Large constellations of low-earth orbit satellites enable frequent high-resolution earth imaging for numerous geospatial applications. They generate large volumes of data in space, hundreds of Terabytes per day, which much be transported to Earth through constrained intermittent connections to ground stations. The large volumes lead to large day-level delay in data download and exorbitant cloud storage costs. We propose DeepSpace, a new deep learning-based super-resolution approach that compresses satellite imagery by over two orders of magnitude, while preserving image quality using a tailored mixture of experts (MoE) super-resolution framework. DeepSpace reduces the network bandwidth requirements for space-Earth transfer, and can compress images for cloud storage. DeepSpace achieves such gains with the limited computational power available on small LEO satellites. We extensively evaluate DeepSpace against a wide range of state-of-the-art baselines considering multiple satellite image datasets and demonstrate the above mentioned benefits. We further demonstrate the effectiveness of DeepSpace through several distinct downstream applications (wildfire detection, land use and cropland classification, and fine-grained plastic detection in oceans).

Fourier features based positional encoding (PE) is commonly used in machine learning tasks that involve learning high-frequency features from low-dimensional inputs, such as 3D view synthesis and time series regression with neural tangent kernels. Despite their effectiveness, existing PEs require manual, empirical adjustment of crucial hyperparameters, specifically the Fourier features, tailored to each unique task. Further, PEs face challenges in efficiently learning high-frequency functions, particularly in tasks with limited data. In this paper, we introduce sinusoidal PE (SPE), designed to efficiently learn adaptive frequency features closely aligned with the true underlying function. Our experiments demonstrate that SPE, without hyperparameter tuning, consistently achieves enhanced fidelity and faster training across various tasks, including 3D view synthesis, Text-to-Speech generation, and 1D regression. SPE is implemented as a direct replacement for existing PEs. Its plug-and-play nature lets numerous tasks easily adopt and benefit from SPE.

The Open RAN architecture, with disaggregated and virtualized RAN functions communicating over standardized interfaces, promises a diversified and multi-vendor RAN ecosystem. However, these same features contribute to increased operational complexity, making it highly challenging to troubleshoot RAN related performance issues and failures. Tackling this challenge requires a dependable, explainable anomaly detection method that Open RAN is currently lacking. To address this problem, we introduce SpotLight, a tailored system archtecture with a distributed deep generative modeling based method running across the edge and cloud. SpotLight takes in a diverse, fine grained stream of metrics from the RAN and the platform, to continually detect and localize anomalies. It introduces a novel multi-stage generative model to detect potential anomalies at the edge using a light-weight algorithm, followed by anomaly confirmation and an explain-ability phase at the cloud, that helps identify the minimal set of KPIs that caused the anomaly. We evaluate SpotLight using the metrics collected from an enterprise-scale 5G Open RAN deployment in an indoor office building. Our results show that compared to a range of baseline methods, SpotLight yields significant gains in accuracy (13% higher F1 score), explain-ability (2.3 -- 4X reduction in the number of reported KPIs) and efficiency (4 -- 7X bandwidth reduction).

The Open RAN architecture, featuring disaggregated and virtualized RAN functions communicating over standardized interfaces, promises a diverse, multi-vendor ecosystem. However, these features also increase operational complexity, complicating the troubleshooting of RAN performance issues and failures. Addressing this challenge requires a reliable, explainable anomaly detection method, which Open RAN currently lacks. To address this problem, we have developed SpotLight, a tailored distributed deep learning method running across the edge and cloud. SpotLight continuously detects and localizes anomalies by analyzing a diverse, fine-grained stream of metrics from the RAN and platform. It employs a novel multi-stage generative model to identify potential anomalies at the edge using a lightweight algorithm, followed by anomaly confirmation and an explainability phase in the cloud, which pinpoints the minimal set of KPIs responsible for the anomaly. In this demo, using a carrier-grade indoor Open RAN testbed with configurable anomaly event generation and replay, we highlight (1) the difficulty of troubleshooting problems in Open RAN and (2) accurate, efficient, and explainable online anomaly detection with SpotLight and corresponding visualization in comparison with prior art.

Drive testing continues to play a key role in mobile network optimization for operators but its high cost is a big concern. Alternative approaches like virtual drive testing (VDT) target device testing in the lab whereas MDT or crowdsourcing based approaches are limited by the incentives users have to participate and contribute measurements. With the aim of augmenting drive testing and significantly reducing its cost, we propose GenDT, a novel deep generative model that synthesizes high-fidelity time series of key radio network key performance indicators (KPIs). The training of GenDT relies on a relatively small amount of real-world measurement data along with corresponding and easily accessible network and environment context data. Through this, GenDT learns the relationship between context and radio network KPIs as they vary over time, and therefore trained GenDT model can subsequently be relied on to generate time series for different KPIs for new drive test routes (trajectories) without having to collect field measurements. GenDT represents an initial attempt at enabling efficient drive testing via generative modeling. Evaluations with real-world mobile network drive testing measurement datasets from two countries demonstrate that GenDT can synthesize significantly more dependable data than a range of baselines. We further show that GenDT has the potential to significantly reduce the drive testing related measurement effort, and that GenDT-generated data yields similar results to that with real data in the context of two downstream use cases - QoE prediction and handover analysis.

Service-level mobile traffic data enables research studies and innovative applications with a potential to shape future service-oriented communication systems and beyond. However, real-world datasets reporting measurements at the individual service level are hard to access as such data is deemed commercially sensitive by operators. APPSHOT is a model for generating synthetic high-fidelity city-scale snapshots of service level mobile traffic. It can operate in any geographical region and relies solely on easily available spatial context information such as population density, thus allowing the generation of new and open traffic datasets for the research community. The design of APPSHOT is informed by an original characterization of service-level mobile traffic data. APPSHOT is a novel conditional GAN design instantiated by a convolutional neural network generator and two discriminators. The model features several other innovative mechanisms including multi-channel and overlapping patch based generation to address the unique challenges involved in generating mobile service traffic snapshots. Experiments with ground-truth data collected by a major European operator in multiple metropolitan areas show that APPSHOT can produce realistic network loads at the service level for areas where it has no prior traffic knowledge, and that such data can reliably support service-oriented networking studies.

The main requirement of the computation offloading service is the low service delay, which would correspond to a high Quality of Service (QoS). Recently, many works show that placing virtual resources (e.g., computing resource) in the Mobile Edge Network (MEN) to form a Mobile Edge Computing (MEC) system contributes to a lower service delay. However, due to the limited virtual resource in the MEN and the constraints of the wireless channel condition, only part of users can be served with low enough service delay. Moreover, considering the independent virtual machine (VM) environment on the MEC server in MEN for each user, efficient virtual resources management is needed. In this paper, we propose two computation offloading strategies combined with the virtual resources management to minimize the average service delay, where the virtual resources consists of the computing resources and the storage resources. One of the proposed strategies guarantees the fairness of users while the other one not. Then the optimal deployments of VMs with different strategies are obtained through the simulations. We also find and calculate the optimal proportion of these two kinds of virtual resource when there is only a limited budget for them.