| Inhalt | Competence Goals
After successfully completing the practical course, students are able to:
- Literature Study & System Design: Independently study a small number of assigned scientific literature to thoroughly understand the theoretical foundations, algorithmic approach, and software architecture requirements.
- Interface and Integrate Hardware/Sensors: Connect, configure, and process raw sensor data (e.g., IMUs, cameras, joint encoders, or distributed/neuromorphic sensors) into functional data pipelines.
- Reimplement Bionic Algorithms: Independently translate state-of-the-art machine learning approaches, perception algorithms, neural network architectures, or control approaches based on scientific literature into efficient, working software implementations.
- Evaluate and Benchmark System Components: Rigorously test and evaluate individual algorithmic components and architectures regarding accuracy and other relevant metrics.
- Integrate and Deploy on Robotic Systems: Combine perception, processing, and/or control components into a functional robotic system and evaluate it in a simulated environment or with a physical robotic platform.
- Document and Present Results: Document software developments and present technical outcomes through live demonstration and/or presentation.
Prerequisites
- Attending the lecture " Biologically Inspired Robots or the theoretical Seminar: Bionic Algorithms and Robot Technologies" is strongly recommended.
- Solid programming skills in Python or C++ and familiarity with Linux, Docker, machine learning frameworks (e.g., PyTorch), or robotics software frameworks (e.g., ROS2) are advantageous.
Content
The focus of this course is software development and system integration, and therefore, the practical application of current State-of-the-Art research.
The primary objective is to bridge the gap between scientific theory and practical implementation. Students build, test, and benchmark modular components, ultimately integrating them into functional systems on physical robotic platforms or in high-fidelity simulation environments.
Throughout the term, students gain hands-on experience with the complete development process, from analyzing a scientific publication to implementing, integrating and evaluating on a robotic system
- Analyse Scientific Literature: Independently review and analyse selected scientific publications in robotics and AI to extract algorithmic concepts, implementation details, and experimental methodologies.
- Algorithmic Understanding and System Design: Derive the required software architecture, data flow, state representations, action spaces, and interfaces needed to reproduce the proposed approach.
- Algorithmic Implementation: Reimplement state-of-the-art robotics algorithms, machine learning policies, perception pipelines, or bionic control approaches using modern software frameworks and robotics libraries.
- Simulation, Hardware and System Integration: Integrate algorithms into simulation environments (e.g., Isaac Sim, Isaac Lab, Gazebo) or physical robotic platforms, including sensor interfaces, robot models, control loops, and software components.
- Evaluation and Benchmarking: Systematically test, debug, and quantitatively evaluate algorithms and robotic systems with appropriate performance metrics, compare results against the original publication, and analyse deviations and limitations.
|
