Yunxiao Shan, Ph.D. from Rutgers University-Wuhan University, associate researcher at Sun Yat-Sen University, review expert for the second China Unmanned Vessel Open, responsible for research on unmanned boat-related technologies of Sun Yat-Sen University, focusing on planning and control technology for unmanned systems And based on single/multi-surface robot-based surface target detection and tracking, water surface passable area extraction and other research areas.
单云霄 副研究员
shanyx@mail.sysu.edu.cn
Academic achievement: Doctoral research mainly focuses on the planning and control technology of unmanned driving, and the post-doctoral research extends to the perception, planning and control technology of platforms such as surface navigation robots and mobile robots. Conducted in-depth research on robot perception, planning, and control. Published 17 academic papers in ITS, TCVST, TVT and other internationally renowned journals, SCI indexed 6 papers (Chinese Academy of Sciences area 2 or above 4 papers), EI indexed 11 papers (4 by the first author), 10 patent applications (4 by the first inventor, 1 authorized).
Representative academic achievements:
Host and participate in scientific research projects:
1. Research on key technologies based on intelligent coastal defense multi-source sensing and interconnected early warning (Shenzhen Science and Technology Innovation Commission), 2 million, ranked second, 2019-2021;
2. Key technologies for autonomous tracking and coordination of multiple unmanned ships in high seas under high sea conditions Research and Industrialization Application Demonstration (Key R&D in Guangdong Province), 5 million, project number: 2018B010108004, sub-project leader of Sun Yat-sen University, 2019-2021;
3. Development of multi-scenario-oriented high-precision, full-steer, large-load mobile operation integrated robot And Application Demonstration (Key R&D in Guangdong Province), 10 million, project number: 2019B090919003, sub-project leader of Sun Yat-sen University, 2019-2021;
4.Vision-based robot environment modeling and positioning and navigation (key research and development by the Ministry of Science and Technology), 10 million , Subject number: 2018YFB1305005, backbone member, 2019-2021;
1. Water surface target detection and tracking based on deep learning
Current research results:
The SiamFPN RPN algorithm was developed to achieve a target tracking accuracy improvement of more than 10% compared with traditional methods under high sea conditions.
| 跟踪模型 | 准确率 | 召回率 | F值 |
| KCF（2014） | 64.88% | 64.88% | 64.88% |
| ECO（2016） | 65.09% | 65.09% | 65.09% |
| SiamRPN（2017） | 58.04% | 57.39% | 57.71% |
| SiamVGG（2018） | 56.27% | 55.88% | 56.07% |
| SiamFPN RPN | 73.7% | 73.67% | 73.69% |
Future research directions:
1. Water surface target detection and tracking method based on deep neural network multi-sensor fusion (including but not limited to: visible light, infrared, lidar, millimeter wave radar)
2. Multi-surface target detection and tracking method based on multi-robot system
2. Research on application algorithms of surface lasers
Current research results:
A lightweight neural network is used to classify riverside targets, and particle filtering and convex polygons are used to extract the drivable area of the target water surface.
Future research directions:
1. Use laser radar to realize 3Dslam positioning, solve the impact of high dynamic water surface SLAM technology, and realize autonomous positioning of unmanned ships under weak GPS conditions.
2. Use lidar to realize accurate perception of water surface objects. Aiming at the characteristics of large swings and large appearance differences of water surface objects, use lidar or fusion with other sensors to accurately realize 3D modeling, positioning and tracking of water surface objects.
3. Robot motion planning and control technology
Mainly focus on the planning and control technology of unmanned vehicles. The three main directions currently under research are motion planning technology considering uncertainty, data-driven planning methods, and end-to-end steering control methods. Among them, the research direction of uncertain motion planning technology welcomes those who have a mathematical background or are interested in optimization control to contact me.
The direction of control technology mainly focuses on path tracking control based on reinforcement learning and adaptive tracking control methods.
