Labeling training data is one of the most costly bottlenecks in developing machine learning-based applications. We present a first-of-its-kind study showing how existing knowledge resources from across an organization can be used as weak supervision in order to bring development time and cost down by an order of magnitude, and introduce Snorkel DryBell, a new weak supervision management system for this setting. Snorkel DryBell builds on the Snorkel framework, extending it in three critical aspects: flexible, template-based ingestion of diverse organizational knowledge, cross-feature production serving, and scalable, sampling-free execution. On three classification tasks at Google, we find that Snorkel DryBell creates classifiers of comparable quality to ones trained with tens of thousands of hand-labeled examples, converts non-servable organizational resources to servable models for an average 52% performance improvement, and executes over millions of data points in tens of minutes.

Indoor localization using mobile sensors has gained momentum lately. Most of the current systems rely on an extensive calibration step to achieve high accuracy. We propose SemanticSLAM, a novel unsupervised indoor localization scheme that bypasses the need for war-driving. SemanticSLAM leverages the idea that certain locations in an indoor environment have a unique signature on one or more phone sensors. Climbing stairs, for example, has a distinct pattern on the phone's accelerometer; a specific spot may experience an unusual magnetic interference while another may have a unique set of Wi-Fi access points covering it. SemanticSLAM uses these unique points in the environment as landmarks and combines them with dead-reckoning in a new Simultaneous Localization And Mapping (SLAM) framework to reduce both the localization error and convergence time. In particular, the phone inertial sensors are used to keep track of the user's path, while the observed landmarks are used to compensate for the accumulation of error in a unified probabilistic framework. Evaluation in two testbeds on Android phones shows that the system can achieve 0.53 meters human median localization errors. In addition, the system can detect the location of landmarks with 0.83 meters median error. This is 62 percent better than a system that does not use SLAM. Moreover, SemanticSLAM has a 33 percent lower convergence time compared to the same systems. This highlights the promise of SemanticSLAM as an unconventional approach for indoor localization.

We present the results, experiences and lessons learned from comparing a diverse set of technical approaches to indoor localization during the 2014 Microsoft Indoor Localization Competition. 22 different solutions to indoor localization from different teams around the world were put to test in the same unfamiliar space over the course of 2 days, allowing us to directly compare the accuracy and overhead of various technologies. In this paper, we provide a detailed analysis of the evaluation study's results, discuss the current state-of-the-art in indoor localization, and highlight the areas that, based on our experience from organizing this event, need to be improved to enable the adoption of indoor location services.

This paper demonstrates that it is possible to leverage WiFi signals from commodity mobile devices to enable hands-free drawing in the air. While prior solutions require the user to hold a wireless transmitter, or require custom wireless hardware, or can only determine a pre-defined set of hand gestures, this paper introduces WiDraw, the first hand motion tracking system using commodity WiFi cards, and without any user wearables. WiDraw harnesses the Angle-of-Arrival values of incoming wireless signals at the mobile device to track the user's hand trajectory. We utilize the intuition that whenever the user's hand occludes a signal coming from a certain direction, the signal strength of the angle representing the same direction will experience a drop. Our software prototype using commodity wireless cards can track the user's hand with a median error lower than 5 cm. We use WiDraw to implement an in-air handwriting application that allows the user to draw letters, words, and sentences, and achieves a mean word recognition accuracy of 91%.

WiFi based indoor positioning has recently gained more attention due to the advent of the IEEE 802.11v standard, requirements by the FCC for E911 calls, and increased interest in location-based services. While there exist several indoor localization techniques, we find that these techniques tradeoff either accuracy, scalability, pervasiveness or cost -- all of which are important requirements for a truly deployable positioning solution. Wireless signal-strength based approaches suffer from location errors, whereas time-of-flight (ToF) based solutions provide good accuracy but are not scalable. Recent solutions address these issues by augmenting WiFi with either smartphone sensing or mobile crowdsourcing. However, they require tight coupling between WiFi infrastructure and a client device, or they can determine the client's location only if it is mobile. In this paper, we present CUPID2.0 which improved our previously proposed CUPID indoor positioning system to overcome these limitations. We achieve this by addressing the fundamental limitations in Time-of-Flight based localization and combining ToF with signal strength to address scalability. Experiments from $6$ cities using $40$ different mobile devices, comprising of more than $2.5$ million location fixes demonstrate feasibility. CUPID2.0 is currently under production, and we expect CUPID2.0 to ignite the wide adoption of WLAN-based positioning systems and their services.

