My research contributions and academic publications
Wu, K., Ni, K., Xu, H., Wang, J., Qin, H., Zhang, X., & Zhang, X.
IEEE Transactions on Microwave Theory and Techniques
2025
We propose a physics-informed long-range attentive propagation network (PLAPN) with a sliding window strategy to accurately and efficiently predict long-range radio wave propagation in tunnels from short-range simulation data.
Wu, K., Zhao, Q, Feng, Z., Qin, H., Zhang, X., & Zhang, X.
IEEE Transactions on Intelligent Transportation Systems
2025
We develop a Deep Reinforce Learning (DRL)-driven framework combining physics-based PWE simulation, conditional Generative Adversarial Network (cGAN)-based data augmentation, and Dueling Deep Q-Network (DQN) optimization to achieve accurate, efficient, and robust access point placement for CBTC tunnel systems.
Wu, K., Zheng, D., Zhu, Y., Ye, Z., & Lin, Y.-S.
Materials Today Communications
2025
We designed a MEMS-tunable π-shaped perfect absorber in the mid-infrared (MIR) range that enables dynamic dual-resonance control, polarization switching, and high-sensitivity refractive index sensing for advanced photonic applications.
Wu, K., Ni, K., Chen, L., Xu, H., Wang, J., Zhou, J., & Zhou, X.
Sensors
2025
We developed an (Transformer-CNN-based) ATTRN machine learning model that directly predicts boiler furnace temperature fields from acoustic measurements (Time of flight), improving accuracy, generalization, and industrial applicability over traditional methods.
Wang, J., Wu, K., & Ouyang, Y.
Information
2025
This paper introduces a Verifiable Emotion Reward (VER) to reduce emotional drift in role-playing with LLMs. We also construct CHARCO, a 230k multi-turn dataset with persona and emotion labels, and design a retrieval-augmented memory module. Experiments show improved emotional consistency, role fidelity, and dialogue coherence.
Ye, Z., Zheng, D., Zhu, Y., Wu, K., & Lin, Y.-S.
APL Materials
2025
Analyzed KI-shaped terahertz resonators, guiding parameter optimization to achieve tunable filtering, signal suppression, and refractive index sensing performance.
Zhou, J., Zhao, Q, Feng, Z, Wu, K., Zhang, L., & Qin, H.
Sensors
2025
We propose a trajectory-based localization framework using ToA and FoA with an enhanced PSO solver. The method improves accuracy and robustness in asynchronous acoustic networks, keeping over 90% of errors within 3 m.
Wu, K., Ni, K., Xu, H., Wang, J., Qin, H., Zhang, X., & Zhang, X.
IEEE MTT-S International Wireless Symposium
2025
Presented a temporal attention model for EM wave propagation in tunnels, driven by the vector parabolic equation. Manuscript accepted.
Wu, K., Zhao, Q, Feng, Z., Qin, H., Zhang, X., & Zhang, X.
IEEE 2nd International Conference on Deep Learning and Computer Vision
2025
Introduced conceptor-augmented regularization to improve the generalization of RNN and LSTM networks.
Wang, J., Zhang, Z., He, Y., Song, Y., Shi, T., Li, Y., Xu, H., Wu, K., Qian, G., Chen, Q., & He, L.
2025
Survey on improving code generation in large language models (LLMs) using reinforcement learning techniques. arXiv preprint arXiv:2412.20367.