardware features and memory resources that enable multiprogrammable systems. Accordingly, microcontroller-based operating systems have not provided important features like fault isolation, dynamic memory allocation, and flexible concurrency. However, an emerging class of embedded applications are software platforms, rather than single purpose devices, and need these multiprogramming features. Tock, a new operating system for low-power platforms, takes advantage of limited hardware-protection mechanisms as well as the type-safety features of the Rust programming language to provide a multiprogramming environment for microcontrollers. Tock isolates software faults, provides memory protection, and efficiently manages memory for dynamic application workloads written in any language. It achieves this while retaining the dependability requirements of long-running applications.

We introduce Harmonium, a novel ultra-wideband RF localization architecture that achieves decimeter-scale accuracy indoors. Harmonium strikes a balance between tag simplicity and processing complexity to provide fast and accurate indoor location estimates. Harmonium uses only commodity components and consists of a small, inexpensive, lightweight, and FCC-compliant ultra-wideband transmitter or tag, fixed infrastructure anchors with known locations, and centralized processing that calculates the tag's position. Anchors employ a new frequency-stepped narrowband receiver architecture that rejects narrowband interferers and extracts high-resolution timing information without the cost or complexity of traditional ultra-wideband approaches. In a complex indoor environment, 90% of position estimates obtained with Harmonium exhibit less than 31 cm of error with an average 9 cm of inter-sample noise. In non-line-of-sight conditions (i.e. through-wall), 90% of position error is less than 42 cm. The tag draws 75 mW when actively transmitting, or 3.9 mJ per location fix at the 19 Hz update rate. Tags weigh 3 g and cost $4.50 USD at modest volumes. Harmonium introduces a new design point for indoor localization and enables localization of small, fast objects such as micro quadrotors, devices previously restricted to expensive optical motion capture systems.

The era of city-scale sensing is dawning. Supported by new sensing capabilities, the capability to detect and measure phenomena throughout a large area will allow deeper insight and understanding into how cities work. The challenge of city-scale sensing is not limited to developing new sensing applications, however. A sensor must be installed in a location. It must be provided power, storage, and communications. All these tasks stand aside from the desired sensing effort, but are necessary nevertheless.

We present Monoxalyze, a keychain-sized, Bluetooth-based, carbon monoxide breathalyzer that aims to enable mobile, scalable smoking cessation intervention programs. These intervention programs have been shown to greatly increase the rate of a quit attempt, which in turn decreases the rate of smoking, a major public health problem that still affects over one billion people around the world. Currently, intervention programs verify cessation compliance by requiring program participants to periodically visit clinics and exhale through large, expensive carbon monoxide breathalyzers---a practice that cannot scale to one billion smokers. Monoxalyze enables mobile cessation verification by working with a user's smartphone to establish a ring of spatio-temporal transitive trust between the Monoxalyze device, the user, and the smartphone, a concept that can be applied to many third-party monitoring applications. In Monoxalyze, the links of this trust are represented by simultaneous exhalation verification, facial recognition, and device-to-phone visible light authentication. In our evaluation, we show that Monoxalyze lasts over 80 days between charges, and has the ability to verify a Monoxalyze user. With a small user study we show that Monoxalyze determines smoking cessation with 92% accuracy, a level comparable with commercial CO breathalyzers. Further contributions describe the design decisions behind creating a low-power BLE device.

We present SurePoint, a system for drop-in, high-fidelity indoor localization. SurePoint builds on recently available commercial ultra-wideband radio hardware. While ultra-wideband radio hardware can provide the timing primitives necessary for a simple adaptation of two-way ranging, we show that with the addition of frequency and spatial diversity, we can achieve a 53% decrease in median ranging error. Because this extra diversity requires many additional packets for each range estimate, we next develop an efficient broadcast ranging protocol for localization that ameliorates this overhead. We evaluate the performance of this ranging protocol in stationary and fast-moving environments and find that it achieves up to 0.08 m median error and 0.53 m 99th percentile error. As ranging requires the tag to have exclusive access to the channel, we next develop a protocol to coordinate the localization of multiple tags in space. This protocol builds on recent work exploiting the constructive interference phenomenon. The ultra-wideband PHY uses a different modulation scheme compared to the narrowband PHY used by previous work, thus we first explore the viability and performance of constructive interference with ultra-wideband radios. Finally, as the ranging protocol requires careful management of the ultra-wideband radio and tight timing, we develop TriPoint, a dedicated "drop-in" ranging module that provides a simple I2C interface. We show that this additional microcontroller demands only marginal energy overhead while facilitating interoperability by freeing the primary microcontroller to handle other tasks.

We present SurePoint, a system for drop-in, high-fidelity indoor localization. SurePoint builds on recently available commercial ultra-wideband radio hardware. While ultra-wideband radios provide a pairwise range estimate natively, we show that with the addition of frequency and spatial diversity, we can achieve a 53% decrease in median range error. Because this extra diversity requires many additional packets for each range estimate, we leverage an efficient broadcast ranging protocol for localization that ameliorates this overhead. In stationary and fast-moving environments SurePoint achieves up to 0.08 m median error and 0.53 m 99th percentile error. As ranging requires the tag to have exclusive access to the channel, we employ a protocol to coordinate the localization of multiple tags in space. This protocol builds on recent work exploiting the constructive interference phenomenon, however SurePoint is the first to demonstrate constructive interface using the 802.15.4a ultra-wideband PHY. Finally, as the ranging protocol requires careful management of the ultra-wideband radio and tight timing, we utilize TriPoint, a dedicated "drop-in" ranging module that provides a simple I2C interface. We show that this additional microcontroller demands only marginal energy overhead while facilitating interoperability by freeing the primary microcontroller to handle other tasks.

