Xin-Liang Zhong （钟心亮）

Master candidate
School of Mechatronic Engineering
Beijing Institute of Technology

Postgraduates' Apartment 2, No. 5 Zhongguancun South Street
Haidian District, Beijing, P.R. China, 100081

Email: xinliangzhong@foxmail.com
Github: https://github.com/TurtleZhong

Biography

I am currently a third-year Master candidate of School of Mechatronic Engineering, Beijing Institute of Technology. My supervisor is Dr. Qingsheng Luo. Prior to that, I received my Bachelor's degree from Beijing Institute of Technology in 2016.

My research interest includes Computer Vision and SLAM.

About me: [Github] [Youtube] [知乎] [Bilibili]

工作经验

2023-至今

硬件在环（HIL）的软硬一体链路实现以及基于HIL的智驾参考实现

1）围绕CARLA仿真器，自研数据注入设备以及域控，实现了仿真器-数据注入设备-域控-底盘-仿真器闭环，搭建了相应的HIL台架，将真实硬件设备接入仿真数据进行上车部署前的测试，加速各个链路验证以及算法的迭代落地
2）编写UE4的着色器使CARLA支持鱼眼相机，整套系统支持1V1R的入门级LKA仿真以及4V的APA泊车仿真
3）CARLA仿真器支持了不同实车底盘can协议
4）基于整套系统开发了LKA/APA参考实现，并在实车上进行了验证
5）...
鱼眼镜头图像

逼真的仿真数据生成用于训练数据的生成以及智驾corner case生成

1）可以添加如下车辆、定义车辆的基本运动方式，也可以让自车运动起来，当然也可以生成一些corner case，例如圆锥桶在地面
2）注意光影细节、同时可以生成天气变化、白天黑夜的图像
2） 更多细节请电话/邮件联系，个人网页只放最基础的demo

2020-2023

单目视觉在基于已知地图的视觉定位工作

一个通用的 基于视觉地图的定位系统. 它包括:

1) 支持传统与深度学习特征的地图创建.
2) 分层定位模块.
3) 支持融合轮速、IMU、GNSS平滑.

[Project page] [ZhiHu] [Code Sample_0] [Code Sample_1] [Related Video_0] [Related Video_1]

多机器人纯自主探索主动式协作快速建图

1）主要是多个机器人基于frontier探索点的方式进行快速地图构建与合并

地外采样返回小车

AVP-SLAM 自主代客泊车视觉定位方案

对论文[AVP-SLAM: Semantic Visual Mapping and Localization for Autonomous Vehicles in the Parking Lot(IROS 2020)]的复现，并将整个Gazebo仿真器开源。
[知乎介绍] [Code]

LVIO-SAM 基于因子图的雷达视觉惯性紧耦合里程计

[Code]
实际效果图

RF-LIO 基于视点可见性的动态物体在线移除LIO方法

基本原理：基于range image的方法将submap和当前激光投影到image，根据深度差判断该点是否为动态点。
1.基本框架与定位结果量化对比
2.建图效果与lio-sam对比，注意地图中的动态车辆拖影
详细方法参考原始论文RF-LIO: Removal-First Tightly-coupled Lidar Inertial Odometry in High Dynamic Environments

基于打滑估计方法的纹理稀疏场景定位

洞穴自主探索与返回机器人

1.设计了基于图结构的分层自主探索与返回方案，并结合高程地图计算机器人的局部可通行性地图，使机器人能够在崎岖环境下安全的纯自主探索与采样返回，方案最终在莫干山洞穴进行了大量测试并得以验证
2.机器人青科论坛之导航专场分享《面向地外探测的机器人定位建图与自主探索技术研究》，直播在线观看人数5k+
图像布局

2019-2020

基于纯视觉的低成本XX无人物流车方案

1) 纯视觉定位方案，包含离线建图以及在线定位两大部分.
2) 基于3D点地图的兼容传统与深度特征点描述子的后端重定位框架，平均定位误差95%以上处于30cm以内.
3) 融合单目鱼眼摄像头的语义分割，检测，深度摄像头的深度信息，超声波信息和IMU信息的感知bev的costmap用于规控.
图像布局

Projects before 2019

 A tool used for calibrate 2D laser range finder (LRF) and camera.

The package is used to calibrate a 2D LiDAR or laser range finder(LRF) with a monocular camera. Specficially, Hokuyo UTM-30LX have been suscessfully calibrated against a mono camera.

[Project page] [Code]
A Multi-State Constraint Kalman Filter for Vision-aided Inertial Navigation

The MSCKF_mono package is a mono version of MSCKF. The software takes in synchronized mono images and IMU messages and generates real-time 6DOF pose estimation of the IMU frame.

[Project page] [Code]
Auto Label Tool for Autonomous Car.

A tool used for automatic label the road sign.

[Project page] [Code]
An offline tool for pose-graph-optimization.

This tool can optimize the pose and eliminate the cumulative error.

[Project page] [Code]
Reconstruction of the scene based on RGBD-Camera.

It is a simple SLAM system based on RGBD-cameras.

[Video]
DJI Robomasters Summer Camp.

This projects aims to design an autonomous MAV and a mobile robot that can grab the doll and place it in the bucket corresponding to the doll pattern.

[News] [Video]
Research on Real-time Location and Construction of Indoor Mobile Robots.

This robot is equipped by the relevant sensor itself can be in a completely unknown environment situation accurately draw the map of the surrounding environment.

[Docs] [Video]

Experiences

Awards and Honors