Yulin Yang

I am currently a Machine Learing Engineer at Apple. My research focuses on robot navigation and state estimation.

I defended my Ph.D. in Robot Perception and Navigation Group at University of Delaware in Oct. 2021, advised by Prof. Guoquan (Paul) Huang. I received Master of Science degree in Mathematics from University of Delaware in 2020, Master of Engineering degree in Mechanical Engineering from Xi'an Jiaotong University in 2012, and Bachelor of Engineering degree in Mechanical Engineering from Shandong University in 2009. I was a recipient of the University Doctoral Fellowship Award (2019-2020) at University of Delaware. I have also completed internship programs at Facebook Reality Lab and Bosch Research Institute, respectively.

Email  /  CV  /  Google Scholar  /  Research Gate  /  GitHub  /  Twitter  /  Homepage@Udel

My research interests are within the areas of robot navigation, simultaneous localization and mapping (SLAM), computer vision, optimization, and multi-sensor fusion, with a concentration on enabling robots to perceive the world through fusing multi-modality information from different sensors (visual cameras, IMUs, sonar, wheel encoder, 2D/3D LiDAR, GPS and so forth).

Please find below a complete list of my publications with representative papers highlighted.

Multi-visual-inertial system with full-parameter calibration
IMU/camera intrinsics, IMU-IMU/camera spatial-temporal and RS calibration
Observability analysis and degenerate motion identification for MVIS with full calibration
Extensive simulations on 3 typical trajectories
Extensive evaluation with a self-made device regarding to Kalibr
Datasets are open-sourced in openmvis.com

Monocular VINS with full-parameter online self-calibration
IMU/camera intrinsics, IMU-camera spatial-temporal and RS calibration
Observability analysis for monocular VINS with full calibration
Comprehensive degenerate motion identification for IMU/camera intrinsics
Extensive simulations to verify different IMU intrinsic models
Extensive real-world experiments on TUM visual-inertial rolling-shutter datasets
Extensive evaluation with regarding to Kalibr
IMU intrinsic calibration code is released in OpenVINS

Autonomous path planning and sensor calibration for unmanned ground vehicle.

Tightly coupled MAV dynamic system with VINS.

VINS tightly coupled with GNSS measurements.

Improve FEJ to compensate the lieanrization errors.

Observability analysis and computation analysis for RI-VINS
Decoupled Right Invariant (DRI) based consistent algorithms (DRI-FEJ and DRI-SW)
Comprehensive evaluations for DRI-FEJ, DRI-SW, DLI-FEJ, FEJ, and standard EKF based VINS

Inertial aided multi-sensor calibration with spatial-temporal and intrinsic calibration.
Observability analysis and degenerate motion identification.
IMU, multi-cameras, LiDAR and wheel encoder.

MINS: a tightly-coupled IMU/CAM/LiDAR/Wheel/GPS navigation system
Plane patch-based efficient feature tracking with 10hz 64 channel LiDAR.
Intensive evaluations with realistic simulations and urban driving real world datasets.

Fully distributed cooperative (DISC)-VIO based on covariance intersection.
Centralized-equivalent cooperative (CEC)-VIO.

Sliding window based Lidar plane tracking.
LiDAR-Inertial-Camera fusion.
Intensive evaluations with Monte-Carlo simulations and real world datasets.

Tightly-coupled Visual-inertial-wheel odometry.
Observability analysis and degenerate motion identification.
Intensive evaluations with Monte-Carlo simulations and real world datasets.

Exploit 2D line for improved loop detection for occupancy grid mapping.
Design EKF framework to fuse inertial, odometry and 2D LiDAR.
Intensive evaluations with Monte-Carlo simulations and real world datasets.

Mono-VINS with online IMU intrinsic calibration.
Observability analysis and degenerate motion identification for IMU intrinsics.
Intensive evaluations with Monte-Carlo simulations and real world datasets.

Efficient visual odometry by leveraging prior LiDRA map.
Leverage ProW-NDT for semidense visual map and prior LiDAR map registration.
Intensive evaluations with simulations and real world datasets.

Modularized and analytic derived IMU integration algorithm.
Graph-based VINS with IMU-CAM time offset calibration.
Evaluations with Monte-Carlo simulations and real world experiments.

A research platform for visual-inertial estimation.
IMU-CAM spatial-temporal calibration and cam intrinsic calibration.
Intensive evaluations with Monte-Carlo simulations and real world datasets.

Tightly-coupled VINS with point and line features.
Two line triangulation algorithms.
Degenerate motion analysis for line triangulation.

A tightly-coupled VI system using prior LiDAR map constraints through MSCKF update.
Intensive simulations and real world experiments.

Unified feature representation (CP/Quaternion) for point/line/plane.
Complete observability analysis for VINS with point/line/plane
Intensive simulations for VINS with point/line/plane

Tightly-coupled visual-inertial system with point and plane features.
Tested point-on-plane constraints.
Both simulations and real world experiments.

Unified feature representation (CP/Quaternion) for point/line/plane.
Observability analysis for VINS with point/line/plane

Observability analysis for VINS with spatial-temporal calibration.
4 degenerate motion profiles for IMU-CAM calibration.
Intensive simulations and real world experiments.

Tightly-coupled visual-inertial target tracking.
Target represented as 3D rigid body.
Intensive simulations and real world experiments.

Closest Point (CP) representation for 3D plane.
Open source code for 3D LiDAR simulator.
LiDAR based 3D plane SLAM (LIPS).

Observability analysis for aided INS with point/line/plane.
5 dof nullspace for VINS with one line and 7 dof nullspace with one plane.
VINS simulation with point/line/plane.

Weighted maximum correntropy criterion (MCC)-based EKF.
Secure Estimation (SE)-EKF with attack detection.
Intensive Monte-Carlo simulations and experiments on real world datasets.

Null-space is equivalent to Schur complement with i.i.d. noise.
Time complexity analysis for Null-space and Schur complement.
Analytic null-space for MSCKF.

2D imaging sonar + IMU for underwater navigation.
Degenerate motion analysis for feature triangulation with sonar.
Online calibration between sonar and IMU.

Special thanks to Jon Barron for the website's source code. Feel free to use as well as contribute to it.