Recupera REHA

Full-body exoskeleton for upper body robotic assistance

The aim of Recupera REHA is the development of an innovative and mobile full-body exoskeleton, as well as and of an active subsystem as an independent unit for robot-assisted rehabilitation of neurological diseases. The full-body system is intended to capture the kinematics of the entire human body, to support itself, and to work power self-sufficient. New methods dealing with actuation, lightweight construction, and control engineering will be elaborated for the development of the system. The mechatronic approaches are combined with a new methods for online-evaluation of EEG/EMG signals, in order to allow an evaluation of the operating person’s condition and a multistage support of the exoskeleton’s control system. Innovations from the full-body system will be used to develop the independent unit and investigate fundamental principles and approaches for treatment and rehabilitation. The close cooperation with the joint project partner rehaworks allows to evaluate the medical suitability of the developed components. Further it opens up new perspectives for development of prototypical robotic rehabilitation devices. The performance and the ergonomics of the systems will be investigated in a subsequent evaluation phase. Finally, criteria for a pilot study in the field of upper body rehabilitation will be defined.

Duration: 01.09.2014 till 31.12.2017
Donee: German Research Center for Artificial Intelligence GmbH
Sponsor: Federal Ministry of Education and Research
German Aerospace Center e.V.
Grant number: This project is funded by the Federal Ministry of Education and Research, DLR, Project Management Agency, Software Systems and Knowledge Technologies, grant no. 01IM14006A.
Partner: rehaworks GmbH
Application Field: Assistance- and Rehabilitation Systems
Related Projects: VI-Bot
Virtual Immersion for holistic feedback control of semi-autonomous robots (01.2008- 12.2010)
Dual-arm exoskeleton (01.2011- 12.2013)
Intelligent Man-Machine Interface - Adaptive Brain-reading for assistive robotics (05.2010- 04.2015)
Related Robots: Dual Arm Exoskeleton
Exoskeleton for upper body robotic assistance (Recupera REHA)
Full Body Exoskeleton
Exoskeleton for upper body robotic assistance
Exoskeleton Passive (CAPIO)
Upper body Human-Machine-Interface (HMI) for tele-operation
Exoskeleton Passive (VI-Bot)
Upper body exoskeleton (right arm) for motion capturing
Related Software: pySPACE
Signal Processing and Classification Environment written in Python
Computer Aided Design To Simulation
Reconfigurable Signal Processing and Classification Environment

Project details

The Recupera REHA Full-body exoskeleton for upper body robotic assistance. (Source: Meltem Yilmaz, DFKI GmbH)
The Recupera REHA subsystem supports movements for instance in an authentic setting. (Source: Annemarie Popp, DFKI GmbH)
The Recupera REHA subsystem. The system covers the entire kinematics of the human arm. (Source: Annemarie Popp, DFKI GmbH)
Sketch of possible exoskeleton kinematic (Source: David Grünwald, DFKI GmbH)
Biosignals before and during one movement. A: Electroencephalogram (EEG) at different scalp positions B: Electromyogram (EMG) of different muscles (Source: Marc Tabie, DFKI GmbH)

Bidirectional man-machine interaction

This specific aspect of the project is focused on the development of application concepts and their specific requirements for dealing with patients in the context of motor rehabilitation.  The aim is to create synergies between man and machine in order to optimize processes and the workflow of rehabilitation, as well as to provide patients and therapists with advanced and innovative therapy options on the basis of this new technology.

Development of a mechatronic system

The task in this research field is the development of an intrinsically safe robotic system. This development work includes the creation of a kinematic structure adapted to the human being, the development of appropriate drives, and provision of an electronic system. These components are then assembled to an overall system and integrated into an additional independent system for upper body rehabilitation.

Kinematics and dynamics

With the aim of building a novel full-body exoskeleton, the specific work in kinematics and dynamics deals with solving different tasks: During the design phase of the exoskeleton, different aspects of the kinematic chain of the system are defined. This comprises the selection of joint types and the length of exoskeleton limbs. The dynamic synthesis deals with the evaluation of the masses in the system and the resulting forces needed from the actuators. In order to support additional aspects in the development and optimal operation of the exoskeleton, further issues are addressed, which deal with e.g. the synchronous evaluation of movement and sensomotoric data, the assessment of the users behavior and the patient dependent control of the system.

