Intelligent Transportation Systems (ITS) demand real-time collision prediction to ensure road safety and reduce accident severity. Conventional approaches rely on transmitting raw video or high-dimensional sensory data from roadside units (RSUs) to vehicles, which is impractical under vehicular communication bandwidth and latency constraints. In this work, we propose a semantic V2X framework in which RSU-mounted cameras generate spatiotemporal semantic embeddings of future frames using the Video Joint Embedding Predictive Architecture (V-JEPA). To evaluate the system, we construct a digital twin of an urban traffic environment enabling the generation of d verse traffic scenarios with both safe and collision events. These embeddings of the future frame, extracted from V-JEPA, capture task-relevant traffic dynamics and are transmitted via V2X links to vehicles, where a lightweight attentive probe and classifier decode them to predict imminent collisions. By transmitting only semantic embeddings instead of raw frames, the proposed system significantly reduces communication overhead while maintaining predictive accuracy. Experimental results demonstrate that the framework with an appropriate processing method achieves a 10% F1-score improvement for collision prediction while reducing transmission requirements by four orders of magnitude compared to raw video. This validates the potential of semantic V2X communication to enable cooperative, real-time collision prediction in ITS.

The integration of unmanned aerial vehicles (UAVs) with mobile edge computing (MEC) and Internet of Things (IoT) technology is crucial for efficient resource management and sustainable agricultural productivity in smart frames. This paper addresses the critical need for optimizing task offloading in secure UAV-assisted smart farm networks, aiming to reduce total delay and energy consumption while maintaining robust security in data communications. We propose a multi-agent deep reinforcement learning (DRL)-based approach using a deep double Q-network (DDQN) with an action mask (AM), designed to manage task offloading dynamically and efficiently. Simulation results demonstrate the superior performance of our method in managing task offloading, highlighting significant improvements in operational efficiency, such as reduced delay and energy consumption. This aligns with the goal of developing sustainable and energy-efficient solutions for next-generation network infrastructures, making our approach an advanced solution for performance and sustainability in smart farming applications.

In precision farming, timely information is vital for effective decision-making in irrigation and pest control processes. Unmanned Aerial Vehicles (UAVs) and Multi-access Edge Computing (MEC) servers optimize data collection from ground sensors. However, integrating sensors, controllers, and actuators complicates data freshness in closed-loop communication. Computational resource optimization is also crucial, especially in energy-constrained equipment and competitive resource scenarios. This work optimizes data freshness and task turnaround time (TAT) with a UAV trajectory and MEC offloading strategy. Analysis includes assessing delays in uplink, downlink, and queues. This gains significance under dynamic network conditions and variable resources. A hierarchical deep reinforcement approach is proposed. Numerical results indicate the proposed solution outperforms baselines, improving data freshness by up to $31 \%$ and simultaneously decreasing TAT by $34 \%$.

Edge computing offers a promising solution to enhance network reliability. In this study, we investigate the integration of mobile edge computing (MEC) technology and unmanned aerial vehicles (UAVs) within the context of smart agriculture. Smart agriculture relies on resource-constrained Internet of Things (IoT) devices for local environmental monitoring and data collection. These IoT devices send the collected data to UAVs for analysis. A central theme of this work is the focus on the applications generated by each UAV and the consideration of their topology to derive our optimization algorithm. To tackle these challenges, we propose harnessing the computational and power resources of UAVs and MEC at the network’s edge to offload and execute resource-intensive tasks in UAV-MEC-assisted networks. Our research focuses on the joint optimization of power allocation and task offloading in these wireless networks. Central to our investigation is the problem of minimizing the energy-time cost (ETC) for the UAVs, considering the interdependencies among tasks. To address this complex problem efficiently, we introduce graph convolutional neural networks (GCNs) and reinforcement learning (RL)-based techniques. We employ a directed acyclic graph (DAG) to model task interdependencies, with GCNs characterizing the DAG. Our approach incorporates an actor-critic method with embedding layers, trained using the compound-action actor-critic (CA2C) algorithm. Our findings reveal a significant improvement in minimizing both delay and energy consumption, with a 27% percent reduction in delay and a 45% reduction in consumed energy for executing complex, interdependent tasks.