With the proliferation of smartphones and tablets, mobile devices are soon becoming a preferred medium of Internet access in Wireless LANs (WLANs). Due to their smaller form factor, these truly mobile devices allow the users to access the wireless networks while undergoing different types of mobility, posing new challenges to wireless protocols. Current history-based protocols that maximize performance in static settings do not work well in mobile settings where wireless conditions change rapidly. Thus, today's WLANs need to be able to determine the type of the client's mobility and employ appropriate strategies in order to sustain high performance. While previous work tried to detect mobility using hints from sensors available in today's mobile devices, in this work, we demonstrate how different mobility modes can be distinguished by using physical layer information -- Channel State Information (CSI) and Time-of-Flight (ToF) -- available at commodity APs, with no modifications on the client side. In addition, we demonstrate how fine-grained mobility determination can be exploited to improve performance of client roaming, rate control, frame aggregation, and MIMO beamforming. Our testbed experiments show that our mobility classification algorithm achieves more than 92% accuracy in a variety of scenarios, and the combined throughput gain of all four mobility-aware protocols over their mobility-oblivious counterparts can be more than 100%.

Indoor navigation is typically achieved by an operational localization system among a range of location-based services it provides. However, the construction of a localization map, which is prerequisite for binding sensed observation in sensing space to individual locations in geographic space, remains a challenging task to date. In this work, we propose an indoor navigation system that alleviates the need for constructing a localization map and instead provides navigation in signal space. The main idea behind our approach is to construct a sensing map consisting of signal observations (WiFi Clusters) and connecting dead-reckoning segments obtained through mobile sensing capabilities (traces of accelerometer and digital compass reading). To this end, we design a prototype to demonstrate effective construction of such sensing map with energy-efficient sensors and crowd-sourcing, and its ability to support accurate navigation.

The continuously increasing number of smartphones and tablets allow the users to access Wireless LANs (WLANs) while undergoing different types of mobility, posing new challenges to wireless protocols. Current history-based WLAN protocols do not work well in mobile settings where wireless conditions change rapidly. Thus, today's WLANs need to be able to determine the type of the client's mobility and employ appropriate strategies in order to sustain high performance. While previous work tried to detect mobility using hints from sensors available in mobile devices, in this work, we demonstrate how different mobility modes can be distinguished by using physical layer information - Channel State Information (CSI) and Time-of-Flight (ToF) - available at commodity APs, with no modifications on the client side. Our testbed experiments show that our mobility classification algorithm achieves more than 92% accuracy in a variety of scenarios. In addition, we demonstrate how fine-grained mobility determination can be exploited to greatly improve performance of client roaming and MIMO beamforming.

This paper presents SAIL, a Single Access Point Based Indoor Localization system. Although there have been advances in WiFi-based positioning techniques, we find that existing solutions either require a dense deployment of access points (APs), manual fingerprinting, energy hungry WiFi scanning, or sophisticated AP hardware. We design SAIL using a single commodity WiFi AP to avoid these restrictions. SAIL computes the distance between the client and an AP using the propagation delay of the signal traversing between the two, combines the distance with smartphone dead-reckoning techniques, and employs geometric methods to ultimately yield the client's location using a single AP. SAIL combines physical layer (PHY) information and human motion to compute the propagation delay of the direct path by itself, eliminating the adverse effect of multipath and yielding sub-meter distance estimation accuracy. Furthermore, SAIL systematically addresses some of the common challenges towards dead-reckoning using smartphone sensors and achieves 2-5x accuracy improvements over existing techniques. We have implemented SAIL on commodity wireless APs and smartphones. Evaluation in a large-scale enterprise environment with 10 mobile users demonstrates that SAIL can capture the user's location with a mean error of 2.3m using just a single AP.

