Bowen Li

Experience

• 

  OSRAM

  May 2018 - Feb 2019

  Working Student

  I participated in a project named LiONS (Lichtsensor-basierte Ortungs- und Navigationsdienste für autonome Systeme) which is about indoor AGV (Automated Guided Vehicle) localization and navigation for industrial application based on intelligent light system. It is a joint research project of OSRAM, Fraunhofer IEM from Paderborn, Götting KG and KEB Automation.OSRAM takes responsibility of the software for localization and navigation as well as the manufacturing of luminaries mounted on the ceiling. The software combines vSLAM (visual Simultaneous Localization and Mapping) and VLC (Visual Light Communication) to precisely calculate the pose of AGV with the accuracy up to centimeter level and then navigates the AGV accordingly.I worked in a team of three people to develop the localization algorithm and finally made a pilot run in a factory of KEB Automation. We took advantage of cutting-edge vSLAM algorithm and combined it with our innovative intelligent light system. Eventually the desired accuracy was achieved and the final demo was a success. 收起

• 

  ZF Group

  Mar 2019 - Aug 2019

  Master Thesis Student

  I wrote my master thesis at the department of autonomous driving. The title is "Multi-Feature Based Localization System Using Graph Optimization For Autonomous Driving".The thesis aims to solve the localization problem for autonomous driving fusing different sensors. GNSS and IMU are used to provide prior pose information and kinematic information of ego vehicle. Radar and Lidar are used to extract multiple features from environment.Different motion models are compared for pose prediction. Various feature extraction and association algorithms are researched as front-end. Graph optimization is investigated as back-end. Combining all mentioned components, a multi-feature based localization system is implemented.The software is capable of doing mapping, map-based localization and SLAM (Simultaneous Localization And Mapping) all based on multiple features and using graph optimization. It is also designed in a multi-thread way in order to achieve real-time processing.The conclusion is that Radar-only and pole feature-only 2D map-based localization can reach around 50 cm accuracy. The accuracy of 3D map-based localization using pole, line and plane features detected by Radar and Lidar is not satisfying. No matter what features are used, SLAM in general has large drift error. 收起

• 

  ZF Group

  Oct 2019 - now

  Software Engineer

  1. Algorithm development for different perception components.2. Software detailed design for different perception components.3. KPI calculation for evaluating algorithm performance.4. Integration of multiple software components.