Master’s Thesis: Visual Pose Estimation for Elastic and Lightweight Robots
At the Institute of Robotics and Mechatronics at the DLR we are developing future robotic systems in a variety of environments: Our robots assemble products in the industries of the future, maintain satellites in space, explore foreign planets, or extend peoples’ capabilities in household environments. To enable the common task of manipulating objects, for instance, emptying a dishwasher in a household setting, one requires to know the location of a robot arm’s end-effector (manipulation tool). While this information can usually be precisely derived from forward kinematics, for some robots, such as neoDavid or LRU, this becomes inaccurate due to design constraints. Thus, we circumvent this shortcoming by fusing visually estimated features with the forward kinematics.
Your Contribution:
We are seeking a highly motivated, hands-on student to conduct research and further develop our PK ROKED approach (Meyer, Klüpfel, et al., “Robust Probabilistic Robot Arm Keypoint Detection Exploiting Kinematic Knowledge”, IROS WS 2022). Therefore, you will extend our/ develop a deep learning algorithm to detect distinctive features on a robot arm based on RGB images and prior kinematic knowledge derived from the system’s forward kinematics.
Project Scope:
- Review literature on existing techniques.
- Familiarize yourself with DLR perception and robotic platforms.
- Extend the existing PK ROKED approach (e.g., but not limited to architectural/ training data improvements, modality (e.g., semantic, depth, temporal) incorporation, geometric-aware loss functions, and uncertainty computations for predictions).
- Evaluate your developed algorithm on simulated and real-world robotic systems, e.g., neoDavid and LRU.
- Analyze and interpret your results to assess performance and limitations.
Your Qualifications:
- Enrolled in a Master's program in Computer Science, Electrical Engineering, Robotics, Aerospace Engineering, or a related field.
- Strong background in computer vision, statistics, and machine learning.
- Programming experience in C++, Python with libraries like OpenCV and PyTorch.
- Familiarity with deep learning concepts is a plus.
- Excellent analytical and problem-solving skills.
- Ability to work independently and as part of a team.
We offer:
- Gain valuable research experience in a cutting-edge field with high demand.
- Make significant contributions to the advancement of pose correction.
- Get hands-on experience on real robotic systems.
- Publish your research findings in a scientific journal or conference (subject to approval