WAISL · DGIST
Research
Our primary interests are AI-native wireless systems, latent communications for edge intelligence, and wireless digital twins. Our broader work also includes non-terrestrial networks.
AI-native wireless
Learning-based radio control and network operation.
We investigate how machine learning can improve the operation of wireless infrastructure. Our work covers online channel synthesis, learning-based beam management, distributed MIMO, and cross-layer intelligence for AI-RAN architectures. The focus is on estimating channels and controlling radio resources under practical system constraints.
RELATED WORK
Latent communications for edge intelligence
Learned signaling for low-latency agent communication.
We study deep-learning-driven signaling for exchanging latent representations in settings including agent-to-agent and base-station-to-agent communication. The focus is on low-latency exchange over noisy, time-varying channels, involving participants that may be mobile or have limited computing or communication resources. This direction emphasizes learned representations, transmitters, and receivers, complementing the radio and network control problems studied in AI-native wireless.
RELATED WORK
Wireless digital twins
RF environment modeling and physical-system calibration.
Our research investigates RF environment representations and phase-coherent digital twin calibration. We combine channel modeling, foundation models, theoretical analysis, and prototype validation to connect simulation with physical wireless systems. These representations also support site-specific learning and evaluation of wireless algorithms.
RELATED WORK
Non-terrestrial networks
Satellite, UAV, and high-altitude platform communications.
Our broader research includes resource allocation and wireless links in terrestrial and non-terrestrial networks. Previous and ongoing work spans satellite-assisted urban air mobility, massive field-data mapping with HAPS, and pointing and trajectory optimization for free-space optical communications.
RELATED WORK
METHODS
Research approach
We combine mathematical analysis, learning-based algorithms, simulation, and prototype validation. Across the four areas, we study both algorithmic performance and the constraints of physical wireless systems.
Publications