Main research topics: V2X, C-ITS, VRU, Semantic Communications, Edge Computing, SL, FL

Our laboratory conducts research on next-generation wireless communications and intelligent transportation systems, focusing on Vehicle-to-Everything (V2X) communications, Cooperative Intelligent Transport Systems (C-ITS), Vulnerable Road User (VRU) safety, semantic communications, edge computing, Split Learning (SL) and Federated Learning (FL).

V2X (Vehicle-to-Everything)

V2X is a technology that exchanges information through wireless communication between vehicles (V2V), road infrastructure (V2I), pedestrians (V2P) and networks (V2N). By sharing surrounding traffic conditions and hazard information that are difficult to detect with a vehicle's own sensors, V2X communication supports autonomous driving and improves traffic efficiency and safety. We study communication technologies and protocols that support the efficient exchange of safety information in V2X environments.

C-ITS (Cooperative Intelligent Transport Systems)

C-ITS is an intelligent transportation system that provides traffic safety and driving assistance services based on real-time information sharing among vehicles, road infrastructure and pedestrians. We analyze traffic conditions using AI-based prediction and optimization techniques, and study the design and implementation of network architectures that support traffic safety services.

VRU (Vulnerable Road User)

VRUs are road users who are especially vulnerable in traffic accidents, such as pedestrians, cyclists and wheelchair users. Real-time exchange of safety information between vehicles and VRUs can be used to detect approaching vehicles and collision risks in advance and to prevent accidents. We study communication technologies that deliver safety messages between VRU devices and vehicles with low latency using Wi-Fi Aware-based short-range direct communication, and that support collision risk assessment and warning services.

Semantic Communications

Semantic communications extract and transmit only the meaning needed for the purpose of communication, reducing the amount of raw data transmitted and improving communication efficiency. In autonomous driving, feature information needed to determine the positions and moving directions of vehicles and pedestrians and to assess collision risks can be extracted from the large volumes of data collected by on-board sensors such as cameras and microphones. Based on semantic communications, we study system architectures that link an Agentic AI decision-making module and an AI-RAN controller to improve the situational-awareness reliability and communication efficiency of autonomous mobility services.

Edge Computing

Edge computing processes data on devices or network equipment close to where the data is generated, instead of sending it to a central cloud server. In autonomous driving and intelligent transportation systems, processing the large volumes of data generated by vehicles and road infrastructure on nearby edge devices reduces transmission latency and improves the responsiveness of real-time services. We apply edge computing and AI to vehicles and road infrastructure to analyze traffic conditions and to support vehicle safety and autonomous driving services that require low-latency communication.

SL (Split Learning)

Split learning divides an AI model into several parts so that devices and servers train it together by exchanging intermediate features and gradients. We study split learning methods that train efficiently by taking into account the computing resources and communication conditions of devices and servers. We also study ways to reduce the amount of transmitted data by applying semantic communications to the intermediate features exchanged during split learning.

FL (Federated Learning)

Federated learning is a technique in which multiple devices train AI models locally and share only the learned model parameters to train a model jointly, without sending raw data to a central server. It protects data privacy and reduces the amount of data sent to the central server. Through cooperative learning among vehicles and edge devices, we study how to make efficient use of distributed data and improve both the training performance of AI models and communication efficiency in V2X environments.