This video presents a demo of indoor localization in multiple settings. In the demo, a user walks with a smartphone and the user's location is shown on the phone's screen in real time. Our system, called Unsupervised Indoor Localization (UnLoc) utilizes the sensor data from smartphones to learn "invisible landmarks" in the environment. Example landmarks could be a unique magnetic fluctuation experienced when the phone is near a water-cooler, or a distinct gyroscope rotation when the user turns a corner. We use these indoor "landmarks" to periodically reset the user's location. To track the user between these landmarks, we use an optimized variant of dead reckoning, ultimately leading to a robust location tracking system. We call our system UnLoc, since the landmarks are generated in an unsupervised manner, requiring no manual effort or floorplan of the building. The demo describes the high level intuitions, shows UnLoc in operation, and shares experiences from running UnLoc in various real-world environments.

Cooperative packet recovery has been widely investigated in wireless networks, where corrupt copies of a packet are combined to recover the original packet. While previous work such as MRD (Multi Radio Diversity) and Soft apply combining to bits and bit-confidences, combining at the symbol level has been avoided. The reason is rooted in the prohibitive overhead of sharing raw symbol information between different APs of an enterprise WLAN. We present Epicenter that overcomes this constraint, and combines multiple copies of incorrectly received “symbols” to infer the actual transmitted symbol. Our core finding is that symbols need not be represented in full fidelity - coarse representation of symbols can preserve most of their diversity, while substantially lowering the overhead. We then develop a rate estimation algorithm that actually exploits symbol level combining. Our USRP/GNURadio testbed confirms the viability of our ideas, yielding 40% throughput gain over Soft, and 25-90% over 802.11. While the gains are modest, we believe that they are realistic, and available with minimal modifications to today's EWLAN systems.

Wireless performance depends directly on the quality of the channel. A wireless transmitter can improve its performance by estimating and transmitting on only the strongest channel, which can be of significantly higher quality than a weak channel (yielding up to 100% rate improvement). It is considered impossible to predict the quality of the unseen channels. Thus, the only way to identify the strongest channel is by probing each channel individually, incurring large over- heads. The key contribution of this paper is a discovery of previously unobserved properties of the wireless channel that makes it possible to predict the the strongest of a set of channels from the measurements collected only on a single channel. We confirm the properties through measurements and present a theoretical analysis that explains their nature. Our proposed system, CSpy, utilizes these observations to predict the strongest channel. CSpy is the first to reliably estimate the strongest channel by utilizing channel responses extracted from off-the-shelf wireless chipsets, without probing any additional channels. By tracking the strongest channel, CSpy improves performance by up to 100% in comparison to channel agnostic schemes.

Despite of several years of innovative research, indoor localization is still not mainstream. Existing techniques either employ cumbersome fingerprinting, or rely upon the deployment of additional infrastructure. Towards a solution that is easier to adopt, we propose CUPID, which is free from these restrictions, yet is comparable in accuracy. While existing WiFi based solutions are highly susceptible to indoor multipath, CUPID utilizes physical layer (PHY) information to extract the signal strength and the angle of only the direct path, successfully avoiding the effect of multipath reflections. Our main observation is that natural human mobility, when combined with PHY layer information, can help in accurately estimating the angle and distance of a mobile device from an wireless access point (AP). Real-world indoor experiments using off-the-shelf wireless chipsets confirm the feasibility of CUPID. In addition, while previous approaches rely on multiple APs, CUPID is able to localize a device when only a single AP is present. When a few more APs are available, CUPID can improve the median localization error to 2.7m, which is comparable to schemes that rely on expensive fingerprinting or additional infrastructure.

Successive interference cancellation (SIC) is a PHY capability that allows a receiver to decode packets that arrive simultaneously. While the technique is well known in communications literature, emerging software radio platforms are making practical experimentation feasible. This motivates us to study the extent of throughput gains possible with SIC from a MAC layer perspective and scenarios where such gains are worth pursuing. We find that contrary to our initial expectation, the gains are not high when the bits of interfering signals are not known a priori to the receiver. Moreover, we observe that the scope for SIC gets squeezed by the advances in bitrate adaptation. In particular, our analysis shows that interfering one-to-one transmissions benefit less from SIC than scenarios with many-to-one transmissions (such as when clients upload data to a common access point). In view of this, we develop an SIC-aware scheduling algorithm that employs client pairing and power reduction to extract the most gains from SIC. We believe that our findings will be useful guidelines for moving forward with SIC-aware protocol research.

