写写代码，搞搞机器人，
想要改造世界，更想认识自己。
Hey，我叫自信，想法很多，有点技术，喜欢钻研各类技术向的东西。然后这里呢，是我的技术和学习博客，欢迎一起交流讨论~
我呢，博士毕业于浙江大学控制学院，目前在浙江人形机器人创新中心做人形机器人相关工作。负责机器人系统架构设计，以及技术和项目团队管理，自己也负责一些核心算法的研究和框架的开发。 博士期间在浙江大学机器人实验室研究多机器人系统以及全局定位中的各类问题。研一的时候和师兄一起做了一套mapping，发了两篇文章，开源了代码，也将这套mapping应用到了几个项目中。在与师兄合作完后，对多机定位与建图进行了一番研究，首先考虑多机系统中传感器异构的前提，利用之前的mapping系统，进一步实现了可微分相位相关算法 DPCN 用于异构定位。其中的一个想法进一步产生了一篇位置重识别的期刊：DiSCO。研三主要在DiSCO的基础上研发可估计位姿的地点重识别方法，和师妹合作发表了RING系列文章：RING/RING++/RING#。同时也开发了一套集中式的实时多机器人SLAM系统MR_SLAM，目前已应用到一个项目中并成功结题。博士后期在阿里达摩院开发了一套环视视觉的全局定位vDiSCO。最后写大论文阶段，就在一起筹备浙江人形机器人创新中心的各种事务。
一些项目与成果
- 人形机器人项目 - 多模态交互、系统架构、项目管理
- vDiSCO
- RING#
- RING
- 实时集中式多机器人SLAM系统-MR_SLAM
- 多模态传感器定位
- 实时2.5D地图构建 / 实时的全局一致性2.5D地图构建
- 可估旋转的地点描述子-DiSCO
- 基于 Quanergy M8 的 3D 地图构建
- 四旋翼多机通讯及控制
- 基于红外信标的四旋翼精准降落
一些能力
- 计算机视觉
- 深度学习 (Pytorch)
- ROS
- C++
- CUDA
- Python
To be a better man.
Hi, I am @maverickp (Xuecheng Xu), a PhD major in Robotics. I am currently working on embodied intelligence. My PhD thesis is about multi-agent SLAM systems.
I hold a Ph.D. from the Control Science and Engineering College at Zhejiang University and currently serve as the Chief Technical Assistant at the Zhejiang Humanoid Robotics Innovation Center. My responsibilities include planning the company’s technical roadmap, and managing the technical team and projects, as well as conducting research on core algorithms and developing frameworks.
During my doctoral studies at Zhejiang University’s Robotics Lab, I focused on multi-robot systems and various issues in global positioning. In my first year of graduate school, I worked with my senior colleague to develop a mapping system, which we published in two papers, open-sourced the code, and applied it in several projects. Following this collaboration, I delved into multi-agent localization and mapping, considering the premise of sensor heterogeneity in multi-agent systems. Building upon our previous mapping work, I further developed the Differentiable Phase Correlation Network (DPCN) algorithm for heterogeneous localization. One of the ideas from this work led to a journal publication on location reidentification: DiSCO. In my third year, I continued working on DiSCO, developing a pose-estimating place recognition method in collaboration with my junior colleague, which resulted in publications of RING and RING++. Due to the need for physical experiments and project requirements, I also developed a centralized real-time multi-robot SLAM system called MR_SLAM, which has been successfully implemented in a project and concluded. In the later stages of my Ph.D., I worked at Alibaba DAMO Academy to develop a panoramic visual global localization system named vDiSCO. During the final stages of writing my dissertation, I was involved in preparing various affairs for the establishment of the Zhejiang Humanoid Robotics Innovation Center.
Projects
- Humanoid Robot Project - HRI, technical architecture and project managing.
- vDiSCO
- RING#
- RING
- Real-time centralized multi-robot system - MR_SLAM
- Heterogeneous sensor modality matching
- Real-time 2.5D mapping system / GEM: Gobally consistent elevation mapping
- Place recognition with Orientation
- 3D Mapping using Quanergy M8
- UAV Cooperation
- UAV Precise Landing
Skills
- Computer Vision
- Deep Learning (Pytorch)
- ROS
- C++
- CUDA
- Python
Specific information about works I’ve done can be found here -> Portfolio, Github