For the last fifteen years, research explored the hardware, software, sensing, communication abstractions, languages, and protocols that could make networks of small, embedded devices---motes---sample and report data for long periods of time while unattended. Today, the application and technological landscapes have shifted, introducing new requirements and new capabilities. Hardware has evolved past 8 and 16 bit microcontrollers: there are now 32 bit processors with lower energy budgets and greater computing capability. New wireless link layers have emerged, creating protocols that support direct interaction with users, but introduce novel limitations that systems must consider. Programming language advances have led to the ability to write system kernels that guarantee safety and reliability while maintaining low overhead. The time has come to look beyond optimizing networks of motes. We look towards new technologies such as Bluetooth Low Energy, Cortex M processors, and capable multi-process operating systems, with new application spaces such as personal area networks, and new capabilities and requirements in security and privacy to inform contemporary hardware and software platforms. It is time for a new, open experimental platform in this post-mote era.

As we show in this paper, I/O has become the limiting factor in scaling down size and power toward the goal of invisible computing. Achieving this goal will require composing optimized and specialized---yet reusable---components with an interconnect that permits tiny, ultra-low power systems. In contrast to today's interconnects which are limited by power-hungry pull-ups or high-overhead chip-select lines, our approach provides a superset of common bus features but at lower power, with fixed area and pin count, using fully synthesizable logic, and with surprisingly low protocol overhead.

Building connected, pervasive, human-facing, and responsive applications that incorporate local sensors, smartphone interactions, device actuation, and cloud-based learning--the promised features of the Internet of Things (IoT)---requires a complete suite of tools spanning both hardware and software. We present a set of these pieces, including a gateway, four hardware building blocks, multiple sensor platforms, an indoor localization system, and software for connecting users and devices. Each piece plays an integral role towards enabling applications, from facilitating rapid development of wireless smart devices to composing data streams and services from a diverse set of components. By providing layered interoperable systems, our toolkit offers cohesive support for moving beyond single-device, cloud-centric applications---typical in today's IoT landscape--and towards richer applications that incorporate multiple data streams, human interaction, cloud processing, location awareness, multiple communication protocols, historical data, access control, and on-demand user interfaces. To show how the pieces in the toolkit cooperate, we demonstrate a location-based access control application where a user's smartphone can control a room's lighting, but only from within the room. Further, data streams from the phone and nearby sensors are used to provide a constant lighting service which attempts to maintain a user-set brightness under variable external lighting conditions.

We present PowerBlade, the smallest, lowest cost, and lowest power AC plug-load meter that measures real, reactive and apparent power, and reports this data, along with cumulative energy consumption, over an industry-standard Bluetooth Low Energy radio. Achieving this design point requires revisiting every aspect of conventional power meters: a new method of acquiring voltage; a non-invasive, planar method of current measurement; an efficient and accurate method of computing power from the voltage and current channels; a radio interface that leverages nearby smart phones to display data and report it to the cloud; and a retro power supply re-imagined with vastly lower current draw, allowing extreme miniaturization. PowerBlade occupies a mere 1" × 1" footprint, offers a 1/16" profile, draws less than 180 mW itself, offers 1.13% error on unity power factor loads in the 2-1200 W range and slightly worse for non-linear and reactive loads, and costs $11 in modest quantities of about 1,000 units. This new design point enables affordable large-scale studies of plug-load energy usage--an area of growing national importance.

We present PowerBlade, the smallest and lowest power AC plug-load meter that measures real, reactive and apparent power, and reports this data, along with cumulative energy consumption, over an industry-standard Bluetooth Low Energy radio. Achieving this design point requires revisiting every aspect of conventional power meters: a new method of acquiring voltage; a non-invasive, planar method of current measurement; an efficient and accurate method of computing power from the voltage and current channels; a radio interface that leverages nearby smart phones to display data and report it to the cloud; and a retro power supply reimagined with vastly lower current draw, allowing extreme miniaturization. PowerBlade occupies a mere 1" by 1" footprint, offers a 1/16" profile, draws 176 mW continuously, offers 1.13% error on unity power factor loads in the 2-1800 W range and slightly worse for non-linear and reactive loads, and costs $11 in modest quantities of about 1000 units. This new design point enables affordable large-scale studies of plug-load energy usage -- an area of growing national importance.

We demonstrate PolyPoint, a high-fidelity RF-based indoor localization system that achieves 28~cm accuracy indoors and tracks a fast-moving quadcopter with only 56~cm average error. PolyPoint uses ultra-wideband signals to achieve high precision RF time-of-flight estimates between nodes. To further improve accuracy, PolyPoint exploits two forms of diversity: frequency diversity, which leverages several ultra-wideband channels to improve channel response, and antenna diversity, which adds three antennas at 120 degree offsets to mitigate the effects of antenna polarization and nulls. PolyPoint introduces an efficient, novel ranging protocol that maximizes these diversity sources with a minimal number of packets.

We are becoming increasingly surrounded by smart and connected devices, popularly known as the Internet of Things. The emerging user interface paradigm for many such things eschews physical buttons, knobs, and displays in favor of virtual interfaces that are downloaded from the web and rendered on remote platforms---like smartphones. However, such smartphone app-based interfaces often require tedious discovery and installation, as well as device discovery, pairing, and configuration before a user can interact with a nearby device. Requiring an explicit app install for each new device type scales poorly with device growth, and particularly hinders casual interactions with ambient devices. Instead of the high-friction, walled-garden approach now taking root, we propose name, a physical web browser that provides a seamless, scalable approach to browsing and interacting with nearby things. name leverages multiple network patterns and modern web technologies to provide users with rich device interfaces, even for devices under network or power constraints. We argue that this approach scales better and that it provides more intuitive and natural functionality for both users and developers. This demo presents the basic concept, allows others to experience our preliminary implementation, and raises several open research questions.