This paper explores the viability of precise indoor localization using physical layer information in WiFi systems. We find evidence that channel responses from multiple OFDM subcarriers can be a promising location signature. While these signatures certainly vary over time and environmental mobility, we notice that their core structure preserves certain properties that are amenable to localization. We attempt to harness these opportunities through a functional system called PinLoc, implemented on off-the-shelf Intel 5300 cards. We evaluate the system in a busy engineering building, a crowded student center, a cafeteria, and at the Duke University museum, and demonstrate localization accuracies in the granularity of 1m x 1m boxes, called "spots". Results from 100 spots show that PinLoc is able to localize users to the correct spot with 89% mean accuracy, while incurring less than 6% false positives. We believe this is an important step forward, compared to the best indoor localization schemes of today, such as Horus.

We propose UnLoc, an unsupervised indoor localization scheme that bypasses the need for war-driving. Our key observation is that certain locations in an indoor environment present identifiable signatures on one or more sensing dimensions. An elevator, for instance, imposes a distinct pattern on a smartphone's accelerometer; a corridor-corner may overhear a unique set of WiFi access points; a specific spot may experience an unusual magnetic fluctuation. We hypothesize that these kind of signatures naturally exist in the environment, and can be envisioned as internal landmarks of a building. Mobile devices that "sense" these landmarks can recalibrate their locations, while dead-reckoning schemes can track them between landmarks. Results from 3 different indoor settings, including a shopping mall, demonstrate median location errors of 1:69m. War-driving is not necessary, neither are floorplans the system simultaneously computes the locations of users and landmarks, in a manner that they converge reasonably quickly. We believe this is an unconventional approach to indoor localization, holding promise for real-world deployment.

We propose UnLoc [1], an unsupervised indoor localization scheme that bypasses the need for war-driving. Our key observation is that certain locations in an indoor environment present an identifiable signature on one or more sensing dimensions. An elevator, for instance, imposes a distinct pattern on a smartphone's accelerometer; a specific spot may experience an unusual magnetic fluctuation. This form of urban sensing and activity recognition has already been demonstrated in literature [2, 3], but not yet applied in pure localization applications. We hypothesize that these kind of signatures naturally exist in the environment and can be envisioned as internal landmarks of a building. Mobile devices that "sense" these landmarks can recalibrate their locations, while dead-reckoning schemes can track them between landmarks. Neither war-driving nor floorplans are necessary - the system simultaneously computes the locations of users and landmarks, in a manner so that they converge reasonably quickly. We believe this is an unconventional approach to indoor localization, holding promise for real-world deployment.

Modern wireless interfaces support a physical-layer capability called Message in Message (MIM). Briefly, MIM allows a receiver to disengage from an ongoing reception and engage onto a stronger incoming signal. Links that otherwise conflict with each other can be made concurrent with MIM. However, the concurrency is not immediate and can be achieved only if conflicting links begin transmission in a specific order. The importance of link order is new in wireless research, motivating MIM-aware revisions to link-scheduling protocols. This paper identifies the opportunity in MIM-aware reordering, characterizes the optimal improvement in throughput, and designs a link-layer protocol for enterprise wireless LANs to achieve it. Testbed and simulation results confirm the performance gains of the proposed system.

A wireless transmitter learns of a packet loss and infers collision only after completing the entire transmission. If the transmitter could detect the collision early [such as with carrier sense multiple access with collision detection (CSMA/CD) in wired networks], it could immediately abort its transmission, freeing the channel for useful communication. There are two main hurdles to realize CSMA/CD in wireless networks. First, a wireless transmitter cannot simultaneously transmit and listen for a collision. Second, any channel activity around the transmitter may not be an indicator of collision at the receiver. This paper attempts to approximate CSMA/CD in wireless networks with a novel scheme called CSMA/CN (collision notification). Under CSMA/CN, the receiver uses PHY-layer information to detect a collision and immediately notifies the transmitter. The collision notification consists of a unique signature, sent on the same channel as the data. The transmitter employs a listener antenna and performs signature correlation to discern this notification. Once discerned, the transmitter immediately aborts the transmission. We show that the notification signature can be reliably detected at the listener antenna, even in the presence of a strong self-interference from the transmit antenna. A prototype testbed of 10 USRP/GNU Radios demonstrates the feasibility and effectiveness of CSMA/CN.

