Doctoral Candidate

Pi-Wei Chen, MSc

Pi-Wei Chen received his Master’s degree in Manufacturing Information and Systems from National Cheng Kung University, with background in managing and advancing complex AI projects. His research expertise spans areas like anomaly detection, collaborative sensor fusion, and federated learning, closely aligning with the TUAI project’s goals.

PW_CHEN_Self-presentaion.pdf

Main Supervisor: Jerry C.-W. Lin (SUT)

Co-Supervisors: Jia-Chun Lin (NTNU), Francesco Piccialli (UNINA), Vicente García Díaz (UNIOVI)

R&D cooperation: AIUT

Objectives: develop novel methodologies to advance perception and vision-based decisionmaking in industrial and AGV-driven environments. 

The project focuses on high-accuracy recognition of objects and obstacles, event detection, and robust adaptation to diverse and dynamic operating scenarios. By fusing multisensor data into a unified perception framework, the system will deliver reliable, real-time environmental awareness to support safe and efficient autonomous navigation.

The feature fusion approach will be used to ensure that critical perception tasks, such as object detection, tracking, and scene understanding, are accurate and explainable in their operational context. Collaborative sensor fusion will provide a framework to ensure consistency and reliability across sensing modalities, while also identifying cases of uncertainty or conflicting data. Measures of accuracy, robustness, and efficiency are essential to evaluate these aspects and ensure that the perception outputs are meaningful and trustworthy.

Research challenges include balancing high perception accuracy with limited computational and energy resources, ensuring adaptability across heterogeneous environments, and developing sensor fusion algorithms that communicate effectively with higher-level AI planning systems.

Expected Results: include a fully integrated perception framework with real-time sensor collaboration,

enabling continuous fault-tolerant operation even if individual sensors fail. The system will support reliable human–

machine interaction (e.g., operators monitoring AGVs) and seamless integration with logistics management software, collaborative robots, and Manufacturing Execution Systems in autonomous and green manufacturing environments.

Applied research: will focus on novel methods for perception-driven sensor fusion and online environmental monitoring, enabling deeper knowledge acquisition of intralogistics and navigation processes. Additionally, methods for agile sensor configuration and communication with third-party systems have the potential to expand the portfolio of technologies used by AIUT.

Planned secondments: NTNU (4 months); UNIOVI(4 months); HVL (4 months)                    

Enrolment in Doctoral degree: SUT