Performance of wireless navigation systems degrade sharply in industrial environments dominated by metallic clutter and heavy multipath. The primary cause lies in how anchors are placed or selected. Methods that succeed in open spaces fail under severe non-line-of-sight (nLoS) and frequent layout changes, for instance, induced by moving forklifts or shifting shelves. We introduce Argos, a multimodal wireless digital twin that fuses visual and RF information to optimize anchor selection. Visual imagery reconstructs the 3D layout, updated continuously via existing surveillance cameras, while RF measurements capture material-specific attenuation and reflections. We show that layout priors alone are insufficient; combining them with the required RF optics yields a material-aware channel model that predicts range errors under severe nLoS. Argos adapts proactively to environmental changes, without the requirement of repeated RF calibration or retraining, sustaining sub-meter localization accuracy in dynamic scenes. We validate Argos in a 120m2factory testbed spanning over 700 locations, where the digital twin is built from 5K RGB images and 0.4M UWB CIR samples. To our knowledge, this is the first system to exploit visual priors for adaptive orchestration of wireless navigation infrastructure.

Visual Analytics Pipelines (VAPs) for real-time surveillance are resource-intensive, consuming high energy and bandwidth. We propose EcoVis, a novel approach that uses mmWave sensing to identify regions of interest (ROIs) for compressing surveillance video frames. This reduces the transmission of static background, optimizing resource usage while maintaining high fidelity of content relevant in traffic surveillance. Unlike conventional methods that rely solely on video frames to detect RoIs, our use of mmWave range-azimuth maps achieves a comparable reduction in network bandwidth while lowering energy consumption by approximately 40%. Moreover, our approach enhances energy efficiency by nearly another 25%, by dynamically controlling the sleep cycle of the camera. For simpler tasks such as vehicle detection or counting, EcoVis works with minimal reliance of the camera. Due to its lower dimensionality compared to video, it allows on-device processing, improving operational speed and network efficiency by roughly 20%. Finally, we introduce both uniform and non-uniform tiling algorithms, utilizing RoIs derived from mmWave analysis. These algorithms enable video encoding with tile-specific Quantization Parameters (QPs), optimizing the overall compression process.

Neural Radiance Fields (NeRF) have emerged as a powerful technique for synthesizing novel views of complex 3D scenes from a sparse set of images. The advances in NeRF has shown prominence in the field of wireless networks as well. This paper explores the application of NeRF in the domain of spectrum sensing, proposing a novel approach that leverages the capabilities of RF based NeRF and extend it to enhance the accuracy and efficiency of spectrum sensing in wireless communication networks. Our proposed solution SpecNeRF is evaluated through extensive experiments, demonstrating significant improvements in terms of scalability, robustness to environmental changes, and adaptability to varying signal conditions. SpecNeRF not only provides a viable solution for current spectrum sensing challenges but also paves the way for innovative applications in future wireless networks, including cognitive radio and 6G technologies.

A critical challenge in deploying UAV-based wireless networks is optimizing the UAV's position in aerial space to ensure robust connectivity. To achieve optimal positioning, the UAV must estimate the propagation loss for each ground terminal across the aerial space, also called Radio Environment Map (REM). Existing literature in this domain emphasizes accuracy enhancement of the REM estimations from sparse and noisy signal measurements - often leveraging deep learning techniques. However, these estimated REMs rapidly become outdated as UEs move, requiring repeated measurements and re-estimation. This introduces significant scalability challenges, especially in environments characterized by high UE mobility.

Edge-assisted wireless sensing is increasingly popular, where complex neural network models perform inference tasks on wireless channel state information (CSI) data streamed from IoT devices. However large volumes of CSI data sent across the network for inference can significantly impact network bandwidth and reduce the Quality of Experience. This paper tackles the challenge of optimizing network resource utilization in wireless sensing systems by compressing and subsampling CSI streams. We evaluate methods that quantize and selectively subsample CSI data before transmission to the edge server, which is then fed to the inference models. Such approach reduces bandwidth and computational load, improving data transmission and processing efficiency. Experiments conducted in two real testbeds (indoors as well as outdoors) show how CSI compression preserves sensing information integrity while enhancing system performance in terms of latency, energy efficiency, and throughput. By integrating quantization and subsampling with edge computing, this work enhances wireless sensing systems, making them more scalable and efficient in utilizing network resources.