This paper shows the viability of precise indoor localization using physical layer information in WiFi systems. We find that channel frequency responses across multiple OFDM sub-carriers can be suitably aggregated into a location fingerprint. While these fingerprints vary over time and environmental mobility, we notice that their core structure preserves certain properties that are amenable to localization. We demonstrate these ideas through a functional prototype, implemented on off-the-shelf Intel 5300 cards (that export per-subcarrier information to the driver). We evaluate the prototype using the existing APs inside a busy building, a cafeteria, and a museum, and demonstrate localization accuracies in the granularity of 1m × 1m boxes, called spots. Results show that our system, PinLoc, is able to localize users to a spot with 90% mean accuracy, while incurring less than 6% false positives. We believe this holds promise towards an important development in indoor localization.

Conventional WiFi networks perform channel contention in time domain. This is known to be wasteful because the channel is forced to remain idle while all contending nodes are backing off for multiple time slots. This paper proposes to break away from convention and recreate the backing off operation in the frequency domain. Our basic idea leverages the observation that OFDM subcarriers can be treated as integer numbers. Thus, instead of picking a random backoff duration in time, a contending node can signal on a randomly chosen subcarrier. By employing a second antenna to listen to all the subcarriers, each node can determine whether its chosen integer (or subcarrier) is the smallest among all others. In fact, each node can even determine the rank of its chosen subcarrier, enabling the feasibility of scheduled transmissions after every round of contention. We develop these ideas into a Back2F protocol that migrates WiFi backoff to the frequency domain. Experiments on a prototype of 10 USRPs confirm feasibility, along with consistent throughput gains over 802.11. at high bit rates. Trace based simulations affirm scalability to larger, real-world network topologies.

This poster shows the viability of precise indoor localization using physical layer information in WiFi systems. We find that channel frequency responses across multiple OFDM subcarriers can be suitably aggregated into a location signature. While these signatures vary over time and environmental mobility, we notice that their core structure preserves certain properties that are amenable to localization. We demonstrate these ideas through a functional system, implemented on off-the-shelf Intel 5300 cards (that are designed to export per-subcarrier information to the driver). We evaluate the system in a busy engineering building, a cafeteria, and the university museum, and demonstrate localization accuracies in the granularity of 1m x 1m boxes, called spots. Results show that our system, PinLoc, is able to localize users to a spot with 90% mean accuracy, while incurring less than 6% false positives. We believe this is an important step forward, compared to the best indoor localization schemes of today.

Conventional WiFi networks perform channel contention in time domain. This is known to be wasteful because the channel is forced to remain idle, while all contending nodes are backing off for multiple time slots. This paper proposes to break away from convention and recreate the backing off operation in the frequency domain. Our basic idea is to pretend that OFDM subcarriers are integer numbers, and thereby, view today's random backoff process as equivalent to transmitting on a randomly chosen subcarrier. By employing a second antenna to listen to all the subcarriers, each node can determine whether its chosen integer (or subcarrier) is the smallest among all others. In fact, each node can even determine the rank of its chosen integer, enabling the feasibility of a TDMA-like schedule from every round of contention. We develop these ideas into a Time to Frequency (T2F) protocol and prototype it on a small testbed of 8 USRPs. Experiments confirm its feasibility, along with promising throughput gains of more than 35% at high bit rates. A fuller design and thorough evaluation of T2F is a topic of ongoing work.

Successive interference cancellation (SIC) is a physical layer capability that allows a receiver to decode packets that arrive simultaneously. While the technique is well known in communications literature, emerging software radios are making practical experimentation feasible. This motivates us to study the extent of throughput gains possible with SIC from a MAC layer perspective. Contrary to our initial expectation, we find that the gains from SIC are not easily available in many realistic situations. Moreover, we observe that the scope for SIC gets squeezed by the advances in bitrate adaptation, casting doubt on the future of SIC based protocols.

