冯凯鑫Kaixin Feng

SLAM · 多传感器融合 · 机器人状态估计SLAM · Multi-Sensor Fusion · Robot State Estimation

Wuhan University · LIESMARS
kaixinfeng@whu.edu.cn
[ CV / 简历 ]

我是武汉大学测绘遥感信息工程全国重点实验室（LIESMARS）硕士研究生，导师李由教授。研究方向为 GNSS/IMU/LiDAR 多源融合定位与机器人状态估计，并探索用学习方法增强经典估计框架中的可靠性与噪声建模。

I am an M.S. student at LIESMARS, Wuhan University, advised by Prof. You Li. My research focuses on GNSS/IMU/LiDAR fusion positioning and robot state estimation, and on augmenting classical estimation frameworks with learning-based reliability and noise modeling.

教育经历Education

武汉大学 — 硕士Wuhan University — M.S.2025 — Present

LIESMARS · 计算机技术 · 导师：李由教授LIESMARS · Computer Technology · Advisor: Prof. You Li

武汉大学 — 本科Wuhan University — B.E.2021 — 2025

遥感科学与技术 · GPA 3.75/4.0 · 排名 10/48（保研本校）Remote Sensing Sci. & Tech. · GPA 3.75/4.0 · Rank 10/48

研究工作Publications

IROS 2026 · Oral · 一作First author

WinTA-GIL: Windowed Trajectory Alignment for GNSS/IMU/LiDAR Heading Refinement in Intermittent Signal Environments

K. Feng, Z. Wen, Z. Liao, X. Xia, Y. Li* · [Paper] · [Demo]

UPINLBS 2025 · Oral · 一作First author

A Modular and Scalable Multi-Sensor Fusion Framework Toward Ubiquitous High-Precision Positioning

K. Feng, Z. Wen, X. Huang, K. Lu, Y. Li* · [Paper]

项目经历Projects

学习型足式里程计 · Per-axis FK 可靠性权重Learning-Augmented Legged Odometry智元创新 · 2026.04–06AgiBot · 2026

Transformer 预测每足正运动学速度可靠性权重（2 维无量纲）。Transformer predicting per-foot FK velocity reliability (2-D dimensionless).
权重到 EKF 观测噪声的自适应映射，对打滑/腾空退化工况各向异性降权。Adaptive weight-to-noise mapping for anisotropic down-weighting under slip / flight.
自监督 FK 损失 + 域随机化缓解 sim-to-real；690K 网络、100Hz 实时推理。Self-supervised FK loss + domain randomization; 690K model at 100Hz.

WinTA-GIL · 航向优化Heading RefinementWHU · 2025.09–2026.03

滑动窗口航向精化，将单次初始对齐扩展为持续轨迹一致性优化。Window-based heading refinement extending one-shot alignment into continuous optimization.
SE(3) 刚体变换 + 杆臂补偿，Ceres 加权优化实现 LIO–GNSS 对齐。SE(3) modeling with lever-arm compensation; weighted Ceres optimization for LIO–GNSS alignment.
相比 KF-GINS 航向 RMSE 改善约 90%（1.04°）。~90% heading-RMSE improvement over KF-GINS (1.04°).

SPARK · 组合导航系统Integrated NavigationWHU · 2024.09–2025.09

GNSS/IMU/LiDAR 多源融合定位；设计 INS 初始对齐与非完整约束（NHC）建模。GNSS/IMU/LiDAR fusion; INS alignment and non-holonomic constraint (NHC) modeling.
基于 ROS 构建多传感器同步与通信框架，实现实时融合与模块化部署。ROS-based multi-sensor synchronization for real-time fusion and modular deployment.

竞赛 · 实习Awards · Internship

华为杯数学建模竞赛Math Modeling (Huawei Cup)

全国二等奖 · 队长 · 2025National 2nd Prize · Team Leader · 2025

SLAM 算法实习SLAM Intern

智元创新 · 机器人多源融合定位 · 2026AgiBot · robot fusion localization · 2026

技能Skills

C/C++ · Python · PyTorch · Matlab · ROS · Eigen · Ceres · OpenCV · PCL · CMake · Docker · Git · Linux