The detection and localization of radiation sources using low-cost, mobile detectors is a challenging application, necessitating new research into sensing devices and detection algorithms. While new sensing that employs small detectors for detecting γ -rays has emerged, the decreased sensitivity of the sensor makes it challenging to maintain reliability compared to larger detectors. Machine learning could be a viable method for enhancing sensitivity by classifying background radiation spectra from anomalous spectra, but this approach can struggle to identify novel radioactive sources or identify sources in dynamic background environments. To address these challenges, we propose the use of adversarial auto encoders (AAEs) for anomaly detection in radiation sensing systems. With the use of our AAE architecture, we eliminate the need for obtaining examples of radiation anomalies for training data and increase the resilience of the sensing when the background radiation is dynamic. We evaluate the system in various contexts using a custom designed detector, showing the AAE model generalizes to various locations and radiation sources. We also show a real-time field test with the detection system in both handheld and drone mounted testing. • Our model improves detection of γ -ray anomalies on cesium iodide detectors. • These improvements extend to simulated sodium iodide detectors. • We demonstrate real-time detection capabilities in both lab and field environments. • These results are competitive with other unsupervised methods.

Drones frequently communicate with devices on the ground, often carried in different ways by users. A user might hold their device near their chest while flying a drone, walk with their phone in their pocket, or walk facing away from the drone while it is tracking them. Despite the proliferation of drones and possibility of such scenarios, the prediction of user orientation and the user equipment (UE) location on or near the user's body in Drone-to-Ground (D2G) channels has not received adequate attention in literature. In scenarios where visibility is no longer available to the drone - due to adversarial attacks or harsh weather - the wireless signal can be used to detect the user's presence, their orientation, and even the location of the device on or near their body. This is the objective of this work. We study how the baseband I/Q samples, converted into spectrogram images that span a relatively short period of time, can be used to predict the on-body device location. We leverage Convolutional Neural Networks (CNNs) to classify three different use cases of holding a device operating at two different carrier frequencies (2.5 GHz and 900 MHz). Specifically, we investigate three on-body locations: near chest while facing the drone, in-pocket while facing the drone, and near chest but facing away from the drone. We show that, using only spectrogram images as input, we can predict these use cases with an average overall accuracy of 87% and 85% at 2.5 GHz and 900 MHz, respectively. We also investigate the classification performance on a dataset that belong to a hovering location that was not seen by the model, and show that the CNN model was able to correctly classify all images belonging to user orientation and 91% of the images that belong to near-chest vs in-pocket for the same orientation. Finally, we study the application of transfer learning and CNNs on classifying the on-body location at a different carrier frequency from the one on which they were trained, and show that while one use case can be correctly predicted, more complex models and hyper-parameter tuning is needed to achieve this goal. This work could be useful for building real-time deep learning models that help drones to make intelligent decisions and adapt to changes in user postures and on-body locations in air-to-ground (A2G) channels.

In drone swarms, where multiple drones collaborate closely to achieve shared objectives within constrained spatial domains, the intricacies of these interrelated actions can lead to potential issues. Despite rigorous pre-deployment planning, the inherent probability of complications persists. These compli-cations stem from onboard computational resources, hardware failures, and network communication disruptions. While the malfunction of an individual drone may seem inconsequential, it can escalate into a substantial predicament when it disrupts the seamless coordination of the entire swarm. Therefore, the need to proactively monitor drones for predictive failure analysis and the subsequent examination of failed drones to mitigate future occurrences becomes imperative. This paper introduces a comprehensive framework for systematically collecting and processing data within drone swarms. The framework gathers critical information about onboard characteristics and commu-nication metrics. These data points are subjected to advanced analysis using Complex Bayesian Networks to probabilistically uncover complex and hidden relationships between random features. The results demonstrate exceptional accuracy, with influences ranging from 99 % to 79 %, that ensures the reliability and effectiveness of the predictive capabilities in enhancing drone safety and network performance.

Understanding the behavior of phase errors between radio frequency (RF) chains in software-defined radios (SDRs) is crucial to the success of implementing many phase-sensitive applications, such as beamforming. Even if SDRs are provided the same clocking signal, initial local oscillator (LO) phase offsets across devices will inevitably be different. Despite its known effect on many wireless applications, there are only few works that experimentally discuss random phase errors in SDRs. To address this issue, we perform experiments and analyze the results of tens of experiments in an attempt to understand the nature of this phase offset. In particular, we target the USRP (Universal Standard Radio Peripheral) N310 platform as it provides up to 4 simultaneous transmit/receive chains that could be attractive for beamforming applications. We first model the system used in this study and demonstrate how phase errors can affect distributed beamforming gains. Then, we introduce our experimental setup, procedures and analysis of the results of the measured phase error. We do so first between two chains of the same/different transceiver boards within the same USRP, and then between chains of distributed USRPs that are geographically separated. We calculate the mean and standard deviation of this phase error, investigate its behavior over time, and demonstrate how the distribution of this error can vary based on whether it is measured in a centralized or a distributed fashion.

This paper explores the nexus of two emerging Internet of Things (IoT) components in precision agriculture, which requires vast amounts of agriculture fields to be monitored from air and soil for food production with efficient resource utilization. On the one hand, unmanned aerial vehicles (UAVs) have gained interest in agricultural aerial inspection due to their ubiquity and observation scale. On the other hand, agricultural IoT devices, including buried soil sensors, have gained interest in improving natural resource efficiency in crop production. In this work, the path loss and fading characteristics in wireless links between a UAV and underground (UG) nodes (Air2UG link) are studied to design a UAV altitude optimization solution. A path loss model is developed for the Air2UG link, including fading in the channel, where fading is modeled using a Rician distribution and validated using the Kolmogorov-Smirnov test. Moreover, Rician-K is found to be dependent on the UAV altitude, which is modeled with a Gaussian function with an RMSE of 0.4 − 1.3 dB. Furthermore, a novel altitude optimization solution is presented to minimize the bit error rate (BER). Results show that the lowest possible altitude does not always minimize the BER. Optimizing the altitude reduces the Air2UG link BER by as much as 8.6-fold. Likewise, altitude optimization can minimize the impacts of increasing burial depth on the BER. Our results and analysis are the first in this field and can be exploited to optimize the altitude and resources of a UAV node to communicate with the sensors embedded in the soil efficiently.