The nature of human mobility demands that mobile devices become agile to diverse operating environments. Coping with such diversity requires the device to assess its environment, and trigger appropriate responses to each of them. While existing communication subsystems rely on in-band wireless signals for context-assessment and response, we explore a lateral approach of using out-of-band sensor information. We propose a relatively novel framework that synthesizes in-band and out-of-band information, facilitating more informed communication decisions. We believe that further research in this direction could enable a new kind of device agility, deficient in today's communication systems. Since such a framework is located at the boundaries of mobile sensing and wireless communication, we call it sensor assisted wireless communication.

A wireless transmitter learns of a packet loss, infers collision, only after completing the entire transmission. If the transmitter could detect the collision early (such as with CSMA/CD in wired networks), it could immediately abort its transmission, freeing the channel for useful communication. There are two main hurdles to realize CSMA/CD in wireless networks. First, a wireless transmitter cannot simultaneously transmit and listen for a collision. Second, any channel activity around the transmitter may not be an indicator of collision at the receiver. This paper attempts to approximate CSMA/CD in wireless networks with a scheme called CSMA/CN (collision notification). Under CSMA/CN, the receiver uses PHY layer information to detect a collision and immediately notifies the transmitter. The collision notification consists of a unique signature, sent on the same channel as the data. The transmitter employs a listener antenna and performs signature correlation to discern this notification. Once discerned, the transmitter immediately aborts transmission. We show that the notification signature can be reliably detected at the listener antenna, even in the presence of a strong self-interference from the transmit antenna. A prototype testbed of 10 USRP/GNURadios demonstrates the feasibility and effectiveness of CSMA/CN

Modern wireless interfaces support a physical layer capability called Message in Message (MIM). Briefly, MIM allows a receiver to disengage from an ongoing reception, and engage onto a stronger incoming signal. Links that otherwise conflict with each other, can be made concurrent with MIM. However, the concurrency is not immediate, and can be achieved only if conflicting links begin transmission in a specific order. The importance of link order is new in wireless research, motivating MIM-aware revisions to link scheduling protocols. This paper identifies the opportunity in MIM-aware reordering, characterizes the optimal improvement in throughput, and designs a link layer protocol to achieve it. Testbed results confirm the performance gains of the proposed system.

Wireless link layer multicast is an important service primitive for emerging applications, such as live video, streaming audio, and other content telecasts. The broadcast nature of the wireless channel is amenable to multicast because a single packet transmission may be received by all clients in the multicast group. However, in view of diverse channel conditions at different clients, the rate of such a transmission is bottlenecked by the rate of the weakest client. Multicast throughput degrades severely. Attempts to increase the data rate result in lower reliability and higher unfairness. This paper utilizes smart beamforming antennas to improve multicast performance in wireless LANs. The main idea is to satisfy the stronger clients with a high-rate omnidirectional transmission, followed by high-rate directional transmission(s) to cover the weaker ones. By selecting an optimal transmission strategy (using dynamic programming), we show that the multicast throughput can be maximized while achieving a desired delivery ratio at all the clients. We use testbed measurements to verify our main assumptions. We simulate our protocol in Qualnet, and observe consistent performance improvements over a range of client topologies and time-varying channel conditions.

In this paper we propose a smart video surveillance system with real-time moving object (primarily human) detection and identification for solving the problem of excessive power consumption in a public place like shopping malls. The proposed detection system is able to handle several known constraints such as changing illumination conditions, occlusion, clutter or even irrelevant extraneous motion. It keeps track of the number of humans in the vicinity and takes measures depending on the number of people within its sight. As the number of humans in the neighbourhood increases the system takes steps to gradually increase the luminosity of the area. Again, when there are a few or no individuals in the area the brightness of the area is gradually decreased. Moreover it keeps track of the members of the working staff, who generally dresses in a fixed uniform, to save power by tuning the lighting to an adequate but not a dazzling level.