Adaptive control system

The aim is to develop a modular, multi-level control structure enabling a dynamic control of the exoskeleton. The basic architecture will consist of three hierarchical layers. The control system will feature selectable single control modules, which can be combined with each other in various ways. One of the core tasks is the development of assistive control strategies for rehabilitation purposes.

Biosignal integration

In this field of research, conditions are created to transfer electroencephalographic and electromyographic activity (EEG and EMG) by means of embedded Brain Reading (see a different application video here) into robot assisted rehabilitation. It is planned that the bio-signals are used as an additional element of the exoskeleton control in order to achieve an optimal adaptation to the patient’s needs. In particular this requires the acquisition of EEG and EMG signals and the development of new automatic procedures for labeling and evaluating the signals in the context of e.g. applied therapy. Further, activity models from healthy persons will be generated in order to compare them with activities of patients and in this way, e.g. analyze the rehabilitation progress. The open source software pySPACE (Signal Processing And Classification Environment written in Python) is used to analyze EEG and EMG data, but also to examine the potential of integrating further data from the robotic system. pySPACE supports the creation of activity models as well as the adaptation of the robotic system to the patient. Numerous processing algorithms can be compared and optimized and the results can be presented to the partners. This includes the adjustment of the processing during runtime. Only after the optimization step, it is possible to have a fixed embedding of the developed processing approach into the exoskeleton.

Embedded Data Processing

In order to achieve the autonomy that is required for rehabilitation applications, all processing (kinematics/dynamics, control, biosignal processing) has to be performed by a small computing system that is embedded into the exoskeleton itself. In order to provide a sufficient amount of computing power and still meet the strict requirements regarding physical space and power consumption, FPGAs will be used. The time-critical computations can then be performed by application specific hardware accelerators. To implement these hardware accelerators, the framework reSPACE (reconfigurable Signal Processing And Classification Environment) will be used. In this context, reSPACE will be extended to support complex control algorithms and kinematical computations.


Recupera REHA: operating principle and usage of the Recupera REHA subsystem

Recupera REHA: Animation of the operating principle and usage of the systems

Recupera REHA: Therapeutic application example of the subsystem

Capio Exoskeleton: Control via biosignals

Demonstration of the Capio exoskeleton control via biosignals: The intended movement of the human operator is detected by the biosignal data processing which triggers the execution of the targeted movement by the exoskeleton. By means of an eye tracker the desired interaction is detected (focusing on a virtual bottle) and by electroencephalographic signals (EEG), the intended movement and the performing limb are determined. Furthermore, by means of electromyographic signals (EMG), the intended movements are verified.



Exoskelette und künstliche Intelligenz in der klinischen Rehabilitation
Elsa Andrea Kirchner, Niels Will, Marc Simnofske, Luis Manuel Vaca Benitez, José de Gea Fernández, Peter Kampmann, Frank Kirchner
Editors: Mario A. Pfannstiel, Patrick Da-Cruz, Harald Mehlich
In Digitale Transformation von Dienstleistungen im Gesundheitswesen V, Springer Nature, chapter 21, pages 413-435, Aug/2019. ISBN: 978-3-658-23986-2.
Modular Design and Decentralized Control of the Recupera Exoskeleton for Stroke Rehabilitation
Shivesh Kumar, Hendrik Wöhrle, Mathias Trampler, Marc Simnofske, Heiner Peters, Martin Mallwitz, Elsa Andrea Kirchner, Frank Kirchner
In Applied Sciences, MDPI, volume 9, number 4 (626), pages 1-23, Feb/2019.
Embedded Multimodal Interfaces in Robotics: Applications, Future Trends, and Societal Implications
Elsa Andrea Kirchner, Stephen Fairclough, Frank Kirchner
Editors: S. Oviatt, B. Schuller, P. Cohen, D. Sonntag, G. Potamianos, A. Krueger
In The Handbook of Multimodal-Multisensor Interfaces, Morgan & Claypool Publishers, volume 3, chapter 13, pages 523-576, 2019. ISBN: e-book: 978-1-97000-173-0, hardcover: 978-1-97000-175-4, paperback: 978-1-97000-172-3, ePub: 978-1-97000-174-7.
Künstliche Intelligenz und robotergestützte Rehabilitation
Niels Will, Elsa Andrea Kirchner, Frank Kirchner
Editors: Heinrich Hanika
In Künstliche Intelligenz, Robotik und autonome Systeme in der Gesundheitsversorgung, Wissenschaft & Praxis, series Schriften zu Gesundheitsökonomie / Gesundheitsmanagement. Hrsg. Manfred Erbsland und Evelin Häusler, pages 101-128, 2019. ISBN: 978-3-89673-759-5.