Rust, a new systems programming language, provides compile-time memory safety checks to help eliminate runtime bugs that manifest from improper memory management. This feature is advantageous for operating system development, and especially for embedded OS development, where recovery and debugging are particularly challenging. However, embedded platforms are highly event-based, and Rust's memory safety mechanisms largely presume threads. In our experience developing an operating system for embedded systems in Rust, we have found that Rust's ownership model prevents otherwise safe resource sharing common in the embedded domain, conflicts with the reality of hardware resources, and hinders using closures for programming asynchronously. We describe these experiences and how they relate to memory safety as well as illustrate our workarounds that preserve the safety guarantees to the largest extent possible. In addition, we draw from our experience to propose a new language extension to Rust that would enable it to provide better memory safety tools for event-driven platforms.

We explore the indoor positioning problem with unmodified smartphones and slightly-modified commercial LED luminaires. The luminaires - modified to allow rapid, on-off keying - transmit their identifiers and/or locations encoded in human-imperceptible optical pulses. A camera-equipped smartphone, using just a single image frame capture, can detect the presence of the luminaires in the image, decode their transmitted identifiers and/or locations, and determine the smartphone's location and orientation relative to the luminaires. Continuous image capture and processing enables continuous position updates. The key insights underlying this work are (i) the driver circuits of emerging LED lighting systems can be easily modified to transmit data through on-off keying; (ii) the rolling shutter effect of CMOS imagers can be leveraged to receive many bits of data encoded in the optical transmissions with just a single frame capture, (iii) a camera is intrinsically an angle-of-arrival sensor, so the projection of multiple nearby light sources with known positions onto a camera's image plane can be framed as an instance of a sufficiently-constrained angle-of-arrival localization problem, and (iv) this problem can be solved with optimization techniques.

We explore the indoor positioning problem with unmodified smartphones and slightly-modified commercial LED luminaires. The luminaires-modified to allow rapid, on-off keying-transmit their identifiers and/or locations encoded in human-imperceptible optical pulses. A camera-equipped smartphone, using just a single image frame capture, can detect the presence of the luminaires in the image, decode their transmitted identifiers and/or locations, and determine the smartphone's location and orientation relative to the luminaires. Continuous image capture and processing enables continuous position updates. The key insights underlying this work are (i) the driver circuits of emerging LED lighting systems can be easily modified to transmit data through on-off keying; (ii) the rolling shutter effect of CMOS imagers can be leveraged to receive many bits of data encoded in the optical transmissions with just a single frame capture, (iii) a camera is intrinsically an angle-of-arrival sensor, so the projection of multiple nearby light sources with known positions onto a camera's image plane can be framed as an instance of a sufficiently-constrained angle-of-arrival localization problem, and (iv) this problem can be solved with optimization techniques. We explore the feasibility of the design through an analytical model, demonstrate the viability of the design through a prototype system, discuss the challenges to a practical deployment including usability and scalability, and demonstrate decimeter-level accuracy in both carefully controlled and more realistic human mobility scenarios.

Mobile computing has deeply influenced the lives of almost every human in the world on a daily basis. In celebration of the 20th anniversary of Mobicom, Mobicom 2014 features an exciting panel on the topic of tackling societal grand challenges using mobile computing. The panel features five excellent researchers who have had a tremendous impact on the field of mobile computing over the years and whose research work has directly addressed pressing societal problems. The panel discussion will feature an in-depth discussion of the grand challenges that we face in society today and the role of mobile computing as a frontier platform for addressing these challenges. The panel will cover both a retrospective and futuristic perspective where the retrospective aspects highlight the diverse contributions of the panelists and the futuristic aspects will feature of a discussion of their individual views of what are the next big societal problems that Mobicom as a community should tackle.

Power meters are critical for sub metering loads in residential and commercial settings, but high installation cost and complexity hamper their broader adoption. Recent approaches address installation burdens by proposing non-invasive meters that easily clip onto a wire, or stick onto a circuit breaker, to perform contact less metering. Unfortunately, these designs require regular maintenance (e.g. Battery replacement) or reduce measurement accuracy (e.g. Work poorly with non-unity power factors). This paper presents Gemini, a new design point in the power metering space. Gemini addresses the drawbacks of prior approaches by decoupling and distributing the AC voltage and current measurement acquisitions, and recombining them wirelessly using a low-bandwidth approach, to offer non-invasive real, reactive, and apparent power metering. Battery maintenance is eliminated by using an energy-harvesting design that enables the meter to power itself using a current transformer. Accuracy is substantially improved over other non-invasive meters by virtualizing the voltage channel -- effectively allowing the meter to calculate power as if it could directly measure voltage (since true power requires sample-by-sample multiplication of current and voltage measurements acquired with tight timing constraints). Collectively, these improvements result in a new design point that meters resistive loads with 0.6 W average error and a range of reactive and switching loads with 2.2 W average error -- matching commercial, mains-powered solutions.

Currently, researchers study face-to-face interactions using wearable sensors and smartphones which provide 2 to 5 m proximity sensing every 20 to 300 s. However, studying interaction distance, which is known to impact disease spread, communication behavior, and other phenomenon, has proven challenging. Smartphones are limited by their inaccurate and/or impractical ranging capabilities, while wearable sensors are limited by their need for infrastructure nodes, bulkiness, and/or inaccurate ranging. To address these challenges, we present Opo, a 14 cm2, 11.4 g "lapel pin" built from commercial components. Opo sensors range neighbors every 2 s up to 2 m away with 5% average error, all while requiring zero infrastructure and improving upon current wearable sensors' accuracy and power usage. The cornerstone of Opo is an ultrasonic wakeup circuit that draws 19 μA when no neighbors are present. This enables Opo sensors to discover and range neighbors without the need for infrastructure nodes and slow or power-hungry RF discovery protocols. Thus, Opo is able to sense interaction distance with high accuracy (5 cm) and temporal fidelity (2 s) on a limited power budget.

