Search for: All records

We consider a decentralized wireless network with several source-destination pairs sharing a limited number of orthogonal frequency bands. Sources learn to adapt their transmissions (specifically, their band selection strategy) over time, in a decentralized manner, without sharing information with each other. Sources can only observe the outcome of their own transmissions (i.e., success or collision), having no prior knowledge of the network size or of the transmission strategy of other sources. The goal of each source is to maximize their own throughput while striving for network-wide fairness. We propose a novel fully decentralized Reinforcement Learning (RL)-based solution that achieves fairness without coordination. The proposed Fair Share RL (FSRL) solution combines: (i) state augmentation with a semi-adaptive time reference; (ii) an architecture that leverages risk control and time difference likelihood; and (iii) a fairnessdriven reward structure. We evaluate FSRL in several network settings. Simulation results suggest that, when we compare FSRL with a common baseline RL algorithm from the literature, FSRL can be up to 89.0% fairer (as measured by Jain’s fairness index) in stringent settings with several sources and a single frequency band, and 48.1% fairer on average. 

The management of RF spectrum resources between heterogeneous RF devices has become more challenging with the advent of 5G, 6G and the desire to enable more spectrum sharing interactions in different bands. Most of the research on Dynamic Spectrum Access (DSA) algorithms considers non-cooperative scenarios with RF devices using omnidirectional antennas. In this paper, we study the effects of antenna directionality on cooperative DSA. Specifically, we develop a custom simulator for large-scale DSA networks that leverages IEEE 1900.5.2 Spectrum Consumption Models (SCMs) to enable coordination and computation of aggregate interference to deconflict spectrum use in large scale scenarios. SCMs offer a mechanism for RF devices to describe the characteristics of their use of spectrum and their needs in terms of interference protection. We create SCMs for RF systems with directional antennas based on measurements from a directional mmWave antenna and from the operational characteristics defined by the European Telecommunications Standards Institute (ETSI). We leverage these SCMs to perform a comparative analysis of spectrum use efficiency in cooperative DSA networks with up-to 300 links of transmitter-receiver RF devices using omnidirectional antennas vs similar networks using directional antennas with different half-power beam widths. The simulation results show the benefits to spectrum use efficiency that can be achieved with directional antennas and how largescale DSA methods can be studied and designed with the use of SCMs that incorporate detailed characteristics of directional antennas. 

This paper presents findings from an extensive 28 GHz mmWave measurement campaign conducted in New York City. The study includes over 20 million power measurements collected from two key scenarios: around-corner (non-line-ofsight due to building blockages) and same-street (nominally lineof-sight without obstructions from street furniture or foliage), covering over 1,300 unique links. For urban macro-cell (UMa) rooftop base stations above local clutter, the dominant angle of arrival (AoA) deviates by only 2 to 3.5 degrees from the direct transmitter/receiver direction. This small deviation allows for effective spatial separation between users, facilitating the future development of Multi-User MIMO algorithms for Beyond5G networks. In the urban micro-cell (UMi) dataset, with base stations below local clutter, a path gain drop of over 20 dB was observed in around-corner segments just 20 meters into a corner. Our Street-Clutter-NLOS path loss model achieves an RMSE of 6.4 dB, compared to 11.9 dB from NLOS 3GPP models. Using the best path loss model to estimate coverage for 90% of users traveling around corners, downlink rates could drop by over 10 times after 50 meters, highlighting the challenges in maintaining consistent user experience over mmWave networks in urban street canyons. 

Urban environments pose significant challenges to pedestrian safety and mobility. This paper introduces a novel modular sensing framework for developing real-time, multimodal streetscape applications in smart cities. Prior urban sensing systems predominantly rely either on fixed data modalities or centralized data processing, resulting in limited flexibility, high latency, and superficial privacy protections. In contrast, our framework integrates diverse sensing modalities, including cameras, mobile IMU sensors, and wearables into a unified ecosystem leveraging edge-driven distributed analytics. The proposed modular architecture, supported by standardized APIs and message-driven communication, enables hyper-local sensing and scalable development of responsive pedestrian applications. A concrete application demonstrating multimodal pedestrian tracking is developed and evaluated. It is based on the cross-modal inference module, which fuses visual and mobile IMU sensor data to associate detected entities in the camera domain with their corresponding mobile device.We evaluate our framework’s performance in various urban sensing scenarios, demonstrating an online association accuracy of 75% with a latency of ≈39 milliseconds. Our results demonstrate significant potential for broader pedestrian safety and mobility scenarios in smart cities. 

