Feifan Liu

ZJU-UIUC Institute, Zhejiang University

Electronic Information

Henan University · Bachelor's degree in Automation

liu.feifan.ai@outlook.com

Hi, I’m Feifan Liu. I am currently a Master’s student in Electronic Information at the ZJU-UIUC Institute (ZJUI), Zhejiang University.

I approach problems with the dual mindset of a researcher and an engineer. Beyond theoretical soundness, I care about what happens when an algorithm is flashed onto a chip or integrated into a real-world system—whether it can withstand physical constraints and actual operating conditions.

From Edge Hardware to Complex Systems My early technical background is rooted in low-level hardware. I have designed PCBs, written control algorithms, and deployed lightweight Transformer models on edge devices. These experiences gave me an intuitive grasp of compute limits, hardware-software co-design, and overall system architecture.

Present and Future: Grid + LLM & Intelligent Decision-Making Currently, my research focuses on the intersection of power systems and artificial intelligence, with a long-term commitment to Grid + LLM and AI Agents. My recent work revolves around GNN-based grid topology control (e.g., busbar splitting and merging). Previously, I built an autonomous AI agent workflow for compiler optimization and developed fine-tuning and RAG architectures for domain-specific LLMs. I am now working on bringing the complex reasoning and scheduling capabilities of large language models into the power grid—a sector with high technical barriers and immense real-world value.

About This Space This is my personal blog, where I document my learning and thoughts. If you have specific engineering problems or research ideas, feel free to reach out and connect.

news

Jan 15, 2016	A simple inline announcement with Markdown emoji!
Nov 07, 2015	A long announcement with details
Oct 22, 2015	A simple inline announcement.

latest posts

Sep 27, 2026	Agent Skills
Sep 25, 2026	【译文】图神经网络入门指南
Apr 28, 2025	a post with marimo snippets

selected publications

Drones
Edge-Deployed Band-Split Rotary Position Encoding Transformer for Ultra-Low-Signal-to-Noise-Ratio Unmanned Aerial Vehicle Speech Enhancement

Feifan Liu, Muying Li, Luming Guo, and 4 more authors

Drones, May 2025

Abs DOI Bib HTML PDF

Addressing the significant challenge of speech enhancement in ultra-low-Signal-to-Noise-Ratio (SNR) scenarios for Unmanned Aerial Vehicle (UAV) voice communication, particularly under edge deployment constraints, this study proposes the Edge-Deployed Band-Split Rotary Position Encoding Transformer (Edge-BS-RoFormer), a novel, lightweight band-split rotary position encoding transformer. While existing deep learning methods face limitations in dynamic UAV noise suppression under such constraints, including insufficient harmonic modeling and high computational complexity, the proposed Edge-BS-RoFormer distinctively synergizes a band-split strategy for fine-grained spectral processing, a dual-dimension Rotary Position Encoding (RoPE) mechanism for superior joint time–frequency modeling, and FlashAttention to optimize computational efficiency, pivotal for its lightweight nature and robust ultra-low-SNR performance. Experiments on our self-constructed DroneNoise-LibriMix (DN-LM) dataset demonstrate Edge-BS-RoFormer’s superiority. Under a −15 dB SNR, it achieves Scale-Invariant Signal-to-Distortion Ratio (SI-SDR) improvements of 2.2 dB over Deep Complex U-Net (DCUNet), 25.0 dB over the Dual-Path Transformer Network (DPTNet), and 2.3 dB over HTDemucs. Correspondingly, the Perceptual Evaluation of Speech Quality (PESQ) is enhanced by 0.11, 0.18, and 0.15, respectively. Crucially, its efficacy for edge deployment is substantiated by a minimal model storage of 8.534 MB, 11.617 GFLOPs (an 89.6% reduction vs. DCUNet), a runtime memory footprint of under 500MB, a Real-Time Factor (RTF) of 0.325 (latency: 330.830 ms), and a power consumption of 6.536 W on an NVIDIA Jetson AGX Xavier, fulfilling real-time processing demands. This study delivers a validated lightweight solution, exemplified by its minimal computational overhead and real-time edge inference capability, for effective speech enhancement in complex UAV acoustic scenarios, including dynamic noise conditions. Furthermore, the open-sourced dataset and model contribute to advancing research and establishing standardized evaluation frameworks in this domain.
@article{liu2025edge, title = {Edge-Deployed Band-Split Rotary Position Encoding Transformer for Ultra-Low-Signal-to-Noise-Ratio Unmanned Aerial Vehicle Speech Enhancement}, author = {Liu, Feifan and Li, Muying and Guo, Luming and Guo, Hao and Cao, Jie and Zhao, Wei and Wang, Jun}, journal = {Drones}, volume = {9}, number = {6}, pages = {386}, year = {2025}, month = may, publisher = {MDPI AG}, doi = {10.3390/drones9060386}, url = {https://doi.org/10.3390/drones9060386}, }