For the last fifteen years, research explored the hardware, software, sensing, communication abstractions, languages, and protocols that could make networks of small, embedded devices---motes---sample and report data for long periods of time unattended. Today, the application and technological landscapes have shifted, introducing new requirements and new capabilities. Hardware has evolved past 8 and 16 bit microcontrollers: there are now 32 bit processors with lower energy budgets and greater computing capability. New wireless link layers have emerged, creating protocols that support rapid and efficient setup and teardown but introduce novel limitations that systems must consider. The time has come to look beyond optimizing networks of motes. We look towards new technologies such as Bluetooth Low Energy, Cortex M processors, and capable energy harvesting, with new application spaces such as personal area networks, and new capabilities and requirements in security and privacy to inform contemporary hardware and software platforms. It is time for a new, open experimental platform in this post-mote era.

Packet collisions have gone from something to be avoided to something that can be embraced. We build upon recent results that employ intentional packet collisions for synchronized flooding to show how multi-hop wireless networks can support real-time actuation and control. First, we show how a network of nodes can synchronize their LED transmissions to extend the range of a visual light communication (VLC) system. Second, we show how buffer-free, streaming audio is possible over a multi-hop wireless mesh network.

Obtaining a detailed, whole-house breakdown of energy usage would allow for homeowners to better understand their energy consumption and opportunities for energy savings. Current solutions are either too course-grained, too difficult to deploy, not networked, or offer incomplete coverage of hard to meter items, such as ceiling lights. We demonstrate a wirelessly networked, energy-harvesting power metering system that draws zero standby power and is power proportional to the load it is metering.

Conventional AC power meters perform at least two distinct functions: power conversion, to supply the meter itself, and energy metering, to measure the load consumption. This paper presents Monjolo, a new energy-metering architecture that combines these two functions to yield a new design point in the metering space. The key insight underlying this work is that the output of a current transformer -- nominally used to measure a load current -- can be harvested and used to intermittently power a wireless sensor node. The hypothesis is that the node's activation frequency increases monotonically with the primary load's draw, making it possible to estimate load power from the interval between activations, assuming the node consumes a fixed energy quanta during each activation. This paper explores this thesis by designing, implementing, and evaluating the Monjolo metering architecture. The results demonstrate that it is possible to build a meter that draws zero-power under zero-load conditions, offers high accuracy for near-unity power factor loads, works with non-unity power factor loads in combination with a whole-house meter, wirelessly reports readings to a data aggregator, is resilient to communication failures, and is parsimonious with the radio channel, even under heavy loads. Monjolo eliminates the high-voltage AC-DC power supply and AC metering circuitry present in earlier designs, enabling a smaller, simpler, safer, and lower-cost design point that supports novel deployment scenarios like non-intrusive circuit-level metering.

We explore the seemingly trivial problem of scaling the AC power meter to a cubic-inch form factor. This small volume, necessary for unobtrusive inline plug-load monitoring, requires aggressive electromechanical codesign and 3-D electronics packaging. However, the key scaling challenge across both size and efficiency lies in the power supply architecture and load power. Hence, contrary to conventional wisdom, power is not free, even if a device is attached to wall power. Since many existing power supplies exhibit high baseline power draw, this leads to poor efficiency and runs counter to the green goals of many applications. We claim there are two paths to scaling the plug-load meter to a space- and power-efficient design point without sacrificing measurement fidelity: (i) by duty cycling in some cases and (ii) by using an off-line switching regulator that is not capacitor-fed. To evaluate our claims, we design a modular power meter architecture, implement several design alternatives, and evaluate their size, efficiency, and cost. We show that an energy-efficient, cubic-inch-sized wireless power meter is viable. Our design is 10x smaller, draws 3x less standby power, and still offers equivalent measurement fidelity as the best previously reported results.

Recent advances in RF ranging techniques have shown superior performance in difficult RF environments but either lack the practicality of a real-time implementation or lack flexibility through purpose-built hardware. We present SR2, a super-resolution ranging platform developed for the software radio environment. SR2 achieves low resource utilization and hardware complexity requirements through a coherent RF design, making SR2 realizable on commodity software radios.

We present μSDR, a compact, inexpensive, and battery-powered software-defined radio (SDR) platform built on a single-chip, flash-based FPGA fabric and ARM Cortex-M3 processor, enabling lower power and tighter hardware/software integration than prior commodity SDR platforms. Our architecture, unlike prior designs, is well-suited to hand-held or battery-operated systems and also supports cleaner partitioning since hardware can be easily mapped into the software addresses space, and vice versa. Building on this flexibility, we show how highly time-critical MAC protocols can be implemented on this platform and deployed using just AAA batteries.

In this work we expose the challenges of interfacing both conventional and new systems with an extremely resource constrained platform. We find that even when attempts are made to utilize an industry standard protocol (I2C), necessary protocol modifications for ultra-low power design means that interfacing remains non-trivial.

Smartphone peripherals like the Square card reader, RedEye mini, and HiJack platform suggest a growing interest in using the headset port for more than just headsets. However, these peripherals only support sporadic activities in an efficient manner. Continuous sensing applications - like monitoring EKG signals - is possible but remains too inefficient for many realistic usage scenarios. We present AudioDAQ, a new sensor data acquisition platform. In contrast with prior work, AudioDAQ requires no hardware or software modifications on the phone, uses significantly less power, and allows continuous data capture over extended periods of time. The design is efficient because we draw all necessary power from the microphone bias voltage, and it is general because this voltage is present in every headset port. Data is modulated within the audible range, captured with the built-in voice recording app, and sent to a server for processing and storage. We show the viability of this approach by demonstrating an EKG monitor that can capture data continuously for hours.

