4-B. Capacity-constrained Distributed Positioning

Goal: To achieve accurate user positioning through distributed access points under limited fronthaul capacity

Approach: To perform local signal processing at each access point and transmit compressed positioning information for centralized fusion

Challenge: Development of central fusion of distributed decisions to optimize position accuracy

1-A. DRL-based Resource Allocation with Partial Observation

Goal: To optimize wireless resource allocation under incomplete and imperfect channel state information

Approach: To leverage learning-based agents that infer latent system states from partial observations for decision-making

Challenge: Partial observability and uncertainty fundamentally complicate state estimation and stable policy learning

1-B. AI-based Channel Estimation

Goal: To accurately estimate wireless channels under limited pilot overhead and rapidly time-varying environments

Approach: To exploit learning-based models that capture latent channel structures from pilot signals and extracted channel features

Challenge: Generalization and robustness under mobility, noise, and model mismatch beyond idealized channel assumptions

2-A. Transciever Design for Federated Learning

Goal: To train a global model collaboratively without sharing raw data across distributed wireless devices

Approach: To exchange local model updates over wireless links and aggregate them at a central server for global learning

Challenge: Wireless channel impairments, privacy leakage risks, and excessive communication overhead fundamentally limit learning accuracy and scalability

3-A. Binary Neural Network for Model Compression

Goal: To significantly reduce model size and inference complexity for efficient deployment on resource-constrained devices

Approach: To apply aggressive model compression techniques such as binary neural networks and low-bit quantization

Challenge: Severe performance degradation and training instability caused by extreme quantization