Kunfu Wang

English / 中文 / 日本語

Contact

Email: [email protected]
Website: https://okonfu.com
GitHub: https://github.com/wkf16

Education

Fuzhou University | B.Eng. in Electronic Information Engineering (Expected Jun 2026)

National Project 211 and Double First Class University

  • Academic Standing: GPA 3.18 / 4.0 (Top 22% in major cohort).
  • Selected Coursework: Discrete Signal Processing (99), C Programming (98), Computer Networks (97), Java Programming (95), Digital Image Processing (95), Microprocessor Principles (91), Electromagnetic Field and Wave (91), Object Oriented Programming (90).

Skills

  • Languages: TOEFL iBT 82 (MyBest 85), JLPT N1 (Perfect score in Reading).
  • Programming and Systems: C, C++, Python, MATLAB, Java, TypeScript, Node.js, CUDA, Linux, Flask, Socket.IO, WebSocket, JSON RPC.
  • Embedded and Hardware: STM32, ESP32, Nordic nRF52 series, Arduino, Verilog, FPGA, KiCad PCB layout, signal chain analysis, circuit debugging.

Research and Engineering Projects

ATOMIC Swarm Robot Platform | HKUST(GZ) ATOMIC Lab

Research Assistant (May 2026 to Present)

  • Full stack swarm robotics infrastructure. Owned the complete swarm platform across four distinct layers: custom KiCad printed circuits, 3D printed mechanical housings, bare metal C firmware on the nRF52833, and a Flask web host powered by Socket.IO. Implemented indoor localization pairing HTC Vive Lighthouse stations with TS4231 optical sensors, broadcasting 50 Hz pose streams via direct 2.4 GHz radio packets without SoftDevice overhead to compress total firmware within 128 KB.
  • Visual verification and spatial calibration. Built an overhead monocular camera rig to establish an absolute spatial reference and correct base station drift. Mapped metric coordinates through homography from multi sheet ArUco marker grids while deploying a custom YOLO model to track robot positions from a single frame, achieving repeatability around 3 mm on markers and 22 mm on robots. This system provides visual ground truth for Lighthouse tracking while streaming refreshed affine calibration parameters to the swarm over the air.
  • Rapid system integration and deployment. Accelerated platform maturity during initial onboarding by shipping three core infrastructure milestones: overhauled the operator interface with responsive viewport mapping and live target feedback, architected a Flask REST and SSE streaming pipeline verified by automated trajectory testing, and established over the air deployment tooling to distribute runtime calibration data across the active fleet.
  • Swarm Chess demo at EWSN 2026. Co developed Swarm Chess, a live demo in which every chess piece is an AtomBot robot and LLM agents play through the physical swarm. The agent runs a ReAct loop and sees the board only through four chess tools (get board, move, setup, resign), while a chessboard engine validates each move and a WHCA* multi robot path planner turns it into target points for the robots. Supports Agent vs. Agent and Agent vs. Human play, with synchronized control hub, wall and floor displays showing agent reasoning, tool calls and robot motion. Presented as an oral demo at EWSN 2026 in Dresden as co first author [1].

Open-Yachiyo: Desktop Agent Execution Framework

Core Architect (2026 to Present) | Open Source (300+ GitHub Stars) | GitHub

  • Runtime loop and execution architecture. Co architected a desktop runtime from scratch around the Reason, Act, and Observe paradigm. Designed an asynchronous Event Bus tool calling pipeline coupled with Server Sent Events and JSON RPC to provide granular, step by step execution traces and state observability for in flight agent decisions.
  • Modular system design. Engineered decoupled tool dispatch and structured context management layers to handle complex agentic workflows under uncertain environments, sustaining rapid feature scaling while maintaining clean system modularity.
  • Academic recognition. Funded as a National College Students Innovation and Entrepreneurship Project (SRTP) and selected as a Fuzhou University Outstanding Graduation Design.

