Master student in robotics, computer science or computer graphics (f/m/x): Global Indoor Localization & Mapping with LiDAR-Inertial systems
What to expect:
Global localization in indoor environments is a skill of critical importance for robots that operate in confined spaces. Robotic systems that are tasked with interacting with a man-made environment, for instance fetching bottles from the fridge or loading a dishwasher, need accurate knowledge of where the target objects are placed and simultaneously of their position in the same environment. Traditional vision-based approaches to mapping, and localization on prior maps, suffer from lack of unique textures and heavy reflections, which are typical of human-made environments.
The objective of this thesis work is to design and implement a localization and mapping approach based on LiDAR sensing, using one or multiple sensors, to provide accurate positioning, unaffected by challenging visual appearance. Starting from unknown, or partially known indoor scenes, the task is to quickly create an accurate map representation, and consequently provide continuous localization within it. The developed approach should build upon recent developments on LiDAR-Inertial SLAM, e.g., FAST-LIO2, potentially including multiple sensors, and agnostic to the specific LiDAR sensor employed.
The developed approach will be deployed and tested on robotic platforms, e.g., Rollin’ Justin, for terrestrial assistance applications developed and in use at DLR-RM.
Tasks:
- In-depth evaluation and testing of alternative LiDAR-Inertial SLAM approaches, suitable for robust indoor localization, considering one or multiple sensor solutions
- Development of an extrinsic calibration pipeline, to reference the LiDAR sensor against existing visual sensors, mounted on the robot, and a central robot frame
- Implementation of a global localization approach, using completed LiDAR maps to provide drift-free localization to the robotic system
- Interface to occupancy or surface mapping, e.g. Octomaps, TSDF, to interface with motion planning architectures
What we expect from you:
- Bachelor’s degree in Computer Science, Electrical Engineering, Robotics Engineering or a similar field
- Academic knowledge and practical experience on Simultaneous Localization and Mapping (SLAM) approaches
- Solid knowledge on point-cloud processing techniques and related data structures, e.g. ICP, kd-trees
- Advanced C++ and Python programming skills
- Fluency with the English language
- Intent to disseminate the results of this work through publications in relevant computer vision or robotics conferences
To apply for this position, please include a resume and transcript of grades.
For application, first make contact (see below) to indicate your interest. We will then get back to you.
Further information:
Starting date: Immediate
Duration of contract: 6 months