In this paper, we propose a multi-band medium access control (MAC) protocol for an infrastructure-based network with an access point (AP) that supports In-Band full-duplex (IBFD) and multiuser transmission to multi-band-enabled stations. The Multi-Band Full Duplex MAC (MB-FDMAC) protocol mainly uses the sub-6 GHz band for control-frame exchange, transmitted at the lowest rate per IEEE 802.11 standards, and uses the 60 GHz band, which has significantly higher instantaneous bandwidth, exclusively for data-frame exchange. We also propose a selection method that ensures fairness among uplink and downlink stations. Our result shows that MB-FDMAC effectively improves the spectral efficiency in the mmWave band by 324%, 234%, and 189% compared with state-of-the-art MAC protocols. In addition, MB-FDMAC significantly outperforms the combined throughput of sub-6 GHz and 60 GHz IBFD multiuser MIMO networks that operate independently by more than 85%. In addition, we study multiple network variables such as the number of stations in the network, the percentage of mmWave band stations, the size of the contention stage, and the selection method on MB-FDMAC by evaluating the change in the throughput, packet delay, and fairness among stations. Finally, we propose a method to improve the utilization of the high bandwidth of the mmWave band by incorporating time duplexing into MB-FDMAC, which we show can enhance the fairness by 12.5 % and significantly reduces packet delay by 80%.

Distributed transmit beamforming (DTBF) can allow a swarm of unmanned aerial vehicles (UAVs) to send a common message to a distant target. DTBF among N nodes can provide N2times the received power compared to a single node and can reduce interference by confining the signal in a certain direction. However, DTBF requires time, frequency, and phase synchronization. Here, we focus on the issue of phase incoherence at the distributed transmit nodes from two sources—different local oscillators (LOs) and hovering position movement—and how to counteract their impact at the receiver via local decisions, namely, rotation. To investigate how the UAV body and its rotation can affect phase coherency, we conduct controlled in-field experiments where we control the phase offset at two distributed antennas and measure the received signal level at four antenna positions on a drone for various rotation angles. We show that significant improvements can be achieved at the receiver through rotation. We also show that there exists an optimal combination of UAV rotation angle and antenna position on the drone to mitigate the effects of phase incoherence among the distributed transmitters. Finally, we demonstrate an interesting trade-off where, due to the heterogeneous nature of the UAV body, rotation angles that yield maximum beamforming gains might not result in the best average (or minimum) beamformed signal level across all possible phase errors at the distributed transmitters.

In this paper, we investigate the user selection techniques for an in-band full-duplex access point that simultaneously transmits and receives data from multiple users for the next-generation wireless local area network. Then, we propose a throughput-fairness tradeoff selection algorithm to enable the AP to maximize the throughput with a maintainable fairness level. In addition, we propose a throughput-fairness medium access control (TFMAC) based on the 802.11 standards to accommodate the requirements of the proposed selection algorithm and support legacy nodes. Our simulation results show that TFMAC improves the throughput compared to multiple state-of-the-art benchmarks while maintaining the desired fairness levels. Also, we study the interplay between the throughput, the uplink fairness, and the downlink fairness for the operation of TFMAC. Finally, we discuss the complexity of the proposed scheme.

In this paper, we investigate a mobile edge computing (MEC) system in which a set of users with intensive computation tasks, and a set of users with high downlink rate requirement, can cooperate to achieve a mutually-beneficial situation where the task completion time is reduced and the downlink users receive more information from the base station (BS). Specifically, by leveraging uplink and downlink non-orthogonal multiple access (NOMA), the user with an intensive computation task can offload its task bits to the edge cloud and the downlink user. Simultaneously this user relays information to the downlink user from the BS. We consider the joint optimization of computational resource allocation at the edge cloud, communication resource allocation, assignment among the two sets of users, the share of computation, and relay bits to minimize the overall completion time of the tasks while guaranteeing downlink users' incentive requirement. A low complexity iterative algorithm is proposed to find efficient locally optimal solutions by utilizing convex optimization, a graph theory matching algorithm, and block coordinate descent technique. Numerical results show that the proposed technique leads to a significant reduction in users' task completion time and increase in the downlink users rate.

Unmanned Aerial Vehicles (UAVs) often lack the size, weight, and power to support large antenna arrays or a large number of radio chains. Despite such limitations, emerging applications that require the use of swarms, where UAVs form a pattern and coordinate towards a common goal, must have the capability to transmit in any direction in three-dimensional (3D) space from moment to moment. In this work, we design a measurement study to evaluate the role of antenna polarization diversity on UAV systems communicating in arbitrary 3D space. To do so, we construct flight patterns where one transmitting UAV is hovering at a high altitude (80 m) and a receiving UAV hovers at 114 different positions that span 3D space at a radial distance of approximately 20 m along equally-spaced elevation and azimuth angles. To understand the role of diverse antenna polarizations, both UAVs have a horizontally-mounted antenna and a vertically-mounted antenna-each attached to a dedicated radio chain-creating four wireless channels. With this measurement campaign, we seek to understand how to optimally select an antenna orientation and quantify the gains in such selections.