Design and Kinematic Analysis of the Novel Almost Spherical Parallel Mechanism Active Ankle
Shivesh Kumar, Bertold Bongardt, Marc Simnofske, Frank Kirchner
In Journal of Intelligent & Robotic Systems, Springer Nature, volume 94, number 2, pages 303-325, Mar/2018.
CAEMO - A Flexible and scalable high performance matrix algebra coprocessor for embedded reconfigurable computing systems
Hendrik Wöhrle, Frank Kirchner
In Microprocessors and Microsystems, Elsevier, volume o.A., pages 47-63, Feb/2018.
Mechatronical design and analysis of a modular developed exoskeleton for rehabilitation purposes
Mehmed Yüksel, Luis Manuel Vaca Benitez, Dinmukhamed Zardykhan, Frank Kirchner
In Proceedings of ELECO 2017, (ELECO-2017), 29.11.-2.12.2017, Bursa, IEEE Xplore, Jan/2018. ISBN: 978-1-5386-1723-6.


Eingebettete Biosignalverarbeitung und integrierte Regelung eines Ganzkörper-Exoskelettes für die Neurorehabilitation
Hendrik Wöhrle, Elsa Andrea Kirchner
In Proceedings of the 2. VDI Fachkonferenz Humanoide Roboter, 05.12.-06.12.2017, München, VDI Fachkonferenz Humanoide Roboter, Dec/2017.
Adaptive multimodal biosignal control for exoskeleton supported stroke rehabilitation
Anett Seeland, Marc Tabie, Su-Kyoung Kim, Frank Kirchner, Elsa Andrea Kirchner
In IEEE International Conference on Systems, Man, and Cybernetics, (SMC-2017), 05.10.-08.10.2017, Banff, IEEE, Oct/2017.
Rotational Data Augmentation for Electroencephalographic Data
Mario Michael Krell, Su-Kyoung Kim
In Proceedings of the 39th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, (EMBC-17), 11.7.-15.7.2017, JeJu Island, South Korea, IEEE, Jul/2017.
A Hybrid FPGA-Based System for EEG- and EMG-Based Online Movement Prediction
Hendrik Wöhrle, Marc Tabie, Su-Kyoung Kim, Frank Kirchner, Elsa Andrea Kirchner
In Sensors - Open Access Journal, MDPI, volume 17, number 7, pages 1552, Jul/2017.
Labelling of Movement onsets based on Exoskeleton Joint Data
Marc Tabie, Anett Seeland, Su-Kyoung Kim, Elsa Andrea Kirchner
In Proceedings of the 1st Neuroadaptive Technology Conference 2017, (NAT-17), 19.7.-21.7.2017, Berlin, o.A., Jul/2017.
Classifier Transfer with Data Selection Strategies for Online Support Vector Machine Classification with Class Imbalance
Mario Michael Krell, Nils Wilshusen, Anett Seeland, Su-Kyoung Kim
In Journal of Neural Engineering, IOP Publishing, volume 14, number 2, pages 025003, Feb/2017.
Kinematic analysis of Active Ankle using computational algebraic geometry
Shivesh Kumar, Abhilash Nayak, Bertold Bongardt, Andreas Mueller, Frank Kirchner
In Computational Kinematics, (CK-2017), 22.5.-24.5.2017, Poitiers, Springer, 2017.
Inverse Kinematics of Anthropomorphic Arms Yielding Eight Coinciding Circles
Bertold Bongardt
In Computational Kinematics, (CK-2017), Poitiers, Springer, 2017.
Integrating Mimic Joints into Dynamics Algorithms – Exemplified by the Hybrid Recupera Exoskeleton
Shivesh Kumar, Marc Simnofske, Bertold Bongardt, Andreas Mueller, Frank Kirchner
In Proceedings of the 2017 Conference on Advances In Robotics, (AIR-2017), 28.6.-02.7.2017, New Delhi, ACM-ICPS, 2017.