Sentiment Analysis Systems (SASs) are data-driven Artificial Intelligence (AI) systems that assign one or more numbers to convey the polarity and emotional intensity of a given piece of text. However, like other automatic machine learning systems, SASs can exhibit model uncertainty, resulting in drastic swings in output with even small changes in input. This issue becomes more problematic when inputs involve protected attributes like gender or race, as it can be perceived as bias or unfairness. To address this, we propose a novel method to assess and rate SASs. We perturb inputs in a controlled causal setting to test if the output sentiment is sensitive to protected attributes while keeping other components of the textual input, such as chosen emotion words, fixed. Based on the results, we assign labels (ratings) at both fine-grained and overall levels to indicate the robustness of the SAS to input changes. The ratings can help decision-makers improve online content by reducing hate speech, often fueled by biases related to protected attributes such as gender and race. These ratings provide a principled basis for comparing SASs and making informed choices based on their behavior. The ratings also benefit all users, especially developers who reuse off-the-shelf SASs to build larger AI systems but do not have access to their code or training data to compare.

A deep reinforcement learning technique is presented for task offloading decision-making algorithms for a multi-access edge computing (MEC) assisted unmanned aerial vehicle (UAV) network in a smart farm Internet of Things (IoT) environment. The task offloading technique uses financial concepts such as cost functions and conditional variable at risk (CVaR) in order to quantify the damage that may be caused by each risky action. The approach was able to quantify potential risks to train the reinforcement learning agent to avoid risky behaviors that will lead to irreversible consequences for the farm. Such consequences include an undetected fire, pest infestation, or a UAV being unusable. The proposed CVaR-based technique was compared to other deep reinforcement learning techniques and two fixed rule-based techniques. The simulation results show that the CVaR-based risk quantifying method eliminated the most dangerous risk, which was exceeding the deadline for a fire detection task. As a result, it reduced the total number of deadline violations with a negligible increase in energy consumption.

Unmanned aerial vehicles (UAVs) with mounted base stations are a promising technology for monitoring smart farms. They can provide communication and computation services to extensive agricultural regions. With the assistance of a Multi-Access Edge Computing infrastructure, an aerial base station (ABS) network can provide an energy-efficient solution for smart farms that need to process deadline critical tasks fed by IoT devices deployed on the field. In this paper, we introduce a multi-objective maximization problem and a Q-Learning based method which aim to process these tasks before their deadline while considering the UAVs’ hover time. We also present three heuristic baselines to evaluate the performance of our approaches. In addition, we introduce an integer linear programming (ILP) model to define the upper bound of our objective function. The results show that Q-Learning outperforms the baselines in terms of remaining energy levels and percentage of delay violations.

We will demonstrate features and capabilities of the PhantomNet testbed. PhantomNet is a mobile testbed, at the University of Utah, aimed at enabling a broad range of mobile networking related research. PhantomNet is remotely accessible and open to the mobile networking research community.

We present Proteus, a mobile network service control platform to enable safe and rapid evolution of services in a mobile software defined infrastructure (SDI). Proteus allows for network service and network component functionality to be specified in templates. These templates are used by the Proteus orchestrator to realize and modify service instances based on the specifics of a service creation request and the availability of resources in the mobile SDI and allows for service specific policies to be implemented. We evaluate our Proteus prototype in a realistic mobile networking testbed illustrating its ability to support service evolution.

Network management tasks remain tedious and error-prone, and often require complex reasoning on the part of the network administrator. With KnowNet we address the challenge of reasoning about network management by approaching it as a set of cooperating applications executing over a knowledge graph which captures data and information about the network and the applications that manage and reason over it. We apply our approach to enterprise network management by developing a suite of cooperating applications that deals with security and application performance management in an enterprise network.

We present an algorithm that takes measurements of the packet Reorder Density (RD) metric and generates reordering sequences. These sequences can be used by a simulator or emulator to precisely and repeatably reorder packets in a way that recreates the original RD. We show that our algorithm is efficient for a range of realistic reordering scenarios, and present an extension to the Dummynet emulator that uses makes use of it.