A new medium access control (MAC) protocol for full-duplex multiuser multi input multi output (FD-MUMAC) is proposed. FD-MUMAC is an 802.11 infrastructure-based MAC protocol where the communication in the network is moderated by a full-duplex access point (AP) that can handle multiple simultaneous uplink and downlink streams from half-duplex users. FD-MUMAC is the first MAC protocol that jointly determines the selection of uplink and downlink users and controls the transmission rate by considering transmit and receive beamforming, channel states, and multiuser-interferences. To demonstrate the usefulness of the proposed protocol, we introduce a joint fairness selection algorithm that ensures that each user obtains a fair allocation of time to utilize the channel and balances its traffic between the uplink and downlink directions. In one specific instantiation, FD-MUMAC achieves a throughput gain of 59%, 177% and 94% compared with a single antenna FD MAC protocol and two half-duplex MU-MIMO protocols, respectively.

The next wave of drone applications is moving from repeatable, single-drone activities such as evaluating propagation environments to team-based, multi-drone objectives such as drone-based emergency services. In parallel, testbeds have sought to evaluate emerging concepts such as highly-directional and distributed wireless communications. However, there is a lack of intersection between the two works to characterize the impact of the drone body, antenna placement, swarm topologies, and multi-dimensional connectivity needs that require in-flight experimentation with a surrounding testbed infrastructure. In this work, we design a Multi-Dimensional Drone Communications Infrastructure (MuDDI) to capture complex spatial wireless channel relationships that drone links experience as applications scale from single-drone to swarm-level networks within a shared three-dimensional space. Driven by the challenges of outdoor experimentation, we identify the need for a highly-controlled indoor environment where external factors can be mitigated. To do so, we first build an open-source drone platform to provide programmable control with visibility into the internal flight control system and sensors enabling specialized coordination and accurate repeatable positioning within the isolated environment. We then design a wireless data acquisition system and integrate distributed software defined radios (SDRs) in order to inspect multi-dimensional wireless behavior from the surrounding area. We achieve and demonstrate the value of measurement perspectives from diverse altitudes and spatial locations with the same notion of time. Finally, we demonstrate how multi-dimensional models from experimental measurements can be implemented to simulate multi-drone networks on a practical scale.

In the next wave of swarm-based applications, unmanned aerial vehicles (UAVs) need to communicate with peer drones in any direction of a three-dimensional (3D) space. On a given drone and across drones, various antenna positions and orientations are possible.We know that, in free space, high levels of signal loss are expected if the transmitting and receiving antennas are cross polarized. However, increasing the reflective and scattering objects in the channel between a transmitter and receiver can cause the received polarization to become completely independent from the transmitted polarization, making the cross-polarization of antennas insignificant. Usually, these effects are studied in the context of cellular and terrestrial networks and have not been analyzed when those objects are the actual bodies of the communicating drones that can take different relative directions or move at various elevations. In this work, we show that the body of the drone can affect the received power across various antenna orientations and positions and act as a local scatterer that increases channel depolarization, reducing the cross-polarization discrimination (XPD). To investigate these effects, we perform experimentation that is staged in terms of complexity from a controlled environment of an anechoic chamber with and without drone bodies to in-field environments where drone-mounted antennas are in-flight with various orientations and relative positions with the following outcomes: (i.) drone relative direction can significantly impact the XPD values, (ii.) elevation angle is a critical factor in 3D link performance, (iii.) antenna spacing requirements are altered for co-located cross-polarized antennas, and (iv.) cross-polarized antenna setups more than double spectral efficiency. Our results can serve as a guide for accurately simulating and modeling UAV networks and drone swarms.

Cellular networks are susceptible to being severely capacity-constrained during peak traffic hours or at special events such as sports and concerts. Many other applications are emerging for LTE and 5G networks that inject machine-to-machine (M2M) communications for Internet of Things (IoT) devices that sense the environment and react to diurnal patterns observed. Both for users and devices, the high congestion levels frequently lead to numerous retransmissions and severe battery depletion. However, there are frequently social cues that could be gleaned from interactions from websites and social networks of shared interest to a particular region at a particular time. Cellular network operators have sought to address these high levels of fluctuations and traffic burstiness via the use of offloading to unlicensed bands, which may be instructed by these social cues. In this paper, we leverage shared interest information in a given area to conserve power via the use of offloading to the emerging Citizens Radio Band Service (CBRS). Our GreenLoading framework enables hierarchical data delivery to significantly reduce power consumption and includes a Broker Priority Assignment (BPA) algorithm to select data brokers for users. With the use of in-field measurements and web-based Google data across four diverse U.S. cities, we show that, on average, an order of magnitude power savings via GreenLoading over a 24-hour period and up to 35 times at peak traffic times. Finally, we consider the role that a relaxation of wait times can play in the power efficiency of a GreenLoaded network.

Precise air-to-ground propagation modeling is imperative for many unmanned aerial vehicle (UAV) applications such as search and rescue, reconnaissance, and disaster recovery. Furthermore, directionalization via MIMO-based beamforming can boost the transmission range by utilizing Channel State Information (CSI). However, the high mobility and flight conditions of drones can threaten the ability to receive accurate CSI in time to achieve such gains. In this work, we design a UAV-based software defined radio (SDR) platform and perform a measurement study to characterize the air-to-ground channel between the aerial platforms and a terrestrial user in practical scenarios such as hovering, encircling, and linear topologies. Our experiments cover multiple carrier frequencies, including cellular (900~MHz and 1800~MHz) and WiFi (5~GHz) bands. Furthermore, we address three baseline issues for deploying drone-based beamforming systems: channel reciprocity, feedback overhead, and update rate for channel estimation. Numerical results show that explicit CSI feedback can increase throughput by 123.9% over implicit feedback and the optimal update rate are similar across frequencies, underscoring the importance of drone-based beamfoming design. We additionally analyze the reciprocity error and find that the amplitude error remained steady while the phase error depends on mobility. Since our study spans many critical frequency bands, these results serve as a fundamental step towards understanding drone- based beamforming systems.

