Observer-based Iterative Learning Control to Improve Path Accuracy of Industrial Robots

Project Focus

  • Modeling of nonlinear effects within the drive train of robots
  • Design of precise model-based state observers
  • Iterative learning control to improve path accuracy at the end-effector


For machining tasks with robots, path accuracy is an essential performance criterion. In many applications, accuracies in the range of tenths of a millimeter and below are required. