Design of a demo experimental setup for human augmentation

State of the art

Modern research on rehabilitation robotics are exploring various ways to cope with the interaction between robots and humans. One of these ways is to avoid forcing the human’s movement to a predefined trajectory, and to permit the human the complete control of the device at any moment. Hence, the robot should detect the user’s movement intention and react accordingly to assist the movement. New techniques are being developped, which can be through hardware adaptations, e.g. elastic elements between the actuators of the device and the human (Series Elastic Elements, SEE), or by a high level control using adaptive oscillators. The latter is able to synchronize to the human’s movement and to estimate the movements parameters (e.g. position, velocity, and acceleration). In consequence, the device can estimate the necessary torque/force to perform the movement and assists the human by a fraction of this torque.

Content of the project

This project is the continuation of the previous work carried out by F. A. Delaloye, who designed the mechanical setup used during this semester project. The goals of this project were defined similar to the structure of this report and were very multifarious going from mechanical design to high level control and some experiments with healthy humans:

  • Finalization of the mechanical design of the actual setup (Chap. 2)
  • Designing a new, flexible harware setup including several safety improvements (Chap. 2)
  • Finalization of the electronics (Chap. 3)
  • Finalization of the low level control (Chap. 4)
  • Establishment of the communication between the Maxon EPOS and MATLAB (Chap. 5)
  • Creating an easy to use library in MATLAB and Simulink with functions / blocks (Chap. 5)
  • Assistance using an adaptive oscillator and an inverse dynamical model for torque estimation (Chap. 5)
  • Setting up an experimental protocol (Chap. 6)
  • Performing experiments with healthy people and analyzing the results (Chap. 6)
  • Writing a publication on the experimental results for the ICORR 2011 taking place in Zurich, Switzerland (corresponds to Chap. 6)

Note: Chapter 5 is thought as an introducing guide about the sofware libraries and its functionalities for new students continuing this project.

Future work

I would be very pleased if I was contacted by the new student at the begining of his/her project in order to arrange a small meeting to discuss my work and to give some personal explanations.

If there are any questions about the new mechanical design or the software, please feel free to contact me by email (mikedomenik.rinderknecht at


 M. D. Rinderknecht, F. A. Delaloye, A. Crespi, R. Ronsse and A. J. Ijspeert, “Assistance using adaptive oscillators: Sensitivity analysis on the resonance frequency”, in Proc. International Conference on Rehabilitation Robotics, Zurich, Switzerland, June 2011. [SUBMITTED]


This paper provides a robustness analysis of the method we recently developed for rhythmic movement assistance using adaptive oscillators. An adaptive oscillator is a mathematical tool capable of extracting high-level features (i.e. amplitude, frequency, offset) of a quasi-sinusoidal measured movement, a rhythmic flexion-extension of the elbow in this case. By the use of a simple inverse dynamical model, the system can predict the torque produced by a human participant, such that a fraction of this estimated torque is fed back through a series elastic actuator to provide movement assistance. This paper objectives are twofold. First, we introduce a new 1 DOF assistive device developed in our lab. Second, we derive model-based predictions and conduct experimental validations to measure the variations in movement frequency as a function of the open parameters of the inverse dynamical model. As such, the paper provides an estimation of the robustness of our method due to model approximations. As main result, the paper reveals that the movement frequency is particularly robust to errors in the estimation of the damping coefficient. This is of high interest for the applicability of our approach, this parameter being in general the most difficult to quantify.