We study the problem of augmenting battery-powered sensornet trees with energy-harvesting leaf nodes. Our results show that leaf nodes that are smaller in size than today's typical battery-powered sensors can harvest enough energy from ambient sources to acquire and transmit sensor readings every minute, even under poor lighting conditions. However, achieving this functionality, especially as leaf nodes scale in size, requires new platforms, protocols, and programming. Platforms must be designed around low-leakage operation, offer a richer power supply control interface for system software, and employ an unconventional energy storage hierarchy. Protocols must not only be low-power, but they must also become low-energy, which affects initial and ongoing synchronization, and periodic communications. Systems programming, and especially bootup and communications, must become low-latency, by eliminating conservative timeouts and startup dependencies, and embracing high-concurrency. Applying these principles, we show that robust, indoor, perpetual sensing is viable using off-the-shelf technology.

Most modern software-defined radios are large, expensive, and power-hungry, and this diminishes their utility in low-power, size-constrained settings like sensor networks and mobile computing. We explore the viability of scaling down the software radio in size, cost, and power, and show that an index card-sized, sub-$150, 'AA' battery-powered system is possible using off-the-shelf components. Key to our approach is that we leverage an integrated, reconfigurable, flash-based FPGA with a hard ARM Cortex-M3 microprocessor which simultaneously enables lower power and tighter hardware/software integration than prior designs. This architecture allows us to implement timing-critical MAC protocols and validate the speculated performance of several recent MAC/PHY primitives and protocols including Backcast, A-MAC, and Glossy using an IEEE 802.15.4-compliant radio implementation that interoperates with commercial radios. The work also identifies several enhancements in the underlying hardware components that could improve power, performance, and flexibility.

We present AudioDAQ, a new platform for continuous data acquisition using the headset port of a mobile phone. AudioDAQ differs from existing phone peripheral interfaces by drawing all necessary power from the microphone bias voltage, encoding all data as analog audio, and leveraging the phone's built-in voice memo application (or a custom application) for continuous data collection. These properties make the AudioDAQ design more universal, so it works across a broad range of phones including sophisticated smart phones and simpler feature phones, enables simple analog peripherals without requiring a microcontroller, requires no hardware or software modifications on the phone itself, uses significantly less power than prior approaches, and allows continuous data capture over an extended period of time. The AudioDAQ design is efficient because it draws all necessary power from the microphone bias voltage, and it is general because this voltage and a voice memo application are present on most mobile phones in use today. We show the viability of our architecture by evaluating an end-to-end system that can capture EKG signals continuously for hours and send the data to the cloud for storage, processing, and visualization.

We present A-MAC, a receiver-initiated link layer for low-power wireless networks that supports several services under a unified architecture, and does so more efficiently and scalably than prior approaches. A-MAC's versatility stems from layering unicast, broadcast, wakeup, pollcast, and discovery above a single, flexible synchronization primitive. A-MAC's efficiency stems from optimizing this primitive and with it the most consequential decision that a low-power link makes: whether to stay awake or go to sleep after probing the channel. Today's receiver-initiated protocols require more time and energy to make this decision, and they exhibit worse judgment as well, leading to many false positives and negatives, and lower packet delivery ratios. A-MAC begins to make this decision quickly, and decides more conclusively and correctly in both the negative and affirmative. A-MAC's scalability comes from reserving one channel for the initial handshake and different channels for data transfer. Our results show that: (i) a unified implementation is possible; (ii) A-MAC's idle listening power increases by just 1.12× under interference, compared to 17.3× for LPL and 54.7× for RI-MAC; (iii) A-MAC offers high single-hop delivery ratios; (iv) network wakeup is faster and more channel efficient than LPL; and (v) collection routing performance exceeds the state-of-the-art.

We present AutoWitness, a system to deter, detect, and track personal property theft, improve historically dismal stolen property recovery rates, and disrupt stolen property distribution networks. A property owner embeds a small tag inside the asset to be protected, where the tag lies dormant until it detects vehicular movement. Once moved, the tag uses inertial sensor-based dead reckoning to estimate position changes, but to reduce integration errors, the relative position is reset whenever the sensors indicate the vehicle has stopped. The sequence of movements, stops, and turns are logged in compact form and eventually transferred to a server using a cellular modem after both sufficient time has passed (to avoid detection) and RF power is detectable (hinting cellular access may be available). Eventually, the trajectory data are sent to a server which attempts to match a path to the observations. The algorithm uses a Hidden Markov Model of city streets and Viterbi decoding to estimate the most likely path. The proposed design leverages low-power radios and inertial sensors, is immune to intransit cloaking, and supports post hoc path reconstruction. Our prototype demonstrates technical viability of the design; the volume market forces driving machine-to-machine communications will soon make the design economically viable.

We present a battery-free sensornet edge node design that operates on ambient indoor lighting, drawing just microwatts, but still delivers readings several times per minute.

Modern software-defined radios are large, expensive, and power-hungry devices and this, we argue, hampers their more widespread deployment and use, particularly in low-power, size-constrained application settings like mobile phones and sensor networks. To rectify this problem, we propose to put the software-defined radio on a diet by redesigning it around just two core chips -- an integrated RF transceiver and a Flash-based, mixed-signal FPGA. Modern transceivers integrate almost all RF front-end functions while emerging FPGAs integrate nearly all of required signal conditioning and processing functions. And, unlike conventional FPGAs, Flash-based FPGAs offer sleep mode power draws measured in the microamps and startup times measured in the microseconds, both of which are critical for low-power operation. If our platform architecture vision is realized, it will be possible to hold a software-defined radio in the palm of one's hand, build it for $100, and power it for days using the energy in a typical mobile phone battery. This will make software radios deployable in high densities and broadly accessible for research and education.