Edge-AI Real-time Rice-Pest Monitoring and Robotic Trapping System

First Author | Published in Frontiers in Plant Science (IF 5.9, JCR Q1) (2024 to 2026)

  • Dataset and edge perception pipeline. Constructed a field dataset of 1,500 images across 12 distinct pest categories collected in natural agricultural environments. Designed an ultra lightweight detection network featuring a Mixed Aggregation backbone and Ghost Conv pre processing head, optimized via TensorRT for edge deployment on the NVIDIA Jetson Xavier NX alongside an ESP32 low power actuation subsystem and LoRa multi node mesh.
  • Empirical performance and validation. Achieved 96.1% mAP at 0.5 with 59.4 FPS directly on edge hardware, boosting inference speed by 19.6% over the standard YOLO baseline. Demonstrated superior noise robustness by securing the highest mAP in 6 out of 9 degraded field scenarios, with a closed loop latency of 14.8 ms and complete system latency of 36.8 ms.
  • Scholarly publication. Published as first author in Frontiers in Plant Science (Volume 2026, DOI: 10.3389/fpls.2026.1853368).

Industrial Robotic Arm for Insole Blanking with Adaptive Edge Detection

Robotics Researcher (2023 to 2024) | GitHub

  • Motion planning and simulation. Developed the FSPS-BIT* motion planning algorithm operating within a 3D workspace for factory insole blanking, benchmarked comprehensively against traditional planners including RRT*, FMT*, Informed-RRT*, and standard BIT*.
  • Implementation and intellectual property. Owned the core MATLAB algorithmic implementation and the UR5 robotic simulation environment. Research outcomes directly contributed to a co authored academic paper and a granted Chinese invention patent.

Publications

  1. K. Wang, P. Li, Y. Gao, J. Liu, J. Dai, T. Chang. “Demo: Swarm Chess — Grounding LLM Agents in a Physical Robot Swarm.” International Conference on Embedded Wireless Systems and Networks (EWSN 2026), Dresden, Germany, 2026. Oral demo presentation. DOI: 10.3217/fvzd-jx74
  2. K. Wang, S. Guo, X. Yang, B. Lin, W. Cai, R. Tang, J. Lin. “An integrated edge AI prototype for smart agriculture: real-time pest detection, physical trapping, and multi-node deployment analysis under field uncertainty.” Frontiers in Plant Science, vol. 17, art. 1853368, 2026. DOI: 10.3389/fpls.2026.1853368
  3. R. Tang, K. Wang, H. Lin, L. Huang, G. Mou, S. Guo. “A framework of insole blanking robot based on adaptive edge detection and FSPS-BIT* path planning.” Scientific Reports, 2024. DOI: 10.1038/s41598-024-71636-4
  4. J. Lin, Z. Wu, K. Wang, et al. “Cascade residual-optimized image super-resolution reconstruction in Transformer network.” Optics and Precision Engineering, 2024. DOI: 10.37188/OPE.20243212.1902

Patents

  1. [P1] G. Mou, K. Wang, R. Tang, W. Liu, L. Huang, S. Guo. Insole Blanking Robot System with Adaptive Edge Detection and Path Planning. China Invention Patent CN118181403B, granted 2025 (PI as first inventor).
  2. [P2] Y. Liao, K. Wang, X. Yan. Subpixel-Edge Method for Circular-Hole Localization and Deformation Detection in Stamped Parts. China Invention Patent CN116542910B, granted 2025 (PI as first inventor).

Honors and Awards

Texas Instruments Cup National Undergraduate Electronic Design Contest, National Second Prize (2023)

  • Completed an end-to-end embedded system for superposed-signal feature recognition and separation, including HW/SW co-design and an FFT-based signal-analysis workflow.
  • Role (Team Leader): designed overall system architecture, led cross-module integration and joint debugging, and took ownership of PCB design and STM32 main-program programming.

National Embedded Chip and System Design Competition, National Third Prize (2024)

  • Implemented an FPGA SoC based on VexRisc-V and validated core system functions.
  • Role: coding, resource integration, and hardware development.

APMCM Mathematical Contest in Modeling, Third Prize (2024)

  • Developed A*-based logistics path-planning models and comparative routing analysis.
  • Role: MATLAB programming.

COMAP Mathematical Contest in Modeling (MCM), Honorable Mention (2022)

  • Completed a cellular-automata-based grassland ecosystem simulation for 2022 MCM Problem C to predict plant-community evolution under irregular weather cycles and drought-period species interactions.
  • Role (Team Leader): led system modeling, contributed to paper writing, completed all data visualization deliverables, and implemented part of the MATLAB programming.