Vision Language models (VLMs) have transformed Generative AI by enabling systems to interpret and respond to multi-modal data in real-time. While advancements in edge computing have made it possible to deploy smaller Large Language Models (LLMs) on smartphones and laptops, deploying competent VLMs on edge devices remains challenging due to their high computational demands. Furthermore, cloud-only deployments fail to utilize the evolving processing capabilities at the edge and limit responsiveness. This paper introduces a distributed architecture for VLMs that addresses these limitations by partitioning model components between edge devices and central servers. In this setup, vision components run on edge devices for immediate processing, while language generation of the VLM is handled by a centralized server, resulting in up to 33% improvement in throughput over traditional cloud-only solutions. Moreover, our approach enhances the computational efficiency of off-the-shelf VLM models without the need for model compression techniques. This work demonstrates the scalability and efficiency of a hybrid architecture for VLM deployment and contributes to the discussion on how distributed approaches can improve VLM performance. Index Terms—vision-language models (VLMs), edge computing, distributed computing, inference optimization, edge-cloud collaboration. 

Recent advances in Visual Language Models (VLMs) have significantly enhanced video analytics. VLMs capture complex visual and textual connections. While Convolutional Neural Networks (CNNs) excel in spatial pattern recognition, VLMs provide a global context, making them ideal for tasks like complex incidents and anomaly detection. However, VLMs are much more computationally intensive, posing challenges for large-scale and real-time applications. This paper introduces EdgeCloudAI, a scalable system integrating VLMs and CNNs through edge-cloud computing. Edge- CloudAI performs initial video processing (e.g., CNN) on edge devices and offloads deeper analysis (e.g., VLM) to the cloud, optimizing resource use and reducing latency. We have deployed EdgeCloudAI on the NSF COSMOS testbed in NYC. In this demo, we will demonstrate EdgeCloudAI’s performance in detecting user-defined incidents in real-time. 

Backscatter power measurements are collected to characterize indoor radar clutter in monostatic sensing applications. A narrowband 28 GHz sounder used a quasi-monostatic radar arrangement with an omnidirectional transmit antenna illuminating an indoor scene and a spinning horn receive antenna offset vertically (less than 1 m away) collecting backscattered power as a function of azimuth. Power variation in azimuth around the local average is found to be within 1 dB of a lognormal distribution with a standard deviation of 6.8 dB. Backscatter azimuth spectra are found to be highly variable with location, with cross-correlation coefficients on the order of 0.3 at separations as small as 0.1 m. These statistics are needed for system-level evaluation of RF sensing performance. 

Backscatter power measurements are collected to characterize indoor radar clutter in monostatic sensing applications. A narrowband 28 GHz sounder used a quasimonostatic radar arrangement with an omnidirectional transmit antenna illuminating an indoor scene and a spinning horn receive antenna offset vertically (less than 1 m away) collecting backscattered power as a function of azimuth. Power variation in azimuth around the local average is found to be within 1 dB of a lognormal distribution with a standard deviation of 6.8 dB. Backscatter azimuth spectra are found to be highly variable with location, with cross-correlation coefficients on the order of 0.3 at separations as small as 0.1 m. These statistics are needed for system-level evaluation of RF sensing performance. 

We present a novel data-driven simulation environment for modeling traffic in metropolitan street intersections. Using real-world tracking data collected over an extended period of time, we train trajectory forecasting models to learn agent interactions and environmental constraints that are difficult to capture conventionally. Trajectories of new agents are first coarsely generated by sampling from the spatial and temporal generative distributions, then refined using state-of-the-art trajectory forecasting models. The simulation can run either autonomously, or under explicit human control conditioned on the generative distributions. We present the experiments for a variety of model configurations. Under an iterative prediction scheme, the way-pointsupervised TrajNet++ model obtained 0.36 Final Displacement Error (FDE) in 20 FPS on an NVIDIA A100 GPU.