Student Projects and Jobs

Rehabilitation Robotics

Dynamic Modelling and Impedance Control Implementation for an Intelligent Balance Board

Description:

One of the most common health problems in Germany is back pain and it is one of the most cost-intensive and often medically difficult health problems to treat [1]. The causes of back pain are mainly lack of exercise and sitting for long periods of time. The percentage of people with back pain has increased significantly over recent years and thus back pain is the second most common single diagnosis for sick leave [2]. On the other hand, a serious problem for older people is their lack of balance, which is often the cause of falls [3]. The coordination of older people can also be improved through targeted training which will result in improving the quality of life as well as the safety of seniors.

Considering the above facts, at MIRMI we are developing a three degree of freedom balance board to strengthen the user’s back muscles, balance, and overall coordination. This will be done by challenging the user to keep his/her balance on the platform during various training modes.

This work will focus on implementing impedance control on a novel platform, which is currently being designed and built as part of a collaborative research project. Specifically, your tasks will involve dynamic modelling of the platform, deriving the impedance control functions, controller implementation in Matlab/Simulink, parameter estimation from CAD drawings, parameter identification, and testing and optimization. Within this framework, you will have the opportunity to work in the exciting field of intelligent rehabilitation devices and exploit your theoretical knowledge of robotics in a practical, real-life application.

Prerequisites of an ideal candidate:

  • Passionate about robotics and its practical applications
  • Strong programming skills in Matlab and Simulink
  • Good knowledge about kinematic and dynamic modelling
  • General familiarity with SOLIDWORKS
  • Knowledge about basics of control theory (ideally including a basic understanding of impedance control)

References and related literature:

  • [1] [Online]. Available: de.statista.com/statistik/daten/studie/37519/umfrage/die-haeufigsten-volksleiden-und-syndrome-in-deutschland/.
  • [2] DAK-Gesundheitsreport 2018, [Online]. Available: www.presseportal.de/pm/50313/3891956.
  • [3] Robert Koch Institut, „Gesundheit in Deutschland aktuelle 2010,“ 2010.
  • [4] Saglia, Jody A., et al. "Design and development of a novel core, balance and lower limb rehabilitation robot: hunova®." 2019 IEEE 16th International Conference on Rehabilitation Robotics (ICORR). IEEE, 2019.
  • [5] Lee, Hyunwook, and Sehoon Oh. "Virtual ground robot for balance control." 2017 14th International Conference on Ubiquitous Robots and Ambient Intelligence (URAI). IEEE, 2017.
  • [6] Saglia, Jody A., et al. "Control strategies for patient-assisted training using the ankle rehabilitation robot (ARBOT)." IEEE/ASME Transactions on Mechatronics 18.6 (2012): 1799-1808.
  • [7] Jamwal, Prashant K., et al. "Impedance control of an intrinsically compliant parallel ankle rehabilitation robot." IEEE Transactions on Industrial Electronics 63.6 (2016): 3638-3647.

Contact:

Reihaneh Mirjalili  - reihaneh.mirjalili@tum.de

 

Robot Learning

Smart Task Board for Reproducible Robot Learning Research

Introduction

Reinforcement learning has been an established approach for solving complex tasks in robotics [1]. A critical issue that hinders the robot learning research is the lack of widely accepted benchmark frameworks for reproducible research, which have been common in other fields of AI, e.g., computer vision and natural language processing. This student project is part of an initiative within MSRM/RSI that aims to provide infrastructure to facilitate reproducible robot learning research, which emphasizes on physical interaction in manipulation.

Different from most existing benchmark frameworks for robot learning that are built upon physics simulation engines, we target for an alternative approach which uses real physical experimental setup for benchmarking. Such a physical learning environment will use IoT sensors to automate data logging and performance measurement. 

The use of IoT technology has been demonstrated in a recent proof-of-conecpt prototype, a smart task board developed for Robothon Challenge at Automatica 2021 [2], which invited competitors to tackle a robotic manipulation problem. In order to improve its next iterations and bridge the hardware to robot learning algorithms, further development and research needs to be carried out. The tasks are detailed as follows.

Type: internship/bachelor thesis/master thesis

Tasks

Below is a list of tasks for consideration. The tasks in your project will depend on the work type (intern, bachelor or master thesis) and the length.