The demand for real-time and accurate mapping is ubiquitous, particularly in complex indoor settings. While SLAM-based methods are popular, Radio Tomographic Imaging (RTI) offers an essential set of advantages, including mapping inaccessible or enclosed spaces, shorter scanning trajectories, or even identifying material properties of structures on the map. However, existing RTI systems typically depend on pre-deployed, precisely calibrated infrastructure with ample computing power, making it challenging to deploy in a ubiquitous setting. We designUbiqMap, a lightweight RTI-based end-to-end system capable of mapping indoor spaces in real-time, with minimal to zero reliance over pre-deployed infrastructure. We evaluate the performance ofUbiqMapin various scenarios, including two real deployments - a moderately complex residential apartment (800 sq. ft) and a large building foyer area (3000 sq. ft) and a few simulated scenarios. We demonstrate howUbiqMapcan benefit over traditional SLAM-based techniques in specific contexts and advocate the fusion of RTI methods with SLAM to improve future mapping technologies. Overall,UbiqMapimproves the quality of the estimated map by 30%–40% over the state-of-the-art with equivalent resource availability.

With the emergence of next generation networks, Network Traffic Classification (NTC) has seen greater importance in network management and security. Recently, Channel State Information (CSI) based WiFi sensing techniques have shown their potential for NTC applications [1], [2] as a privacy-preserving yet effective tool. As CSI could be prone to interference, this paper examines the performance of CSI-based NTC models under interference-induced adversarial attacks. Specifically, the impact of spectral allocation of the interference, underlying interfering network traffic type, and physical location of the interference are studied and quantified. We conducted experiments using off-the-shelf devices to test the NTC performance, with and without the adversarial interference attack of ping, buffered video streaming, and live video streaming network traffics. Subsequently, we found that spectral allocation of the attacking interference and the underlying traffic types of interference could be used to deceive the established NTC model, and different network traffic types show different robustness across the interference cases. Namely, ping suffers the most in the spectrally manipulated attack, with the classification accuracy down to as low as 23.2%, whereas Twitch might be completely misidentified as other traffic in underlying traffic type controlled attack.

Shared spectrum systems is an emerging paradigm to improve spectrum utilization and thus address the unabated increase in mobile data consumption. The paradigm allows the “unused” spectrum bands of licensed Primary Users (PUs) to be shared with Secondary Users (SUs), without causing any harmful interference to the PUs. Allocation of spectrum to the SUs is done based on spectrum availability at the SUs' locations; such allocation of spectrum is greatly facilitated by spectrum occupancy maps. In this work, we address the problem of creating spectrum occupancy maps from spectrum occupancy data over a large number of instants, in the challenging scenario of dynamically (temporally) changing spectrum occupancy due to intermittent transmission of primary users. The problem is particularly challenging when the available occupancy data is very sparse spatially, i.e., only very few locations report sensing data at any particular instant. We design various techniques to create spectrum maps in the above context, including a promising correlation-based merging method that merges observation vectors iteratively in conjunction with careful interpolation. Using extensive simulation over data including real data from cellular and deployed WiFi settings, we show that the correlation-based method is very effective in generating high-accuracy spatiotemporal spectrum maps even with very sparse observation vectors (as long as the number of such vectors is large enough).

We envision a flexible, dynamic airborne LTE infrastructure built upon Unmanned Autonomous Vehicles (UAVs) that will provide on-demand, on-time, network access, anywhere. In this paper, we design, implement and evaluate SkyRAN, a self-organizing UAV-based LTE RAN (Radio Access Network) that is a key component of this UAV LTE infrastructure network. SkyRAN determines the UAV's operating position in 3D airspace so as to optimize connectivity to all the UEs on the ground. It realizes this by overcoming various challenges in constructing and maintaining radio environment maps to UEs that guide the UAV's position in real-time. SkyRAN is designed to be scalable in that it can be quickly deployed to provide efficient connectivity even over a larger area. It is adaptive in that it reacts to changes in the terrain and UE mobility, to maximize LTE coverage performance while minimizing operating overhead. We implement SkyRAN on a DJI Matrice 600 Pro drone and evaluate it over a 90 000 m2 operating area. Our testbed results indicate that SkyRAN can place the UAV in the optimal location with about 30 secs of a measurement flight. On an average, SkyRAN achieves a throughput of 0.9 - 0.95X of optimal, which is about 1.5 - 2X over other popular baseline schemes.

We use a crowdsourcing approach for RF spectrum patrolling, where heterogeneous, low-cost spectrum sensors are deployed widely and are tasked with detecting unauthorized transmissions in a collaborative fashion while consuming only a limited amount of resources. We pose this as a collaborative signal detection problem where the individual sensor's detection performance may vary widely based on their respective hardware or software configurations, but are hard to model using traditional approaches. Still an optimal subset of sensors and their configurations must be chosen to maximize the overall detection performance subject to given resource (cost) limitations. We present the challenges of this problem in crowdsourced settings and present a set of methods to address them. The proposed methods use data-driven approaches to model individual sensors and develops mechanisms for sensor selection and fusion while accounting for their correlated nature. We present performance results using examples of commodity-based spectrum sensors and show significant improvements relative to baseline approaches.