Key Performance Indicators (KPIs) are important measures of the quality of service in cellular networks. There are multiple efforts by cellular carriers and 5G standardization to leverage the KPIs to minimize drive tests (MDT) and self-organize the network for optimal performance via user feedback. Such an approach accounts for user devices in the field of their operation according to their normal usage and circumvents a number of costs (e.g., manpower, equipment) traditionally covered by the carrier, either directly or through a third party. In this paper, we build a Regional Analysis to Infer KPIs (RAIK) framework to establish a relationship between geographical data and user data using crowdsourced measurements. To do so, we use a neural network and crowdsourced data obtained by user equipment (UE) to predict the KPIs in terms of the reference signal's received power (RSRP) and path loss estimation. Since these KPIs are a function of terrain type, we provide a two-layer coverage map by overlaying a performance layer on a 3-dimensional geographical map. As a result, we can efficiently use crowdsourced data (to not overextend user bandwidth and battery) and infer KPIs in areas where measurements have not or can not be performed. For example, we show that RAIK can use only geographical information to predict the KPIs in areas that lack signal quality data with a negligible mean squared error, a seven-fold reduction in error from state-of-the-art solutions.

The last few years have seen a rapid growth in unmanned aerial vehicle (UAV) based innovations and technologies, particularly for smaller drones. The rapid response to natural disasters, high data rate access in public safety situations, and the robustness of long-haul communication relays are highly dependent on airborne communication networks. A more precise channel characterization of air-to-ground links is imperative to establish these drone-based communication networks. However, there have been very limited efforts to understand the unique propagation channels encountered in drone-based communications, especially for wideband beamforming systems. In this paper, we perform a measurement-driven study to characterize air-to-ground wireless channels between UAV platforms and terrestrial users in practical Line of Sight (LOS) and Non Line of Sight (NLOS) scenarios across a wide range of carrier frequencies, including cellular (900 MHz and 1800 MHz), and WiFi (5 GHz) frequency bands. Furthermore, we investigate the feasibility of drone-based beamforming using IEEE 802.11-like signaling. We find that the drone-to-ground path loss differences are frequency dependent and closely related to drone altitude. The drone-based beamforming system can improve throughput significantly over IEEE 802.11 SISO schemes with select carrier frequencies in both LOS and NLOS scenarios up to 73.6% and 120.1%, respectively. Since our study spans many critical frequency bands, these results serve as a fundamental step towards understanding drone-to-ground communications and impact of beamforming-based applications in future aerial networks.

Conducting in-field performance analysis for wireless carrier coverage and capacity evaluation is extremely costly, in terms of equipment, manpower, and time. Hence, there is a growing number of opportunities that exist for crowdsourcing via smart applications, firmware, and cellular standards. These facilities offer carriers feedback about user-perceived wireless channel quality. Crowdsourcing provides the ability to rapidly collect feedback with dense levels of penetration using client smartphones. However, mobile phones often fail to capture the fidelity and high sampling rate of more advanced equipment (e.g., a channel scanner) used when drive testing for analysis of propagation characteristics. In this work, we study the impact of various effects induced by user equipment (UE), when sampling signal quality. These shortcomings include averaging over multiple samples, imprecise quantization, and non-uniform and/or less frequent channel sampling. We specifically, investigate the accuracy of characterizing large-scale fading using crowdsourced data in presence of the aforementioned phone measurement shortcomings. To do so, we conduct extensive in-field experiments across heterogeneous devices and environments to empirically quantify the perceived channel characteristics by phone measurements. Analyzing the quality of the smartphone measurements in LTE indicates that the inferred radio propagation models, is comparable with models obtained by advanced equipment.

Ensuring cellular coverage is an important and costly concern for carriers due to the expense of in-field experimentation (i.e., drive testing). With the ubiquity of smartphones, apps, and social media, there has been an explosion of crowdsourcing to understand a vast array of trends and topics at a minimal cost to the organization. While cellular carriers might seek to replace the expensive act of drive testing with the nearly cost-free crowdsourcing, questions remain as to: (i) ~the accuracy of crowdsourcing, considering the lack of user control, (ii) ~the detection of when drive testing might still be required, and (iii) ~the quantification of how many additional in-field measurements to perform for a certain accuracy level. In this work, we use geographical features of a region to reduce in-field propagation experimentation by predicting the number of measurements required to accurately characterize its path loss. In particular, we study the path loss prediction accuracy of drive testing and crowdsourcing by taking millions of measurements in a suburban and downtown region. We then use statistical learning to build a relationship between these geographical features and the measurements required. In doing so, we find that the number of measurements collected to achieve a certain path loss accuracy over the entire region can be reduced by up to $58%$ in a high density drive testing scenario.

Understanding user-induced effects on signal reception across multiple frequency bands is of great scientific and military importance to the wireless industry. Various onbody locations and directional heading of the user are believed to impact the performance of mobile devices, but there has been little work across multiple frequency bands to quantify these user-induced effects. In this work, we perform a measurement study to explore user effects on radio wave propagation with varying line-of-sight conditions and environments across multiple frequency bands, including white space (500 and 800 MHz), cellular (1800 MHz), and WiFi (2400 MHz) frequency bands. To do so, we first conduct a baseline experiment that characterizes the propagation channel in this environment. We show that the propagation differences for ground-to-ground communication (common in Ad Hoc and WiFi scenarios) and tower-to-ground communication (common in cellular scenarios) are frequency dependent. Then, we measure signal quality as a function of the on-body location of the receiver, directional heading of the user with respect to the transmitter, vegetation type, frequency band, and propagation distance. Our assessment reveals that the user directionality with respect to the transmitter can reduce received signal strength up to 20 dB and reduce throughput by 20.9% at most. We also find that the body can act like an antenna, increasing reception by 4.4 dB and throughput by 14.4% over a reference node at the same distance. Since our study spans many critical (UHF) frequency bands, we believe these results will have far-reaching impact on a broad range of network types.

