UAV Swarm-Based Antenna Design for Intelligent Edge (iEDGE): The proposed UAV swarm-based antenna design integrates seamlessly with Intelligent Edge (iEDGE) systems to enhance wireless communication and sensing capabilities in dynamic and distributed environments. By leveraging a coordinated UAV swarm to form reconfigurable antenna arrays, such as Yagi-Uda configurations, this approach enables high-gain, directional communication tailored to the needs of iEDGE systems. The mobility and adaptability of the UAV swarm allow the iEDGE to dynamically adjust its communication links in response to real-time environmental and operational changes. This project aims to research and develop an efficient, scalable, and cost-effective UAV swarm-based antenna solution to support the high-performance requirements of iEDGE systems in applications like edge computing, autonomous networks, and mission-critical IoT deployments.
Optimizing Generative AI Deployment in Intelligent Edge (iEDGE) Networks: Generative AI services are transforming digital content creation, but deploying these models efficiently in mobile edge networks faces significant challenges. Edge servers promise reduced latency and bandwidth usage compared to cloud solutions but struggle with resource limitations such as storage, GPU memory, and energy. The heterogeneity of generative AI models, including their resource demands and delay factors, adds complexity to deployment decisions. Moreover, dynamic users request further complicated effective resource allocation. This project aims to address these challenges by designing an optimized adaptive, collaborative edge-cloud/ edge-edge framework for deploying generative AI models. The goal is to minimize service delay while balancing resource constraints through a feature-aware algorithm.
Environment-Aware Mobile Edge Computing: This project enables students to practice cutting-edge edge computing techniques on Jetson edge devices and Hiwonder robotic cars, focusing on how environmental factors like temperature and wind speed affect mobile edge devices. These mobile edge devices experience fluctuating computing capacity due to environmental conditions, with GPU temperatures rising during compute-intensive tasks, causing performance degradation and increased latency. Students will evaluate these effects and develop an environment-aware iEDGE system that dynamically adjusts computing power based on real-time environmental data to minimize inference latency. This hands-on experience prepares students for real-world challenges in edge computing while advancing adaptive, sustainable mobile edge systems.
Scalable Real-Time Scheduling for EV Charging using Intelligent Edge (iEDGE): The rapid growth in plug-in electric vehicles (PEVs) has introduced challenges in managing their charging demands efficiently. Uncontrolled charging can lead to grid congestion, demand peaks, and power quality issues, threatening the reliability of power systems. Current methods for real-time scheduling at charging stations (CS), including heuristic rules, decentralized approaches, and learning-based techniques, and face limitations. These include scalability issues, lack of systematic performance guarantees, and difficulty adapting to dynamic and uncertain PEV charging demands. This project aims to propose an efficient and scalable real-time PEV charging scheduling mechanism. By leveraging renewable energy sources, the solution will dynamically allocate charging resources to minimize operation costs while addressing constraints like energy variability and charging station capacity. The project will simulate the PEVs and their charging behavior. The goal is to propose cost-effective, real-time scheduling that adapts to the dynamic nature of PEV requests and supports sustainable energy practices.
AI-RAN: Testbed Integration of O-RAN with Edge Computing: This project will integrate the Edge Radio Access Network Intelligent Controller (EdgeRIC) [https://edgeric.github.io/] into the Open Artificial Intelligence Cellular (OAIC) platform while leveraging software-defined radios (SDRs) to analyze and optimize application-specific quality of service (QoS). Integrating EdgeRIC’s real-time capabilities with OAIC’s near-real-time RIC allows intelligent edge processing to be overseen and guided by AI controllers. Using the current version of srsRAN employed by EdgeRIC, the experiment will evaluate different applications, such as VoIP, gaming, and video streaming, to classify these applications using AI. Transmission parameters can be dynamically adjusted by identifying the application type to support QoS flows better. The laboratory testbed at MSU that will be deployed for this project will use smartphones and SDR-based user equipment for data collection and testing to develop reproducible experiments. This project will extend prior O-RAN Edge computing research, produce new results and data, and enhance community research infrastructure.
Trust-Aware Task Offloading for Connected Vehicles: Connected vehicle networks are revolutionizing transportation by enabling vehicles to communicate with each other and with infrastructure. However, the dynamic and heterogeneous nature of connected vehicle environments presents significant challenges in ensuring reliable and efficient task offloading for real-time applications. These challenges are worsened by malicious behaviors, such as spoofing attacks, collusion attacks, and false data injections, which can compromise system reliability and safety. Furthermore, current task offloading strategies for connected vehicles cannot adapt dynamically to varying trust levels, resource availability, and application requirements, resulting in inefficient resource utilization and increased latency. This project aims to develop a robust, adaptive, trust-based task offloading strategy tailored for connected vehicle networks. The proposed solution will ensure a reliable and efficient delivery service by incorporating dynamic trust evaluation, resource clustering, and task allocation mechanisms. The project will address key challenges, including identifying malicious nodes, optimizing resource allocation, and minimizing system costs. The goal is to enhance connected vehicle networks’ safety, reliability, and performance in complex and dynamic environments.
Collective Learning in Intelligent Edge (iEDGE): Within iEDGE, an IoT cluster may consist of various mobile devices, such as cellular phones, sensors, vehicles, etc., each with diverse computing capabilities. These IoT devices, which contain different data with different distributions, would like to jointly complete a learning task over a network through an edge server. The performance of this collective learning over a network highly depends on how the local learning results of the IoT devices can be effectively collected. The project aims to investigate how to select the local learning results of IoT devices to help IoT devices achieve fast learning over a network.