  1. Write interfaces/APIs for implementing learning algorithms to solve the task board manipulation challenge. There will need an API to communicate with the task board and retrieve reward signals. A Python interface for controlling the robot needs to be integrated. The outcome could be an ‘Environment’ class that can take control actions and execute on the real robot, read measurement data and provide reward signals to the learning algorithms.
  2. Implementation and evaluation of existing methods in reinforcement learning for robot manipulation.
  3. Write benchmark protocol for using the smart task board for robot learning research, with some examples.
  4. Various research directions could be explored, e.g., skill representation, learning adaptive impedance control policy, sequential manipulation planning/learning

Prerequisites:

For task 1

  • good Python
  • embedded programming would be a plus;

For task 2-4 

  • Basic knowledge about machine learning and/or reinforcement learning
  • Basic knowledge about robotics and control theory
  • Very good Python and C++ skills

If you are interested, please contact: f.wu@tum.de

[1] J. Kober, J. A. Bagnell, and J. Peters, “Reinforcement learning in robotics: A survey,” Int. J. Robot. Res., vol. 32, no. 11, pp. 1238–1274, Sep. 2013.

[2] Robothon Grand Challenge, www.robothon-grand-challenge.com

Human-Robot Interaction

Biomechanics Aware Sequential Planner for Human-Robot Interaction

Empowering robots to physically engage and interact with humans is one of the key challenges of today’s robotics. When autonomously interacting with humans, robots need better decision-making which is hindered by the lack of a reliable quantitative assessment of human physical capabilities and limitations. At MSRM, we are developing planning and control algorithms for robot manipulators/humanoids based on human biomechanics and ergonomics which will push service robots and intelligent manufacturing to a new era of physical human-robot collaboration (pHRC).
This research internship/master thesis program is focused on implementing, developing, and contributing to pHRC algorithms driven by human biomechanics response, ergonomics, and motion recognition and prediction.
In this work, you will have the opportunity to work with state-of-the-art robots and algorithms and be part of the exciting community that drives robotics forwards. Indeed, be excited and passionate about the research is a fundamental prerequisite. You have the right support, but you will also be expected to work as hard as we do.


Tasks may include few of the below (to be discussed depending on your interest and background): 

- Implement/integrate grasp planners with motion planners;
- Implement/design impedance/admittance controllers;
- Implement sequential motion planner on a dual-arm system and/or humanoid robot;
- Integrate existing biomechanics model into ROS for biomechanics-aware planning;
- Implementing biomechanics based human-manipulability;  
- Design ergonomics-based hand-over and pHRC;
- Implement/Design biomechanics-intention-aware reactive motion planning;
- Integrate motion capture systems for pHRC experiments;
- Skeleton tracking integration with biomechanics metrics;
- Human motion intention recognition;
- EMGs setup for experiments and post-processing of the EMG data;
- Inverse muscular-activity estimation based on experimental data.


Type: Research Internship, Master Thesis, Master Internship
(The tasks can be split into different workpackages)


Pre-requisites:
- Motivation to work in a team and driving environment
- Good C++ / Python / Matlab skills
- Basic knowledge about robotics OR biomechanics (particularly, for the upper-limb) 


Helpful depending on the tasks/position: 
- Experience with grasp planners 
- Experience with control or planners in robotics 
- Experience with Opensim 
- Knowledge of sEMG data capture and analysis
- Experience with motion capture systems  


Related literature

[1] Chen, L., Figueredo, L.F.C. & Dogar, M.R. Manipulation planning under changing external forces. Auton Robot 44, 1249–1269 (2020). doi.org/10.1007/s10514-020-09930-z

[2] L. F. C. Figueredo, R. C. Aguiar, L. Chen, S. Chakrabarty, M. R. Dogar and A. G. Cohn, "Human Comfortability: Integrating Ergonomics and Muscular-Informed Metrics for Manipulability Analysis During Human-Robot Collaboration," in IEEE Robotics and Automation Letters, vol. 6, no. 2, pp. 351-358, April 2021, doi: 10.1109/LRA.2020.3043173.

[3] L. Chen, L. F. C Figueredo and M. R. Dogar, "Planning for Muscular and Peripersonal-Space Comfort During Human-Robot Forceful Collaboration," 2018 IEEE-RAS 18th International Conference on Humanoid Robots (Humanoids), 2018, pp. 1-8, doi: 10.1109/HUMANOIDS.2018.8624978.

[4] Riddhiman Laha, Luis F.C. Figueredo, Juraj Vrabel, Abdalla Swikir, and Sami Haddadin. "Reactive Cooperative Manipulation using Set Primitives and Circular Fields." 2021 IEEE International Conference on Robotics and Automation (ICRA), Xi'an, China.