Recupera-Reha: Exoskeleton Technology with Integrated Biosignal Analysis for Sensorimotor Rehabilitation
Elsa Andrea Kirchner, Niels Will, Marc Simnofske, Luis Manuel Vaca Benitez, Bertold Bongardt, Mario Michael Krell, Shivesh Kumar, Martin Mallwitz, Anett Seeland, Marc Tabie, Hendrik Wöhrle, Mehmed Yüksel, Anke Heß, Rüdiger Buschfort, Frank Kirchner
In 2. Transdisziplinäre Konferenz "Technische Unterstützungssysteme, die die Menschen wirklich wollen", 12.12.-13.12.2016, Hamburg, Elsevier, pages 504-517, Dec/2016.
hyperSPACE:Automated Optimization of Complex Processing Pipelines for pySPACE
Torben Hansing, Mario Michael Krell, Frank Kirchner
In BayesOpt2016 - Bayesian Optimization: Black-box Optimization and Beyond, (BayesOpt-2016), 10.12.2016, Barcelona, n.n., Dec/2016.
Task space controller for the novel Active Ankle mechanism
Shivesh Kumar, Bertold Bongardt, Marc Simnofske, Frank Kirchner
In International Conference on Robotics and Automation for Humanitarian Applications, (RAHA-16), 18.12.-20.12.2016, Amritapuri, Kerala, IEEE, series RAHA 2016 Poster Proceedings, pages 22, Kerala, India, Dec/2016. Amrita University.
Active Ankle - an Almost-Spherical Parallel Mechanism
Marc Simnofske, Shivesh Kumar, Bertold Bongardt, Frank Kirchner
In 47th International Symposium on Robotics (ISR 2016), (ISR), 21.6.-22.6.2016, Munich, VDE Verlag, pages 37-42, Jun/2016.
Rekonfigurierbare Datenflussarchitekturen in der Robotik - Zukünftige robotische Systeme benötigen dezentrale und verteilte Rechenarchitekturen für Intelligenz und Autonomie
Hendrik Wöhrle, Frank Kirchner
In Industrie 4.0 Management, GITO Verlag, volume 2, number 4, pages 25-28, Mar/2016.
Embedded Brain Reading - Sichere und intuitive Mensch-Maschine-Interaktion
Elsa Andrea Kirchner, Rolf Drechsler
In Industrie 4.0 Management, Gito mbH Verlag für Industrielle Informationstechnik und Organisation, volume 4, number 2/2016, pages 37-40, Mar/2016.
Movement identification based on exoskeleton sensor data for event marking of the electroencephalogram
Nils Eckardt, Marc Tabie, Anett Seeland, Elsa Andrea Kirchner, P. Rostalski
In Student Conference Proceedings 2016: 5th Conference on Medical Engineering Science and 1st Conference on Medical Informatics, 09.3.-11.3.2016, Lübeck, Infinite Science Publishing, pages 151-154, Mar/2016. ISBN: 3945954185.


Backtransformation: A new representation of data processing chains with a scalar decision function
Mario Michael Krell, Sirko Straube
In Advances in Data Analysis and Classification, Springer, volume 11, number 2, pages 415-439, Dec/2015.
Comparison of Data Selection Strategies For Online Support Vector Machine Classification
Mario Michael Krell, Nils Wilshusen, Andrei Cristian Ignat, Su-Kyoung Kim
In Proceedings of the International Congress on Neurotechnology, Electronics and Informatics (, (NEUROTECHNIX-2015), 16.11.-17.11.2015, Lissabon, SciTePress, pages 59-67, Nov/2015.
Unified View on Complex Numbers and Quaternions
Bertold Bongardt
In The 14th IFToMM World Congress, 25.10.-30.10.2015, Taipei, The 14th IFToMM World Congress, Oct/2015.
Spatio-temporal Comparison between ERD/ERS and MRCP-based Movement Prediction
Anett Seeland, Laura Manca, Frank Kirchner, Elsa Andrea Kirchner
In Proceedings of the 8th International Conference on Bio-inspired Systems and Signal Processing (BIOSIGNALS-15), (BIOSIGNALS-15), 12.1.-15.1.2015, Lisbon, ScitePress, pages 219-226, Jan/2015.
raxDAWN: Circumventing Overfitting of the Adaptive xDAWN
Mario Michael Krell, Anett Seeland, Hendrik Wöhrle
In Proceedings of the International Congress on Neurotechnology, Electronics and Informatics (, (NEUROTECHNIX-2015), 16.11.-17.11.2015, Lissabon, SciTePress, pages 68-75, 2015.

Back to the list of projects
last updated 04.01.2024