Understanding user-induced effects on signal reception across multiple frequency bands is of great scientific and military importance to the wireless industry. Various on-body locations and directional heading of the user are believed to impact the performance of mobile devices, but there has been little work across multiple frequency bands to quantify these user-induced effects. In this work, we perform a measurement study to explore user effects on radio wave propagation with varying line-of-sight conditions and environments across multiple frequency bands, including white space, cellular, and WiFi frequency bands. To do so, we first conduct a baseline experiment that investigates the propagation differences for ground-to-ground communication and tower-to-ground communication. Then, we measure signal quality as a function of the on-body location of the receiver, directional heading of the user with respect to the transmitter, vegetation type, frequency band, and propagation distance. Our assessment reveals that the user directionality can reduce received signal strength up to 20 dB and reduce throughput by 20.9% at most. We also find that the body can act like an antenna, increasing reception by 4.4 dB and throughput by 14.4% over a reference node at the same distance.

While there were high hopes for multihop wireless networks (mesh) to provide ubiquitous Wi-Fi in many cities, in-field trials revealed the node spacing required for Wi-Fi propagation induced a prohibitive cost model for network carriers to deploy. However, the digitization of TV channels and new FCC regulations have reapportioned spectrum for data networks with far greater range than WiFi due to lower carrier frequencies. In this paper, we analyze our in-field measurements in the Dallas-Fort Worth metroplex of channel occupancy in both WiFi and white space frequencies to deploy a wireless multihop backhaul tier. We design a measurement-driven heuristic algorithm, Band-based Path Selection (BPS), to approach optimal channel assignment of both white space and WiFi spectrum with reduced computational complexity. Numerical results show that BPS nearly doubles the served traffic of existing multi-channel, multi-radio algorithms, which are agnostic to diverse propagation characteristics across bands. Most importantly, this paper lays a foundation for the optimal use of white space and WiFi bands in the backhaul tiers of mesh networks across diverse population densities.

Environmental factors that lead to the movement and type of obstacles in and around wireless links are well-known to directly affect channel characteristics. However, while mobile users typically have repeatable daily or weekly patterns with common locations being frequently visited, many protocols along the network stack do not attempt to identify when physical locations are revisited. If wireless channels could be recognized as previously visited, the observance of good and bad decisions in that particular context could dramatically improve some network protocols. In this paper, we present a channel recognition framework which uses the geometrical shape of the link-level performance in a particular context across transmission modes and channel qualities. When attempting to recognize a channel condition, the performance of data transmissions is observed and compared against known channel types to detect similar behavior. The matching channel type can be used as an input to the link adaptation training and resulting decision structure. We perform extensive experimentation on controlled repeatable channels as well as in-field channels to show the validity of the classification algorithm.

In anticipation of the increasing demand of wireless traffic, WiFi standardization efforts have recently focused on two key technologies for capacity improvement: multi-user MIMO and wider bandwidth. However, users experience heterogeneous channel orthogonality characteristics across sub-carriers in the same channel bandwidth, which prevents ideal multi-user gain. Moreover, frequency selectivity increases as bandwidth scales and correspondingly severely deteriorates multi-user MIMO performance. In this work, we consider the frequency selectivity of current and emerging WiFi channel bandwidths to optimize multi-user MIMO by dividing the occupied channel bandwidth into equally-sized sub-channels according to the level of frequency selectivity. In our selectivity-aware multi-user MIMO design, SAMU, each sub-channel is allocated according to the largest bandwidth that can be considered frequency-flat, and an optimal subset of users is chosen to serve in each sub-channel according to spatial orthogonality, achieving a significant performance improvement for all users in the network. Additionally, we propose a selectivity-aware very high throughput (SA-VHT) mode, which is based on and an extension to the existing IEEE 802.11ac standard. Over emulated and real indoor channels, even with minimal mobility, SAMU achieves as much as 80 percent throughput improvement compared to existing multi-user MIMO schemes, which could serve as a lower bound as bandwidth scales.

Existing rate adaptation protocols have advocated training to establish the relationship between channel conditions and the optimal modulation and coding scheme. However, wireless devices for outdoor and vehicular communications frequently enter environments they have not yet encountered and therefore, have insufficient training for rate adaptation decisions. In addition, protocols are often optimally tuned for indoor environments but, when taken outdoors, perform poorly. In both cases, the decision structure formed offline lacks the ability to acclimate to a new situation on the fly. The diverse and ever-changing environments of increasingly mobile wireless devices call for a rate adaption scheme that can quickly adjust accordingly to form a unique environment set established by the user. In this paper, we propose an on-the-fly, in-situ training (FIT) mechanism which addresses the challenges of making rate decisions with unpredictable fluctuation and lack of repeatability of real wireless channels. We design and conduct extensive experiments on emulated and in-field wireless channels to evaluate the in-situ training process, showing that the rate decision structure can be updated as channel conditions change using existing traffic flows.