 

For more information, please contact:

Dr Luis Figueredo

 

 

Point-to-Point Motion Planning and Visual Servoing using User Guidance

Despite the increasing number of collaborative robots in human-centered manufacturing, up to today, industrial robots are still largely preprogrammed with very few autonomous features. At MSRM, we developing novel strategies based on single user guidance motion generation that facilitates changes in the production line in a timely and easy-to-implement fashion.
This research internship/master thesis program is focused on implementing, developing, and contributing to point-to-point motion generation and visual servoing based on purely geometric information from a single user demonstration.
In this work, you will have the opportunity to work with state-of-the-art robots and algorithms and be part of the exciting community that drives robotics forwards. Indeed, be excited and passionate about the research is a fundamental prerequisite. You have the right support, but you will also be expected to work as hard as we do.  


Tasks may include a few of the below (to be discussed depending on your interest and background):
- Implement motion interpolation algorithms using existing dual quaternions lib. (available in c++, python, matlab);  
- Implement motion interpolation from one-shot user demonstrations;
- Extract geometric features from robotic kinesthetic demonstration;  
- Extract geometric features from skeleton tracking;  
- Design/implement path and motion planning from interpolation poses;
- Design/implement visual tracking algorithms;
- Design interpolation algorithms for bimannual manipulator;


Type: Research Internship, Master Thesis, Master Internship, Bachelor Thesis
(The tasks can be split into different work packages)


Pre-requisites:
- Motivation to work in a team and driving environment
- Good C++ / Matlab skills
- Good math(algebra) and/or control skills
- Basic experience with RGBD sensors

Helpful but not required
- Experience with ROS
- Experience with robot manipulators
- Python


Related literature

[1] B. V. Adorno and M. Marques Marinho, "DQ Robotics: A Library for Robot Modeling and Control," in IEEE Robotics & Automation Magazine, doi: 10.1109/MRA.2020.2997920.

[2] A. Sarker, A. Sinha and N. Chakraborty, "On Screw Linear Interpolation for Point-to-Point Path Planning," 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2020, pp. 9480-9487, doi: 10.1109/IROS45743.2020.9341651.

[3] Riddhiman Laha, Luis F.C. Figueredo, Juraj Vrabel, Abdalla Swikir, and Sami Haddadin. "Reactive Cooperative Manipulation using Set Primitives and Circular Fields." 2021 IEEE International Conference on Robotics and Automation (ICRA), Xi'an, China.

[4] R. Laha, A. Rao, L. F.C. Figueredo, Q. Chang, S. Haddadin, N. Chakraborty, "Point-to-point Path Planning based on User Guidance and Screw Linear Interpolation," in ASME International Design Engineering Technical Conferences and Computers and Information in Engineering Conference IDETC/CIE2021.

 

For more information, please contact:

Dr Luis Figueredo

Riddhiman Laha (riddhiman.laha@tum.de)

Human-Motion Aware Collision Handling and Avoidance during Motion Planning

Recent advances in robotics technologies are closing the gap between humans and robots. Still, fluent and safe interactions of humans and robots require both partners to understand and anticipate each other's actions and to be able to move and react accordingly in a timely fashion. At MSRM, we are developing state-of-the-art human models, robotic controllers, and planners that will enhance robot capabilities and autonomy during human-robot interaction and collaboration.
One of the bottlenecks in existing methods is a proper integration between human motion estimation and prediction with reactive motion behavior from the robotic system. In this work, you will have the opportunity to work with state-of-the-art robots and algorithms, and be part of the exciting community that drives robotics forwards. Indeed, be excited and passionate about the research is a fundamental prerequisite. You have the right support, but you will also be expected to work as hard as we do.

Tasks may include a few of the below (to be discussed depending on your interest and background):
- RGBD-based skeleton tracking;
- Trajectory and skeleton prediction for the human motion (unsupervised approach);  
- Imitation learning from human demonstration;
- Probabilistic intention-aware human motion from demonstrations;
- Motion-intention-aware reactive motion planning;
- Biomechanics-intention-aware reactive motion planning;
- Intention-aware human-robot collaboration.


Type: Research Internship, Master Thesis, Master Internship, Bachelor Thesis
(The tasks can be split into different work packages)


Pre-requisites:
- Motivation to work in a team and driving environment
- Good C++ / Python skills
- Good experience with RGBD sensors


Helpful but not required
- Basic experience with ROS
- Basic knowledge about robotics/manipulators


Related literature

[1] Butepage, J., Kjellstrom, H., & Kragic, D. (2018). Anticipating Many Futures: Online Human Motion Prediction and Generation for Human-Robot Interaction. Proceedings - IEEE International Conference on Robotics and Automation, 4563–4570. doi.org/10.1109/ICRA.2018.8460651

[2] D. Koert, J. Pajarinen, A. Schotschneider, S. Trick, C. Rothkopf and J. Peters, "Learning Intention Aware Online Adaptation of Movement Primitives," in IEEE Robotics and Automation Letters, vol. 4, no. 4, pp. 3719-3726, Oct. 2019, doi: 10.1109/LRA.2019.2928760.

