Students of Computer Science, Computer Engineering and Information Systems at CIn can join through undergraduate research.
Openings
Come and decide what the network will do
We are looking for master's and doctoral students to work on 5G and beyond (NextG) networks: AI-orchestrated Open RAN, aerial networks and observability.
Why here
The conditions to carry an idea through, and a publication record that opens doors
- Analytical modeling, simulation and testbed, with software-defined radio, Open RAN, network slicing and 5G and beyond core.
- Publication in IEEE Transactions, IEEE Network, ACM Computing Surveys and Computer Networks, and at national and international events of the field.
- Projects funded by CNPq, CAPES, FACEPE and FINEP.
- Forty master's and doctoral graduates since 2009, now in universities, operators and industry.
Open topics
Eight fronts
-
Autonomous networks from the RAN to the core
An agent perceives the network, reasons about its state, plans, and decides, using the open interfaces of Open RAN and the service-based core. In the Internet of Vehicles, multiple agents already decide whether a vehicle talks directly to its neighbor over sidelink or goes through the base station, cutting loss and latency. Open problems: conflicts between intents, guardrails that audit every decision, and digital twins that test a reconfiguration before it goes live.
An O-RAN-Assisted MARL Approach for Dynamic Sidelink and Infrastructure Selection in V2X Communications · IEEE Transactions on Vehicular Technology, 2026
-
Agentic AI in the mobile core
One agent plans and another critiques and executes. Language models translate intent into action over the core, in a closed loop that corrects itself by the real state of the network, in place of the centralized analytics that 3GPP offers today.
AgentxGCore: Agentic AI for Next-Generation Mobile Core Network · IEEE Network, 2026
-
Routing under uncertainty in the transport network
Critical traffic crosses several routers between the user and the application, even with the edge nearby. Segment Routing defines the path, and a bandit model chooses which one, learning from the measured loss and latency.
Segment Routing Path Optimization for URLLC in NextG Mobile Transport Network · IEEE Transactions on Network and Service Management, 2026
-
xApps and rApps for mobility
xApps running on the near-real-time RIC, deciding the handover before the base station change interrupts the video or the car's update. Compared against the standard 3GPP procedure, over real urban traffic data.
Open RAN-Enabled Deep Learning-Assisted Mobility Management for Connected Vehicles · ICOIN, 2025
-
Aerial networks, 5G and beyond
The base station rises into the air and takes the problem with it: a smaller cell, more ping-pong effect and neighboring drones interfering. Predicting trajectory and signal is what decides the handover, and weight and battery limit how much processing fits on board.
Aero5GBS · IEEE Access, 2023 · Resource allocation for UAV-enabled multi-access edge computing · Journal of Supercomputing, 2024
-
Network observability and analytics
The NWDAF observes the data and control planes in real time. The next question is what to do with that: scale a network function after the load arrives, as Kubernetes autoscaling does, or before, by predicting the curve.
Monitoramento dos planos de dados e controle 5G utilizando NWDAF · SBrT, 2025 · Reagir ou antecipar? HPA e ML para balanceamento de carga · SBrT, 2025
-
Slicing that adjusts and defends itself
Opportunistic slices use the spectrum that the primary ones leave free, and deciding which resources to allocate is an NP-hard problem. On the other side, the slice has to defend itself: a neighboring slice is a new attack surface.
A real-time attack defense framework for 5G network slicing · Software: Practice and Experience, 2020 · A Fuzzy-Genetic Approach for 5G/6G Opportunistic Slicing · LATINCOM, 2021
-
Energy and QoS in open infrastructure
The RAN consumes more energy than anything else in a 5G network, and the existing ways to save it are proprietary solutions for switching radios on and off. Open RAN makes it possible to base that decision on the real state of the network, without breaking the QoS of URLLC and eMBB.
SEQUOIA: solução para energia e qualidade de serviço otimizadas em infraestruturas abertas · SBRC, 2025
Profile
Someone who enjoys formulating, simulating and measuring an autonomous network
- A background in computer networks, and comfort with probability, optimization or machine learning, depending on the topic.
- Hands-on experience with programming, Linux, containers and simulators. Nobody has to arrive knowing ns-3 or OMNeT++, but they do have to want to learn.
- English to read papers and write your own.
- A willingness to publish. The work is only finished when the research community reads it.
How to join
The path, in three steps
-
Before
Write
Send an email to with your CV, saying which of the topics interests you and why. Being specific is worth more than a generic letter. This is the moment to get to know the research areas better and to ask questions.
-
Selection
Apply to the program
Admission to the master's and doctoral programs is through the Graduate Program in Computer Science at CIn-UFPE, which runs its own admissions process and schedule at secpos.cin.ufpe.br.
-
Funding
Financial support
Scholarships through CNPq, CAPES and FACEPE. Availability varies by call and by year.
Undergraduate
There is undergraduate research too
Contact
Write and tell us which topic is yours
Prof. Kelvin Lopes Dias · Centro de Informática, UFPE.