Build an intelligent assistant that ingests live telemetry from network devices and uses a Large Language Model (LLM) to diagnose and suggest fixes for faults. It should detect anomalies like latency spikes, link failures, or misconfigurations and present actionable recommendations in natural language.

The Industrial Internet of Things (IIoT) holds immense potential for transforming manufacturing and facilitating "factories of the future." However, IIoT applications usually demand ultra-high connection density, ultra-low communication latency, and ultra-high reliability, which presents significant challenges to existing networks. 5G addresses ultra-reliable low-latency communications and massive machine-type communications as separate cases and thus cannot simultaneously meet the stringent connection density, latency, and reliability demands of IIoT. This project aims to advance medium access control (MAC) for IIoT in the context of 6G networks by investigating the role of graph neural networks (GNNs) in access scheduling. Specifically, by representing access points (APs) in an IIoT network as graph nodes in a GNN, the team aims to enable APs to exchange information and intelligently coordinate their MAC settings.

5G is the fifth generation of mobile networks. The objective of this project is to implement and deploy a virtual 5G network using open-source tools. The network will include multiple 5G slices that will run network services (e.g. billing, ISP management tools) as well as user services (video streaming, connected cars support) over secure and isolated slices.
Network slices in a 5G network allow running multiple virtual networks on shared physical resources. The students are expected to deploy and run 5G network slices using open-source tools and perform experiments using different user traffic types to evaluate the performance. The tools students are required to use include Free5GC [1] and Containerlab [2]. Free5GC is a Linux-based open-source tool that allows the deployment of a 5G network. Containerlab is an open-source tool from Nokia that enables the deployment of container-based custom network topologies. It supports several network Operating Systems (OSes) from Nokia, Juniper, Cisco, SONiC, etc.

An indoor localization and tracking system, also known as an Indoor Positioning System (IPS) or Real-Time Location System (RTLS), is technology that determines and monitors the location of people and objects within buildings or confined spaces where outdoor GPS signals fail. These systems use a combination of wireless communication technologies like Wi-Fi and Bluetooth, ultrasonic or infrared signals, and sensors to provide real-time location data, enabling applications from asset tracking and inventory management to indoor navigation and emergency response. In this project, the team builds an indoor localization and tracking system for health care. This includes localization and tracking for patients and medical equipment.

Create a virtual replica of a physical or simulated network that can model topology, routing, and traffic patterns. The twin will allow “what-if” experiments—such as link failures or routing changes—to predict performance impacts before changes are applied.

A smart home security system is a network of connected electronic devices, such as cameras, sensors, and smart locks, that work together to monitor and protect a home. These systems can be controlled and monitored remotely through a mobile app, allowing users to receive real-time alerts, view live video feeds, and manage various security and even home automation features from anywhere. In this project, you will build a smart home security system with camera and intrusion detection sensors (such as magnetic reed switches for doors/windows or PIR - passive infrared - sensors for motion detection). You also develop a smart phone app or a web application allowing user to receive notifications, view live video, etc.

The advancement of machine learning provides new approaches to training defense agents for enhanced network security. In this project, we extend existing results on reinforcement learning (RL)-based single defense agent training for TTCP CAGE Challenge 2 to the case of multi-agent RL (MARL) for TTCP Cage Challenge 4, which involves multiple defense agents distributed over different segments of a network. Experiments will be conducted on the Cyber Operations Research Gym (CybORG), an experimental platform using the OpenAI Gym interface.

In modern warehouses, robots play a critical role in accomplishing tasks assigned by a central controller. These robots maintain constant communication with the warehouse controller via access points (APs) connected to routers or local area networks (LANs). This project involves emulating robot mobility in a warehouse environment using the SONiC FRRouting platform. Robots will be treated as hosts, while the IP fabric will utilize eBGP underlay with a VxLAN-EVPN control plane. The project's goal is to 1) Characterize current mobility techniques and 2) Develop and evaluate new techniques based on criteria such as warehouse size, robot movement rate, communication reliability, and BGP route management.