We describe an end-to-end platform called SpecSense to support large scale spectrum monitoring. SpecSense crowdsources spectrum monitoring to low-cost, low-power commodity SDR/embedded platforms and provides necessary analytics support in a central spectrum server. In this work, we describe SpecSense and address specific challenges related to accurately estimate spectrum occupancy on demand with low overhead. To address the accuracy question, we augment state-of-the-art spatial interpolation techniques to accommodate scenarios where RF propagation characteristics change across space. To address the overhead question, we solve the sensor selection problem to select the minimum number of spectrum sensors that can best estimate the spectrum at the requested locations.

Enterprise wireless networks face significant challenges to deliver Quality-of-Experience (QoE) with the variety of mobile applications. One of the fundamental challenges is that the traditional definition of network capacity (often defined as throughput capacity) is not sufficient to reflect applications' requirements in wireless networks. In this paper, we propose to rethink the network capacity of wireless networks to better incorporate QoE. Specifically, we first propose a novel concept of an Experiential Capacity Region (ExCR) for wireless networks. ExCR is defined as a set of simultaneous application flows whose QoE requirements can be satisfied by the network. Next, we present the infrastructure based ExBox system that measures per-application QoE metrics and determines the ExCR for wireless networks to better serve a set of mobile application flows. In its core, ExBox employs light-weight machine learning techniques that are tailored for dynamic wireless environments. Through both large-scale simulations and extensive real-life experiments on WiFi and LTE networks, we show that ExBox delivers QoE in admission control decision with a precision of ≈ 0.8 - 0.9, even when clients experience diverse channel quality. Moreover, ExBox quickly adapts to changing network environments without much overhead.

We argue that spectrum sensing on mobile clients will be both necessary and feasible if we wish to manage the white space spectrum optimally in indoor spaces. We demonstrate the necessity with a set of empirical measurements showing the need for fine grained sensing. We demonstrate the feasibility by building a spectrum sensing infrastructure that collects measurements from sensing devices to analyze and better use spectrum resources. The infrastructure consists of mobile spectrum sensors that are built using DTV receiver dongles interfaced with Android-based mobile devices and a cloud-based central server to manage such sensing devices. We also show results about resource consumption (energy, network overhead) involved in operating such sensors. The vision is ultimately creating a system where mobile devices perform part-time spectrum sensing in a coordinated fashion under the control of a central spectrum manager. We lay out the research challenges based on our initial prototyping and benchmarking experience.

We address the problem of network-side localization where cellular operators are interested in localizing cellular devices by means of signal strength measurements alone. While fingerprinting-based approaches have been used recently to address this problem, they require significant amount of geo-tagged (`labeled') measurement data that is expensive for the operator to collect. Our goal is to use semi-supervised and unsupervised machine learning techniques to reduce or eliminate this effort without compromising the accuracy of localization. Our experimental results in a university campus (6 sq. km) demonstrate that sub-100m median localization accuracy is achievable with very little or no labeled data so long as enough training is possible with `unlabeled' measurements. This provides an opportunity for the operator to improve the model with time. We present extensive analysis of the error characteristics to gain insight and improve performance, including understanding spatial properties and developing confidence measures.

Spectrum databases used to estimate TV white space availability often provide inaccurate and largely conservative estimates as they are primarily based on empirical propagation models. This leads to 'loss' of white space spectrum that is critical in urban areas with large spectrum demand. While alternatives are possible in terms of incorporating direct spectrum measurements, the measurement locations must be judiciously chosen so that measurement effort is not prohibitive. Fundamentally, this boils down to addressing the estimation accuracy vs measurement effort question. We present a rigorous data driven analysis to address this using measurement data collected in parts of New York City metro area. We show that it is possible to develop models that estimate whether the current database estimates are reliable in a given location. Following this, we provide a recipe for developing a `measurement-augmented' spectrum database that takes the help of measurements where needed and falls back on the current propagation model-based database technique in the rest of the areas. The final takeaway is that it is possible to improve database accuracy significantly with only modest amount of measurements.

Recent industry trends suggest a new phenomenon in the mobile market: mobile virtual network operators or MVNOs that operate on top of existing cellular infrastructures. While MVNOs have shown significant growth in the US and elsewhere in the past two years and have been successful in attracting customers, there is anecdotal evidence that users are concerned about cellular performance when choosing MVNOs over traditional cellular operators. In this paper, we present the first systematic measurement study to shed light on this emerging phenomenon. We study the performance of 3 key applications: web access, video streaming and voice, in 2 popular MVNO families (a total of 8 carriers) in the US, where each MVNO family consists of a major base carrier and 3 MVNOs running on top of it. We observe that some MVNOs do indeed exhibit significant performance degradation and that there are key differences between the two MVNO families.

In this paper, we envision a scenario where mobile devices perform at least part-time spectrum sensing in a collaborative fashion under the control of a central server. The goal is to create an adequate `radio environment map' for the `white spaces' that will be useful for spectrum management decisions. We lay out the research challenges, describe a prototype implementation using a DTV receiver dongle interfaced with an Android-based mobile device, and present preliminary performance measurements.