We present Virtual High-resolution Time (VHT), a power-proportional time-keeping service that offers a baseline power draw of a low-speed clock (e.g. 32 kHz crystal), but provides the time resolution that only a higher frequency clock could offer (e.g. 8 MHz crystal), and scales essentially linearly with access (i.e. the "reading" and "writing" of the clock). We achieve this performance by revisiting a basic assumption in the design of time-keeping systems -- that to achieve a given time-stamping resolution, a free-running timebase of equivalent frequency is needed. We show that this assumption is false and argue that the dependence is not on usage (i.e. whether on or off) but rather on access (i.e. reading and writing). Therefore, it is possible to duty cycle the free-running timebase itself, and augment it with a lower-frequency, temperature-compensated one, which achieves comparable resolution, at a fraction of the power, for typical workloads. The key technical challenge lies in duty cycling the fast clock and synchronizing the fast and slow clocks. To assess the viability of the approach, we explore how VHT could be implemented on several different platform architectures, and to study the power/performance tradeoff, we characterize VHT on one particular architecture in detail. Our results show power-proportional operation with a 10x improvement in average power and a synchronization accuracy exceeding 1 μs at duty cycles below 0.1%.

We introduce the Egs mote platform based on the Cortex M3 microcontroller that focuses on medical sensing applications. Egs uses an Atmel SAM3U microcontroller that runs up to 96 MHz and has up to 52 KB of RAM and 256 KB of Flash. Egs combines this microcontroller with two radios (802.15.4 and Bluetooth), external flash, on board sensors, and a LCD touchscreen to enable a rich set of wireless sensing applications.

We present A-MAC, a receiver-initiated link layer for low-power wireless networks that supports several services under a unified architecture, and does so more efficiently and scalably than prior approaches. A-MAC's versatility stems from layering unicast, broadcast, wakeup, pollcast, and discovery above a single, flexible synchronization primitive. A-MAC's efficiency stems from optimizing this primitive and with it the most consequential decision that a low-power link makes: whether to stay awake or go to sleep after probing the channel. Today's receiver-initiated protocols require more time and energy to make this decision, and they exhibit worse judgment as well, leading to many false positives and negatives, and lower packet delivery ratios. A-MAC begins to make this decision quickly, and decides more conclusively and correctly in both the negative and affirmative. A-MAC's scalability comes from reserving one channel for the initial handshake and different channels for data transfer. Our results show that: (i) a unified implementation is possible; (ii) A-MAC's idle listening power increases by just 1.12x under interference, compared to 17.3x for LPL and 54.7x for RI-MAC; (iii) A-MAC offers high single-hop delivery ratios, even with multiple contending senders; (iv) network wakeup is faster and far more channel efficient than LPL; and (v) collection routing performance exceeds the state-of-the-art.

We present AutoWitness, a system to deter, detect, and track personal property theft, improve historically dismal stolen property recovery rates, and disrupt stolen property distribution networks. A property owner embeds a small tag inside the asset to be protected, where the tag lies dormant until it detects vehicular movement. Once moved, the tag uses inertial sensor-based dead reckoning to estimate position changes, but to reduce integration errors, the relative position is reset whenever the sensors indicate the vehicle has stopped. The sequence of movements, stops, and turns are logged in compact form and eventually transferred to a server using a cellular modem after both sufficient time has passed (to avoid detection) and RF power is detectable (hinting cellular access may be available). Eventually, the trajectory data are sent to a server which attempts to match a path to the observations. The algorithm uses a Hidden Markov Model of city streets and Viterbi decoding to estimate the most likely path. The proposed design leverages low-power radios and inertial sensors, is immune to intransit cloaking, and supports post hoc path reconstruction. Our prototype demonstrates technical viability of the design; the volume market forces driving machine-to-machine communications will soon make the design economically viable.

The mobile phone is the most pervasive personal communications and computing platform ever created and yet, among its various analog interfaces, only one is open, standardized, and widely accessible: the headset port. In this demo, we augmente the mobile phone with a range of phone-powered peripherals. We show that the mobile phone headset port can be used to efficiently power external peripherals and communicate with them, enabling many new phone-centric applications. But, why use the headset port at all? One reason is that it is an open, simple, and ubiquitous interface with documented electrical and mechanical specifications. Perhaps even more important, the headset interface is backward-and forward-compatible with most mobile phones in use today.

We present our experiences evaluating the power-performance tradeoffs of a sensornet network protocol on a power-aware testbed. We characterize the power draw of the entire network while running the Collection Tree Protocol (CTP), as a function of low-power-listening interval. We find that message transmission counts are poor predictors for energy consumption on the CC2420 radio, that CTP routinely creates energy hotspots in the routing tree, and that conclusions based on protocol evaluation performed without low-power listening enabled provide little insight about the same protocol performance using low-power listening.

To achieve more accurate global time synchronization, this paper argues for decoupling the clock distribution network from the routing tree in a multihop wireless network. We find that both flooding and routing-integrated time synchronization rapidly propagate node-level errors (typically due to temperature fluctuations) across the network. Therefore, we propose that a node chooses synchronization neighbors that offer the greatest frequency stability. We propose two methods to estimate a neighbor’s stability. The first approach selects the neighbor whose Frequency Error Variance, or simply FEV, is smallest with respect to the local clock. The second approach selects the neighbor that reports the lowest FEV relative to its synchronization parent. We also propose the node-level time-variance FEV as an additive metric for selecting more stable clock trees than either naïve flooding or routing-integrated time synchronization can provide. We incorporate these techniques into FTSP, a widelyused time synchronization protocol, and show that the mean error in global time significantly improved (by a factor of five) when some nodes are warmed and others are not.

We present empirical measurements of the packet delivery performance of the latest sensor platforms: Micaz and Telos motes. In this article, we present observations that have implications to a set of common assumptions protocol designers make while designing sensornet protocols—specifically—the MAC and network layer protocols. We first distill these common assumptions in to a conceptual model and show how our observations support or dispute these assumptions. We also present case studies of protocols that do not make these assumptions. Understanding the implications of these observations to the conceptual model can improve future protocol designs.

Poor air quality is a global health issue, causing serious problems like asthma, cancer, and heart disease around the world. Earlier this decade, the World Health Organization estimated that three million people die each year from the effects of air pollution [6]. Unfortunately, while variations in air quality are significant, today's air quality monitors are very sparsely deployed. To address this visibility gap, the Common Sense project is developing participatory sensing systems that allow individuals to measure their personal exposure, groups to aggregate their members' exposure, and activists to mobilize grassroots community action.

