Master-Thesis: System Identification of a Variable Stiffness Actuator using Linear Parameter Varying

2 July 2013

Nowadays, novel types of robotic systems are able to adjust in addition to an actively controlled impedance a passive intrinsic impedance, namely stiffness. Thus, similar to the human these robots are able to interact in an uncertain environment and change their compliance with respect to whatever is required by the task. The mechanisms for adjusting the stiffness are human-inspired. However, the behavior of how human change their stiffness with respect to different tasks aren’t yet fully understood, because of lacking the possibility to measure human stiffness continuously. Thus, we at DLR developed in a student project a measurement device that is able to apply a set of high frequency perturbations and measure reaction forces simultaneously. From this set of information it will be possible to calculate the full impedance and thus stiffness and we will be able to investigate human stiffness adaption behavior, e.g. during a reach-to-grasp tasks.


DLR Hand Arm System --- capable
of changing the intrinsic joint stiffness

Arm Perturbator for applying and
measure continuous perturbations

The topic of this thesis is to investigate several methods of system identification, as e.g. Linear Parameter Varying. As a ground truth and to prove the capability of the system identification methods the measurement device will be coupled with a robotic joint, which is able to change its impedance, its active as well as its passive part. This setup will be able to set a certain impedance which can then be checked if it can be identified. If successful, the thesis will be completed with measurements on human subjects.


  • Knowledge in system identification and control theory
  • Affinity for system identification and controlling robotic joints

Field of study:

  • Mechatronics, Electrical and Mechanical Engineering


  • MATLAB/ Simulink
  • Knowlegde of C/C++ is helpful, but not essential
  • be able to work independently

URL for this article