Frequently, client-side wireless devices have a view of multiple WiFi access points, whether from open residential and commercial networks, corporate networks, or mesh networks. Given the increasing number of radios and antennas in today's wireless devices, residual capacity from these multiple APs could be leveraged if client devices communicate with multiple APs simultaneously. In this paper, we exploit multi-user multi-input multi-output (MU-MIMO) technology to improve throughput and reliability in both directions of a wireless connection. For uplink, we use multi-user beamforming to enable the client devices to send multiple data streams to multiple APs simultaneously. For downlink, we leverage interference nulling technology to allow the client devices to decode parallel packets from multiple APs. This iBeam system requires no changes to existing APs or backhaul networks and is compatible with the IEEE 802.11 standards. We experimentally evaluate iBeam and show significant throughput improvements over both single-AP connections and multi-AP connections in a time division mode. The client's reliability and stability are also significantly improved due to the multi-AP diversity gain.

Repeatability, isolation and accuracy are the most desired factors while testing wireless devices. However, they cannot be guaranteed by traditional drive tests. Channel emulators play a major role in filling these gaps in testing. In this paper we present an efficient channel emulator which is better than existing commercial products in terms of cost, remote access, support for complex network topologies and scalability. We present the hardware and software architecture of our channel emulator and describe the experiments we conducted to evaluate its performance against a commercial channel emulator.

While many metropolitan areas sought to deploy city-wide WiFi networks, the densest urban areas were not able to broadly leverage the technology for large-scale Internet access. Ultimately, the small spatial separation required for effective 802.11 links in these areas resulted in prohibitively large up-front costs. The FCC has reapportioned spectrum from TV white spaces for the purposes of large-scale Internet connectivity via wireless topologies of all kinds. The far greater range of these lower carrier frequencies are especially critical in rural areas, where high levels of aggregation could dramatically lower the cost of deployment and is in direct contrast to dense urban areas in which networks are built to maximize spatial reuse. Thus, leveraging a broad range of spectrum across diverse population densities becomes a critical issue for the deployment of data networks with WiFi and white space bands. In this paper, we measure the spectrum utility in the Dallas-Fort Worth metropolitan and surrounding areas and propose a measurement-driven band selection framework, Multiband Access Point Estimation (MAPE). In particular, we study the white space and WiFi bands with in-field spectrum utility measurements, revealing the number of access points required for an area with channels in multiple bands. In doing so, we find that networks with white space bands reduce the number of access points by up to 1650% in sparse rural areas over similar WiFi-only solutions. In more populated rural areas and sparse urban areas, we find an access point reduction of 660% and 412%, respectively. However, due to the heavy use of white space bands in dense urban areas, the cost reductions invert (an increase in required access points of 6%). Finally, we numerically analyze band combinations in typical rural and urban areas and show the critical factor that leads to cost reduction: considering the same total number of channels, as more channels are available in the white space bands, less access points are required for a given area.

Channel emulators are valuable tools for controllable and repeatable wireless experimentation. Often, however, the high cost of such emulators preclude their widespread usage, especially in large-scale wireless networks. Moreover, existing channel emulators offer either very realistic channels for simplistic topologies or complex topologies with highly-abstracted, low-fidelity channels. To bridge the gap in offering a low-cost channel emulation solution which can scale to a large network size, in this paper, we study the tradeoff in channel emulation fidelity versus the hardware resources consumed using both analytical modeling and FPGA-based implementation. To reduce the memory footprint of our design, we optimize our channel emulation using an iterative structure to generate the Rayleigh fading channel. In addition, the channel update rate and word length selection are also evaluated in the paper which greatly improve the efficiency of implementation. We then extend our analysis of a single channel to understand how the implementation scales for the emulation of a large-scale wireless network, showing that up to 24 vehicular channels can be emulated in real-time on a single Virtex-4 FPGA.

A number of rate adaptation protocols have proposed using instantaneous channel quality to select the physical layer data rate. However, due to fast channel variations, even aggressive probing of the channel before each data packet is often unable to offer an up-to-date notion of channel quality. In this paper, we propose a coherence-aware channel indication and prediction algorithm for rate adaptation (CIPRA) and evaluate it analytically and experimentally, considering both the measurement errors and the staleness of the channel quality indicator. CIPRA uses Minimum-Mean-Square-Error (MMSE), first-order prediction and jointly considers the time interval over which the prediction will occur and the coherence time of the channel to determine the optimal window size for previous channel quality indicator measurements. We also implement a Doppler shift estimation method in hardware to assist the proposed channel prediction algorithm. We show that CIPRA outperforms existing methods in terms of prediction fidelity and throughput via experimental results from an FPGA-based platform on emulated and in-field wireless channels. In our experiments in the field, CIPRA achieves up to 1.66 times the throughput achieved by the indication and prediction method currently used by off-the-shelf cards. CIPRA could be easily applied to other channel indicators, although we only evaluated RSSI-based rate adaptation in our experiments to isolate our gains.

In hardware, packet loss may happen due to overflow from a finite-depth transmit buffer. To prevent such losses and further improve rate selection, we exploit statistical knowledge of transmit buffer occupancy and source packet distribution in IEEE 802.11-based systems, which have variable frame slots. We consider a traditional method of rate adaptation based on channel quality information and evaluate the throughput gain in hardware when the buffer occupancy and source packet distribution information are known. Our optimization objective is to maximize the throughput with constant transmit power since most IEEE 802.11 APs and nodes operate in this manner. We also derive an upper bound of the improvement introduced by exploiting the offered load distribution and buffer status information. By evaluating the effect of diverse buffer sizes with different packet arrival distributions, both our theoretical analysis and our experimental results show that the throughput can be greatly improved in many cases when the source packet distribution and buffer status information are exploited.