We describe the design, deployment, and experience with a wireless sensor network for high-fidelity monitoring of electrical usage in buildings. A network of 38 mote-class AC meters, 6 light sensors, and 1 vibration sensor is used to determine and audit the energy envelope of an active laboratory. Classic WSN issues of coverage, aggregation, sampling, and inference are shown to appear in a novel form in this context. The fundamental structuring principle is the underlying load tree, and a variety of techniques are described to disambiguate loads within this structure. Utilizing contextual metadata, this information is recomposed in terms of its spatial, functional, and individual projections. This suggests a path to broad use of WSN technology in energy and environmental domains.

The AutoWitness project aims to deter, detect, and track theft of everyday objects using a combination of ultra low-power mobile tags and a wide-area network of static anchors. Key research challenges include dramatically driving down the cost and size of tags and increasing their lifetime, discriminating between normal activities and theft using motion detection and classification algorithms, reconstructing getaway trajectories from sparse anchor rendezvous, and ensuring sufficient coverage and connectivity in a sparse, wide-area network of anchors. The demonstration will show AutoWitness in operation including motion detection, classifying theft signatures, and tracking the trajectories of "stolen" objects near the conference venue.

We present Epic, an open mote platform for application-driven design. Sensornet platforms, like most embedded systems, are tightly coupled to their applications. This coupling can make it difficult for general-purpose platforms to address application-specific needs, forcing platform designers to repeatedly reimplement functionality. Inspired by the hierarchical nature of software and integrated circuit design, we propose sensornet platforms be composed hierarchically from a family of modular components. This approach makes platform development accessible to a much wider community; developers do not need to be analog, sensor, or radio frequency experts, and can instead reuse components that encapsulate the needed functionality.

We present Disco, an asynchronous neighbor discovery and rendezvous protocol that allows two or more nodes operating their radios at low duty cycles (e.g. 1%) to discover and communicate with each other during opportunistic encounters and without any prior synchronization information.

We present iCount, a new energy meter design. For many systems that have a built-in switching regulator, adding a single wire between the regulator and the microcontroller enables real-time energy metering. iCount measures energy usage by counting the switching cycles of the regulator. We show that the relationship between load current and switching frequency is quite linear and demonstrate that this simple design can be applied to a variety of regulators. Our particular implementation exhibits a maximum error of less than plusmn20% over five decades of current draw, a resolution exceeding 1 muJ, a read latency of 15 mus, and a power overhead that ranges from 1% when the node is in standby to 0.01 % when the node is active, for a typical workload. The basic iCount design requires only a pulse frequency modulated switching regulator and a microcontroller with an externally-clocked counter.

We present Disco, an asynchronous neighbor discovery and rendezvous protocol that allows two or more nodes to operate their radios at low duty cycles (e.g. 1%) and yet still discover and communicate with one another during infrequent, opportunistic encounters without requiring any prior synchronization information. The key challenge is to operate the radio at a low duty cycle but still ensure that discovery is fast, reliable, and predictable over a range of operating conditions. Disco nodes pick a pair of prime numbers such that the sum of their reciprocals is equal to the desired radio duty cycle. Each node increments a local counter with a globallyfixed period. If a node's local counter value is divisible by either of its primes, then the node turns on its radio for one period. This protocol ensures that two nodes will have some overlapping radio on-time within a bounded number of periods, even if nodes independently set their own duty cycle. Once a neighbor is discovered, and its wakeup schedule known, rendezvous is just a matter of being awake during the neighbor's next wakeup period,for synchronous rendezvous, or during an overlapping wake period, for asynchronous rendezvous.

We present a building block approach to hardware platform design based on a decade of collective experience in this area, arriving at an architecture in which general-purpose modules that require expertise to de sign and incorporate commonly-used functionality are integrated with application-specific carriers that satisfy the unique sensing, power supply, and mechanical constraints of an application. Of course, modules are widespread, but our focus is far less on the performance of any individual module and far more on an overall architecture that supports the prototype, pilot, and production stages of design, and preserves the artifacts and learnings accumulated along the way.

A home where every major appliance can be monitored for energy consumption and individually controlled wirelessly has long been a dream of gadgeteers and the green-conscious alike. Research has shown that real-time, per-appliance electricity usage feedback can induce behavior changes that lead to 10% to 20% reduction in usage [2].

We present SPOT, a scalable power observation tool that enables in situ measurement of nodal power and energy over a dynamic range exceeding four decades or a temporal resolution of microseconds. Using SPOT, every node in a sensor network can now be instrumented, providing unparalleled visibility into the dynamic power profile of applications and system software. Power metering at every node enables previously impossible empirical evaluation of low power designs at scale. The SPOT architecture and design meet challenges unique to wireless sensor networks and other low power systems, such as orders of magnitude difference in current draws between sleep and active states, short-duration power spikes during periods of brief activity, and the need for minimum perturbation of the system under observation.

We present Flush, a reliable, high goodput bulk data transport protocol for wireless sensor networks. Flush provides end-to-end reliability, reduces transfer time, and adapts to time-varying network conditions. It achieves these properties using end-to-end acknowledgments, implicit snooping of control information, and a rate-control algorithm that operates at each hop along a flow. Using several real network topologies, we show that Flush closely tracks or exceeds the maximum goodput achievable by a hand-tuned but fixed rate for each hop over a wide range of path lengths and varying network conditions. Flush is scalable; its effective bandwidth over a 48-hop wireless network is approximately one-third of the rate achievable over one hop. The design of Flush is simplified by assuming that different flows do not interfere with each other, a reasonable restriction for many sensornet applications that collect bulk data in a coordinated fashion, like structural health monitoring, volcanic activity monitoring, or protocol evaluation. We collected all of the performance data presented in this paper using Flush itself.