[3] Riddhiman Laha, Luis F.C. Figueredo, Juraj Vrabel, Abdalla Swikir, and Sami Haddadin. "Reactive Cooperative Manipulation using Set Primitives and Circular Fields." 2021 IEEE International Conference on Robotics and Automation (ICRA), Xi'an, China.

[4] L. F. C. Figueredo, R. C. Aguiar, L. Chen, S. Chakrabarty, M. R. Dogar and A. G. Cohn, "Human Comfortability: Integrating Ergonomics and Muscular-Informed Metrics for Manipulability Analysis During Human-Robot Collaboration," in IEEE Robotics and Automation Letters, vol. 6, no. 2, pp. 351-358, April 2021, doi: 10.1109/LRA.2020.3043173.

 

For more information, please contact:

Dr Luis Figueredo

Riddhiman Laha (riddhiman.laha@tum.de)

Prosthetics

Mechatronics System Developement

Forschungspraxis: Experimental Control Performance Testing and Model Identification of a Pneumatic Acuation Unit

I am always looking for talented students who are interested in doing their internship (Forschungspraxis) in the following fields:

  • mechatronic systems and robot development (CAD + system design)
  • optimization and identification, 
  • control and observers,
  • pneumatic acutators,
  • modeling of multibody systems and actuators,
  • experimental data analysis and signal processing.

It would be benefitial if you have already gained some experience in Matlab Simulink or Solidworks. It might also be possible to write a master thesis after the internship.

Please contact me by the following address:
alexander.toedtheide@tum.de

Development of compliant actuator development algorithm and toolbox

Compliant actuators are a hot topic in robotics. Even though most of the developed systems are unique in their fields, they are mostly a product of the integration of off-the-shelf parts. However, the development process relies on experience, intuition, and trial error. The main goal of this research opening is to develop an inclusive algorithm to select necessary components for the compliant actuator such as gearbox, motor, cooling system, etc. 

 

Type: Research Internship, Master Thesis, Master Internship

(The tasks can be split into different work packages)

 

Pre-requisites:

  • Prior knowledge of Matlab/Simulink
  • Basic knowledge of actuator models
  • Basic knowledge of mechanics, fatigue, wear, thermal models
  • Strong background in dynamic modeling
  • Knowledge of optimization

 

Application:

Please send these documents to [mehmet (dot) yildirim (at) tum (dot) de]

  • Transcript
  • CV
  • Max. 200 words, letter of intention
  • Portfolio of previous projects

 

Legged Locomotion

Dynamics Algorithms – Closed form Computation of the EoM

Type: Research Internship

Description:

Due to the increasing complexity of robots, especially in the field of legged robots, the computation of the equations of motion has received more and more attention throughout the last decades. The spatial formulation of dynamic quantities and algorithms such as the recursive Newton-Euler algorithm (RNEA) and the composite rigid-body algorithm (CRB) are some examples that underline the progress that has been made in the field. However, the “algorithmic toolbox” that is available in most frameworks today still lacks for example the computation of state and partial derivatives, the computation of the regressor, the explicit computation of the Coriolis matrix, etc.

The Rigid Body Dynamics Library (RBDL) developed by [1] is based on Featherstone’s spatial formulation of the RNEA and CRB. It also contains methods to compute the forward kinematics and Jacobians for an arbitrary tree-like structure. However, like most other libraries, it lacks the explicit numerical computations of some quantities that might result in better control laws. To get past these shortcomings, the goal of this student project is to augment the RBDL library with a novel iterative method introduced by [2]. With this algorithm, it is possible to obtain the closed-form solution of the EoM as well as their derivatives.

Prerequisites:

  • Strong background in kinematic and dynamic modeling
  • Strong C programming skills

Literature:

[1] Felis, Martin L. "RBDL: an efficient rigid-body dynamics library using recursive algorithms." Autonomous Robots 41.2 (2017): 495-511.

[2] Garofalo, Gianluca, Christian Ott, and Alin Albu-Schäffer. "On the closed form computation of the dynamic matrices and their differentiations." 2013 IEEE/RSJ International Conference on Intelligent Robots and Systems. IEEE, 2013.