The high mobility of vehicular networks makes channel estimation and phase tracking challenging and an important problem for OFDM receiver design. In IEEE 802.11p vehicular networks, the channel estimation based on a long preamble in the PHY frame performs poorly in fast-fading channels. Moreover, the pilot-based phase tracking usually suffers from large residual phase errors. In this paper, we propose and implement a novel phase tracking algorithm, leveraging the decoded data. We also provide a detailed hardware design for a decoder-based channel estimation algorithm with a pipeline structure on an FPGA-based platform. Through diverse experiments via an advanced channel emulator, our results show that the proposed algorithm could significantly improve the system performance in terms of packet error rate, while adding less than 3% of additional hardware resources. We also jointly evaluate the packet error rate versus the bitwidth of the data in the FPGA, which is important to achieve a good balance between the hardware cost and the system performance.

A novel and efficient method for emulating a Weibull fading channel is presented. The proposed Weibull fading channel generator employs CORDIC technology which can work without the need for a multiplier, a scarce FPGA resource when emulating wireless channels. Using the proposed hardware implementation structure, the resource consumption is effectively reduced. Further, the proposed generator is able to produce the Weibull fading channel with arbitrary values of the shape parameter γ, in contrast to existing state-of-the-art methods, which can only generate Weibull channels with fixed values of γ. In simulation, the PDF of the Weibull fading channel agrees closely with the theoretical values. The simulation results also show the effectiveness of the proposed structure to generate Lognormal and Suzuki fading channels.

An increasing number of adaptive protocols use training data to learn optimal parameter choices for adaptation in wireless communication networks. For instance, several recent papers have studied link adaptation protocols based on context information such as node velocity and SNR. However, a number of embedded sensors providing context information frequently report erroneous values, e.g., GPS errors and accelerometer lag, producing incorrect information about motion. As a result, the relationship between the context information and optimal parameter choices that the adaptive algorithm is attempting to establish is erroneous. In this paper, we propose an outlier detection algorithm, which detects the corrupted information due to system errors. The proposed outlier detection algorithm is based on an alternating minimization approach. To evaluate the performance of the proposed algorithm, we apply it to a link-level context-aware rate adaptation system. Numerical results on emulated channels and in-field testing demonstrate that the proposed algorithm increases the prediction accuracy of the optimal transmission mode by 87% and the throughput by 18%.

Context awareness has received increasing attention with the proliferation of various types of sensors on mobile devices. However, while wireless performance is known to be highly correlated with environmental settings, mobile devices have yet to fully exploit the awareness of context to improve wireless performance. In this paper, we leverage available context information to improve link-level adaptation via decision-tree classifiers and extensively evaluate its performance over emulated channels as well as with in-field trials. We first propose a classification method based on decision trees to select the optimal transmission parameters such as modulation, coding rate and packet size. We then quantify the throughput improvement using the proposed scheme and show that in some scenarios the throughput increases by over 100% compared to traditional SNR-based rate adaptation protocols. Second, we analyze the amount of training to assess the classification scheme. Third, we validate classification-based method by implementation on two different test platforms for extensive experimentation. We reveal the importance of the various contextual attributes used and identify channel type as a key parameter that affects classification performance. Finally, we study and quantify the use of context information across multiple different frequency bands and demonstrate the significant throughput gains that can be obtained.

The application of machine learning algorithms in wireless communications has attracted increasing attention due to the promising performance gains recently achieved. Static classification algorithms have been successfully applied to training protocols that adapt transmission parameters according to context information. However, in reality, there are many time-varying reasons for fading channel quality including mobility of sender, receiver, and/or obstacles within the environment. Moreover, time-varying noise further exacerbates the dynamics of the channel. These problems pose new challenges for the application of static classification algorithms in context-aware algorithms and suggest that sequential classifiers which leverage the temporal dynamics and correlation of context information might be more appropriate. In this paper, we apply sequential training to rate adaptation (ASTRA), leveraging the temporal correlation of context information. In particular, linear and non-linear sequential coding schemes are used in the training process for selecting the modulation/coding rate that achieves the highest throughput for the given context. Experimental results on measurements from emulated and in-field channels demonstrate that ASTRA can significantly increase the accuracy of selecting these target rates by up to 175% and increase the resulting throughput by up to 66% over rate adaptation training which uses static classifier-based methods.

Contending flows in multihop 802.11 wireless networks compete with two fundamental asymmetries: 1) channel asymmetry, in which one flow has a stronger signal, potentially yielding physical layer capture; and 2) topological asymmetry, in which one flow has increased channel state information, potentially yielding an advantage in winning access to the channel. Prior work has considered these asymmetries independently with a highly simplified view of the other. However, in this paper, we perform thousands of measurements on coupled flows in urban environments and build a simple yet accurate model that jointly considers information and channel asymmetries. We show that if these two asymmetries are not considered jointly, throughput predictions of even two coupled flows are vastly distorted from reality when traffic characteristics are only slightly altered (e.g., changes to modulation rate, packet size, or access mechanism). These performance modes are sensitive not only to small changes in system properties, but also small-scale link fluctuations that are common in an urban mesh network. We analyze all possible capture relationships for two-flow subtopologies and show that capture of the reverse traffic can allow a previously starving flow to compete fairly. Finally, we show how to extend and apply the model in domains such as modulation rate adaptation and understanding the interaction of control and data traffic.

Contending flows in multi-hop 802.11 wireless networks compete with two fundamental asymmetries: (i) channel asymmetry, in which one flow has a stronger signal, potentially yielding physical layer capture, and (ii) topological asymmetry, in which one flow has increased channel state information, potentially yielding an advantage in winning access to the channel. Prior work has considered these asymmetries independently with a highly simplified view of the other. However, in this work, we perform thousands of measurements on coupled flows in urban environments and build a simple, yet accurate model that jointly considers information and channel asymmetries. We show that if these two asymmetries are not considered jointly, throughput predictions of even two coupled flows are vastly distorted from reality when traffic characteristics are only slightly altered (e.g., changes to modulation rate, packet size, or access mechanism). These performance modes are sensitive not only to small changes in system properties, but also small-scale link fluctuations that are common in an urban mesh network. We analyze all possible capture relationships for two-flow sub-topologies and show that capture of the reverse traffic can allow a previously starving flow to compete fairly. Finally, we show how to extend and apply the model in domains such as modulation rate adaptation and understanding the interaction of control and data traffic.

