Date : 12/05/2026
Hour: 14.00
Student Name & ID: Ebrar Sude Başer & 255105114
Supervisor: Dr. Öğr. Üyesi Metin Öztürk
Topic: Learning-Based Adaptive Mode Selection in Heterogeneous Data Traffic
Online Meeting Link : A-424
Abstract: Efficient resource use is severely hampered by the growing volume of data and the extensive usage of multimodal data sources in contemporary wireless communication systems. Traditional methods rely on rule-based and static transmission schemes, which are not flexible enough to adjust to changing network circumstances and traffic loads. Specifically, treating all data types equally frequently results in inefficient use of capacity and possible loss of important information. This work explores a semantic communication-based method in which data is dynamically conveyed utilizing various modes and represented at the semantic level. Depending on how important the data is, the system can adaptively switch between several transmission modalities, such as different compression levels, resolutions, or semantic abstractions. A state-action-based Q-learning algorithm is used in a reinforcement learning-based framework to facilitate adaptive decision-making. To identify the best transmission mode in real time, the model takes into account factors including network capacity, traffic volume, and semantic significance. By combining semantic representation with learning-based mode selection methods, the suggested method seeks to improve data transmission flexibility, maximize resource consumption, and guarantee the preservation of crucial information