Contact:
Dennis Ossadnik
dennis.ossadnik@tum.de

Robot Safety: Bachelor/Master Thesis | Research/Engineering-Internship

Collision Analysis and Safe Control in Human-Robot Interaction

Currently, increasing effort is taken in the robotics community to understand injury mechanisms during physical human-robot interaction (pHRI). This is motivated by the fact that human and robot will work intensively and closely together, and therefore, one has to be aware of the potential threats in case such a close cooperation takes place and take appropriate countermeasures to ensure human safety via planning and/or control. In the context of safety in pHRI, possible topics that can be addressed in the thesis/internship are:

  • Design and analysis of collision experiments and/or testing devices
  • Development and verification of collision simulations
  • Survey of biomechanics and forensics literature
  • Development of robot motion planning and/or control schemes for ensuring human safety

Prerequisites:

  • Studies in Mechanics, Mechatronics, Electronics, Computer Science 
  • Knowledge in robotics & control (for topics on planning & control)
  • Good C++ programming skills
  • Matlab/Simulink
  • Working knowledge in ROS
  • Ability to work well structured and organized
  • Creativity

Contact:
Mazin Hamad, M.Sc.
mazin.hamad@tum.de
Chair of Robotics Science and Systems Intelligence

AI-Enabled Lab-Automation

AI unterstützter robotischer Lab Assistent: Zuverlässiges automatisiertes Flüssigkeitshandling für den Laboralltag der Zukunft

Typ: Forschungs-/Ingenieurpraktikum

→ Für weitere Informationen kontaktieren Sie uns!

Die Automatisierung von Laborprozessen in der Chemie, Bio-, Pharma- und Lebensmitteltechnologie sowie in der Medizin ist bereits heute Realität. Doch viele Lösungen, die heute auf dem Markt angeboten werden, sind entweder zu hochpreisig und/oder nur speziell und unflexibel für bestimmte Prozesse im Labor entwickelt und optimiert. Damit die Laborautomatisierung für den dynamischen Laboralltag als Werkzeug für Jedermann genutzt werden kann, soll in diesem Anwendungsforschungsprojekt ein mit AI unterstützter Automatisierungsablauf für einen robotischen Laborassistenten entwickelt und im Labor umgesetzt werden. Als zu automatisierender Prozess soll vorerst das zuverlässige automatisierte Flüssigkeitshandling angegangen werden. Um dies zu erreichen steht einer der neusten kollaborativen Roboter zur Verfügung sowie 3D-Druck-Möglichkeiten zur flexiblem Gripperfingerentwicklung.

Folgender Ablauf ist geplant:

  • Ausführliche Literaturrecherche Robotik in der Laborautomatisierung
  • Analyse und Entwicklung von Roboterfingersystemen zur Nutzung von Laborwerkzeugen
  • Analyse, Automatisierung und Evaluierung von Prozessabläufen in Bezug auf Flüssigkeitshandling im Labor
  • Und vieles mehr …

Voraussetzungen:

  • Aus dem Fachbereich Elektrotechnik und Informationstechnik, Maschinenbau oder Mechatronik
  • Grundwissen in Robotik, Regelungstechnik und Systemtheorie
  • Gute Programmierkenntnisse in C/C++, Python, Matlab
  • Gute Fähigkeiten in CAD-Design (SolidWorks usw.)
  • Erfahrung mit 3D-Druck

Kontakt:
Dennis Knobbe
dennis.knobbe@tum.de
+49 (89) 289 - 29412

Human modeling

3D-Modellierung und Echtzeitvisualisierung der Muskelverformung

Es gibt heutzutage zahlreiche 3D computergraphische Modelle für muskuloskelettale Systeme, z.B.
- statisches Modell,
- bewegliches aber rechenintensives Modell
- echtzeitfähiges Modell mit aber abstrahierter Muskeldarstellung.
Diese Arbeit handelt sich um eine algorithmische Kombination der vorteilhaften Merkmale von den Stand-der-Technik-Modelle, und zwar eine rechnerische Lösung zur echtzeitfähiger 3D-Visualisierung mit sowohl anatomisch korrekter (Muskel-)Darstellung als auch diversen Bewegungsfreiheitsgrade. Hierbei ist es eine rechnerisch effiziente Muskelverformung während der Bewegungen eine Herausforderung. 

Wir erwarten von Ihnen
- Grundkenntnis der Mechanik und der Mehrkörpersysteme (Kinematik, Statik)
- Grundkenntnis numerischer Mathematik
- Objektorientierte Programmierung (C++ oder C#)
- Interesse an Visualisierung der Anatomie und Entwicklung von Computerspielen

Wir bieten
- Gut gestalteten Arbeitsplatz
- Fachliche Betreuung

Kontakt:
M.Sc. M.Sc. Tingli Hu
tingli.hu@tum.de