Accurately selecting modulation rates for time-varying channel conditions is critical for avoiding performance degradations due to rate overselection when channel conditions degrade or underselection when channel conditions improve. In this paper, we design a custom cross-layer framework that enables: 1) implementation of multiple and previously unimplemented rate adaptation mechanisms; 2) experimental evaluation and comparison of rate adaptation protocols on controlled, repeatable channels as well as residential urban and downtown vehicular and nonmobile environments in which we accurately measure channel conditions with 100- s granularity; and 3) comparison of performance on a per-packet basis with the ideal modulation rate obtained via exhaustive experimental search. Our evaluation reveals that SNR-triggered protocols are susceptible to overselection from the ideal rate when the coherence time is low (a scenario that we show occurs in practice even in a nonmobile topology), and that “in situ” training can produce large gains to overcome this sensitivity. Another key finding is that a mechanism effective in differentiating between collision and fading losses for hidden terminals has severely imbalanced throughput sharing when competing links are even slightly heterogeneous. In general, we find trained SNR-based protocols outperform loss-based protocols in terms of the ability to track vehicular clients, accuracy within outdoor environments, and balanced sharing with heterogeneous links (even with physical layer capture).

In this paper, we perform an extensive measurement study on a multi-tier mesh network serving 4,000 users. Such dense mesh deployments have high levels of interaction across heterogeneous wireless links. We find that this heterogeneous backhaul consisting of data-carrying (forwarding) linksandnon- data-carrying (non-forwarding) links creates two key effects on performance. First, we show that low-rate management and control packets can produce a disproportionally large degradation in data throughput. We define a metric for this effect called Wireless Overhead Multiplier and use it to quantify the impact of MAC and PHY mechanisms on the the throughput degradation. Surprisingly, we show that these multiplicative effects are primarily driven by the non-forwarding links where, in the worst case, data packets lose physical layer capture to the overhead, yielding disproportionate throughput degradation. Finally, we show that when data flows contend in this worst-case scenario, the loss-based autorate policy is unnecessarily triggered, causing throughput imbalance and poor network utilization.

Significant progress has been made in understanding the behavior of TCP and congestion-controlled traffic over multi- hop wireless networks. Despite these advances, however, no prior work identified severe throughput imbalances in the basic scenario of mesh networks, in which one-hop flows contend with two-hop flows for gateway access. In this paper, we demonstrate via real network measurements, test-bed experiments, and an analytical model that starvation exists in such a scenario, i.e., the one-hop flow receives most of the bandwidth while the two- hop flow starves. Our analytical model yields a solution consisting of a simple contention window policy that can be implemented via mechanisms in IEEE 802.11e. Despite its simplicity, we demonstrate through analysis, experiments, and simulations, that the policy has a powerful effect on network-wide behavior, shifting the network's queuing points, mitigating problematic MAC behavior, and ensuring that TCP flows obtain a fair share of the gateway bandwidth, irrespective of their spatial locations.

Accurately selecting modulation rates for time-varying channel conditions is critical for avoiding performance degradations due to rate overselection when channel conditions degrade or underselection when channel conditions improve. In this paper, we design a custom cross-layer framework that enables (i) implementation of multiple and previously unimplemented rate adaptation mechanisms, (ii) experimental evaluation and comparison of rate adaptation protocols on controlled, repeatable channels as well as residential urban and downtown vehicular and non-mobile environments in which we accurately measure channel conditions with 100-μs granularity, and (iii) comparison of performance on a per-packet basis with the ideal modulation rate obtained via exhaustive experimental search. Our evaluation reveals that SNR-triggered protocols are susceptible to overselection from the ideal rate when the coherence time is low (a scenario that we show occurs in practice even in a nonmobile topology), and that "in-situ" training can produce large gains to overcome this sensitivity. Another key finding is that a mechanism effective in differentiating between collision and fading losses for hidden terminals has severely imbalanced throughput sharing when competing links are even slightly heterogeneous. In general, we find trained SNR-based protocols outperform loss-based protocols in terms of the ability to track vehicular clients, accuracy within outdoor environments, and balanced sharing with heterogeneous links (even with physical layer capture).

Multihop wireless mesh networks can provide Internet access over a wide area with minimal infrastructure expenditure. In this work, we present a measurement driven deployment strategy and a data-driven model to study the impact of design and topology decisions on network-wide performance and cost. We perform extensive measurements in a two-tier urban scenario to characterize the propagation environment and correlate received signal strength with application layer throughput. We find that well-known estimates for pathloss produce either heavily overprovisioned networks resulting in an order of magnitude increase in cost for high pathloss estimates or completely disconnected networks for low pathloss estimates. Modeling throughput with wireless interface manufacturer specifications similarly results in severely underprovisioned networks. Further, we measure competing, multihop flow traffic matrices to empirically define achievable throughputs of fully backlogged, rate limited, and web-emulated traffic. We find that while fully backlogged flows produce starving nodes, rate-controlling flows to a fixed value yields fairness and high aggregate throughput. Likewise, transmission gaps occurring in statistically multiplexed web traffic, even under high offered load, remove starvation and yield high performance. In comparison, we find that well-known noncompeting flow models for mesh networks over-estimate network-wide throughput by a factor of 2. Finally, our placement study shows that a regular grid topology achieves up to 50 percent greater throughput than random node placement.