Projecte ofert
Títol: Reinforcement Learning for Adaptive Switching Between OTFS and OFDM
Per assignar-vos el projecte us heu de dirigir al director/a perquè us l'assigni.
Director/a: ALONSO ZÁRATE, LUIS
Departament: TSC
Títol: Reinforcement Learning for Adaptive Switching Between OTFS and OFDM
Data inici oferta: 04-07-2025 Data finalització oferta: 04-03-2026
Estudis d'assignació del projecte:
MU AI4CI- MU DRONS
- MU EM CODAS 1
- MU EM CODAS 2
- MU MASTEAM 2015
| Tipus: Individual | |
| Lloc de realització: EETAC | |
| Segon director/a (UPC): CHAUDHRI, SULEMAN FAROOQ | |
| Paraules clau: | |
| Wireless communications, OFDM, OTFS, Doppler, Mobility, Performance | |
| Descripció del contingut i pla d'activitats: | |
| This Master's thesis project explores the application of reinforcement learning (RL) to dynamically optimize waveform selection between Orthogonal Frequency Division Multiplexing (OFDM) and Orthogonal Time Frequency Space (OTFS) in high-mobility, beyond-5G communication systems. OFDM, though widely adopted in current standards such as LTE and 5G, suffers performance degradation under high Doppler shifts, making it less suitable for rapidly changing environments like vehicular or UAV-based communications. OTFS, which operates in the delay-Doppler domain, offers higher robustness in such conditions but is computationally more complex and not yet fully integrated into legacy infrastructure. This project aims to bridge that gap by developing an intelligent, adaptive system that switches between OTFS and OFDM in real-time based on observed network conditions such as Doppler spread, user velocity, and channel variability.
The core objective of this project is to build a simulation environment that mimics real-world wireless conditions where channel dynamics and user mobility constantly change. Using this environment, the student will implement a reinforcement learning agent-initially with a simple Q-learning algorithm and later extending to more advanced deep reinforcement learning techniques like Deep Q-Networks (DQN). The agent's goal will be to observe environment states (e.g., estimated Doppler shift, SNR, or latency constraints) and learn a policy that selects the optimal waveform (OTFS or OFDM) to maximize performance. Performance will be evaluated using key metrics such as Bit Error Rate (BER), Spectral Efficiency, and Latency, MATLAB will be used to implement and train the RL models. This project provides a multidisciplinary platform combining wireless communications, signal processing, and machine learning, making it ideal for students aiming to gain expertise in the development of intelligent 6G systems. By the end of the thesis, the student will have built a complete reinforcement learning-based adaptive communication module, a robust s, and a technical report documenting the performance of the system across various mobility and channel conditions. The project emphasizes both practical implementation and conceptual understanding, offering a rich learning experience that can serve as a foundation for future research or industry applications. Optional extensions may include scaling the system to multi-user MIMO scenarios, integrated beamforming strategies, or evaluating real-time inference performance for lightweight deployment in embedded or edge-computing environments. Given its technical depth and research relevance, the project aligns well with cutting-edge trends in AI-native 6G, Integrated Sensing and Communication (ISAC), and autonomous wireless systems. |
|
| Orientació a l'estudiant: | |
| Requereix activitats hardware: No | |
| Requereix activitats software: Sí Sistema operatiu: Disc (Gb): | |
| Horari d'atenció a estudiants per a l'assignació de projecte: |
|