Abstract: In the future, wireless systems are expected to support diverse traffic patterns, dense multi-cell deployments, and high UE mobility, causing conventional wireless procedures less effective. This work addresses two such problems from the UE perspective: adaptive power saving through connected-mode discontinuous reception (C-DRX), and joint beam and base station (BS) prediction in multi-cell high-mobility scenarios. To improve energy efficiency, a reinforcement learning (RL)- based framework is developed to select suitable C-DRX configurations based on traffic characteristics. Both context-free and contextual bandit methods are studied and compared with the conventional static configuration method. Simulation results show that the proposed learning-based method can outperform conventional one by reducing UE power consumption while maintaining throughput performance. The second part of the work is position-aided inter-cell beam prediction for high-mobility multi-cell networks. A transformer-based framework is proposed to jointly predict the future serving base station and its optimal beam using historical beam information, serving BS history, and UE positional data. Simulation results based on a ray-tracing-based urban dataset show that incorporating position information significantly improves joint BS-beam prediction accuracy by 16% compared to those without position data, while remaining robust under localization errors. These two works together demonstrate the potential of AI/ML based methods for UE-centric processes in 6G with improved energy efficiency, reduced signaling overhead, and enhanced mobility support.

Event Details
Title: AI/ML-Based C-DRX for Energy Efficient UE and Position-Aided Beam Prediction for High Mobility Multi-cell Networks
Date: April 15, 2026 at 2:00 PM
Venue: ESB 244 / Google Meet (http://meet.google.com/bqc-pope-rgg)
Speaker: Mr. SAYAN RUDRA PAL (EE23S046)
Guide: Dr. Radha Krishna Ganti
Type: MS seminar

Updated: