Advanced Robot Control and Learning
|Language of instruction||English|
|Position within curricula||See TUMonline|
- 18.10.2022 11:30-13:00 Online: Videokonferenz / Zoom etc.
- 25.10.2022 11:30-13:00 Online: Videokonferenz / Zoom etc.
- 08.11.2022 11:30-13:00 Online: Videokonferenz / Zoom etc.
- 15.11.2022 11:30-13:00 Online: Videokonferenz / Zoom etc.
- 22.11.2022 11:30-13:00 Online: Videokonferenz / Zoom etc.
- 29.11.2022 11:30-13:00 Online: Videokonferenz / Zoom etc.
- 06.12.2022 11:30-13:00 Online: Videokonferenz / Zoom etc.
- 13.12.2022 11:30-13:00 Online: Videokonferenz / Zoom etc.
- 20.12.2022 11:30-13:00 Online: Videokonferenz / Zoom etc.
- 10.01.2023 11:30-13:00 Online: Videokonferenz / Zoom etc.
- 17.01.2023 11:30-13:00 Online: Videokonferenz / Zoom etc.
- 24.01.2023 11:30-13:00 Online: Videokonferenz / Zoom etc.
- 31.01.2023 11:30-13:00 Online: Videokonferenz / Zoom etc.
- 07.02.2023 11:30-13:00 Online: Videokonferenz / Zoom etc.
After successful participation in the module, the students are able to: - Understand robotics problems from differential geometry point of view and consequently interpret control objectives - Identify and determine the Robot Dynamics - Solve modern control problems with advanced methods - Understand and exploit the concept of redundancy in robotics - Design robust controllers - Design bio-inspired controllers - Understand and use learning methods in robotics
The Control section covers advanced topics in robot control, modeling and identification. Starting with the fundamental concepts such as differential geometry, it then covers the essential ideas behind the state-of-the-art of control methods in real robotic applications. Moreover, advanced methods are presented for the modeling and identification of physical systems and in particular robots. Finally, in the Adaptive Impedance Control section, the theories behind the human motor control are concisely explained and the associated bio-inspired control algorithms are described. The second part of the course covers typical problems in robot learning and approaches to solve them. Beyond that, many practical implications and problems regarding real-world applications are covered. The outline for this course is as follows. I. Differential Geometry in Robotics - Manifolds - Rigid Body Configuration - Tangent and Cotangent Space - Tensors II. Task Space Control - Robot Equations in Task Coordinates - SE3 Coordinates and Jacobian Matrix - Decoupled Position Control in Task Space III. Redundant Robots - Dynamics of Redundant Robots - Motion/Torque Null space - Null space and Stability IV. Passivity-based Robot Control - Passivity - 1-DoF Robot - Passive Representation of a Robot - PD-g(q) Control - Passivity & Robustness - PD-g(q)-Feedforward & PD+ Control - Damping Design - Slotine & Li Control - Joint Control Summary - Cartesian Impedance Control - Impedance vs. Admittance Control - Local Nature of Cartesian Control V. Port-based Modeling in Robotics - Dirac Structure & Power Ports - Port-based Modeling of a Manipulator - Passivity Analysis VI. Linear Parametric Modeling and Identification of Robot Dynamics - Linear Parameterization of the Manipulator’s Dynamic Model - The Minimum Parameter Set - Parameter Identification - Trajectory Optimization - Modeling of Friction - Identification Procedure - Adaptive Control VII. Adaptive Impedance Control - Bio-inspired formulation - Adaptive impedance control for a manipulator VIII. Learning for Physical Systems – Overview - Latest advances in machine learning for physical systems - Recent applications for autonomous systems - Literature overview IX. Real-world Problem Classes - Motion and manipulation skill learning - Blind manipulation - Vision-based manipulation - Sensitive grasping - Compound manipulation tasks X. Complexity reduction - Expert knowledge vs. data-driven approaches - Smart data vs. big data - Interaction control architectures - Interaction skill formalisms - System limits and task context - Latent spaces and dimensionality reduction XI. Cost function design - Feasibility and optimality - Unknown constraints - Confidence in real-world problems XII. Practical implications - Simulation vs. real world - Experiment design - Limitations of physical systems
- Fundamentals of control engineering - Fundamentals of robotics - Fundamentals machine learning - Fundamentals of statistics
Teaching and learning methods
Exercises are made available, presented and discussed during the lessons. Sample solutions are provided. The Robothon introduces the students to real-world problems that ought to be solved in the form of a project. Have a look at previous Robothons when Prof. Haddadin was still in Hannover: https://www.roboterfabrik.uni-hannover.de During the Robothon, students will gain hands-on experience in solving real-world problems using robotic systems. Students will work together in teams, where each team comes up with their own suggestion of a relevant problem which they may solve within one week. The students are responsible for planning out their solution to the respective problem and implementing it. They will be supported throughout the one-week course but are self-sufficient. At the end they will demonstrate their solution live and give a presentation about the methods and approaches they have used. Learning and control are the main theme of the student projects. Example problems are related to the future of robotics in human-robot shared working areas and industries (e.g., performing the peg-in-hole procedure, packing, etc. in the real world), future of mobility and health systems. - Presentations - Exercises with solutions - Robot hands-on experience - Tutorials
The students will learn the advanced topics in control and learning methods, and will be able to apply these methods to real robotic systems. The module grade is based on the student’s performance in a one-week hands-on course (practical demonstration and oral presentation), as well as a final oral exam (20 min). This evaluates the students’ knowledge in theoretical aspects of learning and control, and examines whether they are able to apply this to real-world robotic problems, and present results.
- Ploen, Scott Robert. Geometric algorithms for the dynamics and control of multi-body systems. Diss. UNIVERSITY OF CALIFORNIA IRVINE, 1997. - Dullemond, Kees, and Kasper Peeters. "Introduction to Tensor calculus." Kees Dullemond and Kasper Peeters - (1991). - Khatib, Oussama. "A unified approach for motion and force control of robot manipulators: The operational space formulation." IEEE Journal on Robotics and Automation 3.1 (1987) - Ott, Christian. Cartesian impedance control of redundant and flexible-joint robots. Springer, 2008. - Siciliano, Bruno. "Kinematic control of redundant robot manipulators: A tutorial." Journal of Intelligent and Robotic systems 3.3 (1990) - Hatanaka, Takeshi, et al. Passivity-Based Control and Estimation in Networked Robotics. Springer, 2015 - Duindam, Vincent, et al., eds. Modeling and control of complex physical systems: the port-Hamiltonian approach. Springer Science & Business Media, 2009 - Khalil, Wisama, and Etienne Dombre. Modeling, identification and control of robots. Butterworth Heinemann, 2004 - Burdet, Etienne, David W. Franklin, and Theodore E. Milner. Human robotics: neuromechanics and motor control. MIT Press, 2013 - Argall, Brenna D., et al. "A survey of robot learning from demonstration." Robotics and autonomous systems 57.5 (2009): 469-483. - Asfour, Tamim, et al. "Imitation learning of dual-arm manipulation tasks in humanoid robots." International Journal of Humanoid Robotics 5.02 (2008): 183-202. - Kober, Jens, J. Andrew Bagnell, and Jan Peters. "Reinforcement learning in robotics: A survey." The International Journal of Robotics Research 32.11 (2013): 1238-1274. - Van Hoof, Herke, et al. "Learning robot in-hand manipulation with tactile features." 2015 IEEE-RAS 15th International Conference on Humanoid Robots (Humanoids). IEEE, 2015. -...