Ultrawideband radar-enabled wireless sensor networks have the potential to address key detection and classification requirements common to many surveillance and tracking applications. However, traditional radar signal processing techniques are mismatched with the limited computational and storage resources available on typical sensor nodes. The mismatch is exacerbated in noisy, cluttered environments or when the signals have corrupted spectra. To explore the compatibility of ultrawideband radar and mote-class sensor nodes, we designed and built a new platform called the Radar Mote. An early prototype of this platform was used to detect, classify, and track people and vehicles moving through an outdoor sensor network deployment. This paper describes the sensor's theory of operation, discusses the design and implementation of the Radar Mote, and presents sample signal waveforms of people, vehicles, noise, and clutter. We demonstrate that radar sensors can be successfully integrated with mote-class devices and imbue them with an extraordinarily useful sensing modality.

A number of multi-hop, wireless, network programming systems have emerged for sensor network retasking but none of these systems support a cryptographically-strong, public-key-based system for source authentication and integrity verification. The traditional technique for authenticating a program binary, namely a digital signature of the program hash, is poorly suited to resource-contrained sensor nodes. Our solution to the secure programming problem leverages authenticated streams, is consistent with the limited resources of a typical sensor node, and can be used to secure existing network programming systems. Under our scheme, a program binary consists of several code and data segments that are mapped to a series of messages for transmission over the network. An advertisement, consisting of the program name, version number, and a hash of the very first message, is digitally signed and transmitted first. The advertisement authenticates the first message, which in turn contains a hash of the second message. Similarly, the second message contains a hash of the third message, and so on, binding each message to the one logically preceding it in the series through the hash chain. We augmented the Deluge network programming system with our protocol and evaluated the resulting system performance.

We present the philosophy, design, and initial evaluation of the Trio Testbed, a new outdoor sensor network deployment that consists of 557 solar-powered motes, seven gateway nodes, and a root server. The testbed covers an area of approximately 50,000 square meters and was in continuous operation during the last four months of 2005. This new testbed in one of the largest solar-powered outdoor sensor networks ever constructed and it offers a unique platform on which both systems and application software can be tested safely at scale. The testbed is based on Trio, a new mote platform that provides sustainable operation, enables efficient in situ interaction, and supports fail-safe programming. The motivation behind this testbed was to evaluate robust multi-target tracking algorithms at scale. However, using the testbed has stressed the system software, networking protocols, and management tools in ways that have exposed subtle but serious weaknesses that were never discovered using indoor testbeds or smaller deployments. We have been iteratively improving our support software, with the eventual aim of creating a stable hardware-software platform for sustainable, scalable, and flexible testbed deployments.

A main challenge with developing applications for wireless embedded systems is the lack of visibility and control during execution of an application. In this paper, we present a tool suite called Marionette that provides the ability to call functions and to read or write variables on pre-compiled, embedded programs at run-time, without requiring the programmer to add any special code to the application. This rich interface facilitates interactive development and debugging at minimal cost to the node.

We present empirical measurements of the packet delivery performance of the Telos and MicaZ sensor platforms. At a high level, their behavior is similar to that of earlier platforms. They exhibit a reception "grey region," and temporal variations in packet loss. Looking more deeply, however, there are subtle differences, and looking deeper still, the patterns behind these complexities become clear. Environmental noise (802.11b) has high spatial correlation. Packet loss occurs when a receiver operating near its noise floor experiences a small decrease in received signal strength, rather than an increase in environmental noise. These variations cause the reception "grey region." Packet losses are highly correlated over short time periods, but are independent over longer periods. Based on these findings, current practices could be easily changed that would greatly improve efficiency and performance.

The operation of wireless sensor networks is fundamentally constrained by available energy sources. The underlying hardware determines the power draw of each possible mode of operation. System software attempts maximize the use of the lowest possible modes of each of the subsystems. This tutorial paper describes the system software techniques used at several levels. At the application sensing level, this includes duty-cycling, sensor hierarchy, and aggregation. At the communication level, it includes low-power listening, communication scheduling, piggybacking, post-hoc synchronization, and power-aware routing. At the node OS level, it includes event driven execution with split-phase operation and cooperative power management interfaces. At the lowest level, it includes management of primary and secondary energy storage devices coupled with intelligent charge transfer scheduling. All of these aspects must be integrated in a systematic software framework.

We present the design of the extreme scale mote, a new sensor network platform for reliably detecting and classifying, and quickly reporting, rare, random, and ephemeral events in a large-scale, long-lived, and ret askable manner. This new mote was designed for the ExScal project which seeks to demonstrate a 10,000 node network capable of discriminating civilians, soldiers and vehicles, spread out over a 10 km/sup 2/ area, with node lifetimes approaching 1,000 hours of continuous operation on two AA alkaline batteries. This application posed unique functional, usability, scalability, and robustness requirements which could not be met with existing hardware, and therefore motivated the design of a new platform. The detection and classification requirements are met using infrared, magnetic, and acoustic sensors. The infrared and acoustic sensors are designed for low-power continuous operation and include asynchronous processor wakeup circuitry. The usability and scalability requirements are met by minimizing the frequency and cost of human-in-the-loop operations during node deployment, activation, and verification through improvements in the user interface, packaging, and configurability of the platform. Recoverable retasking is addressed by using a grenade timer that periodically forces a system reset. The key contributions of this work are a specific design point and general design methods for building sensor network platforms to detect exceptional events.

Project ExScal (for extreme scale) fielded a 1000+ node wireless sensor network and a 200+ node peer-to-peer ad hoc network of 802.11 devices in a 13km by 300m remote area in Florida, USA during December 2004. In comparison with previous deployments, the ExScal application is relatively complex and its networks are the largest ones of either type fielded to date. In this paper, we overview the key requirements of ExScal, the corresponding design of the hardware/software platform and application, and some results of our experiments.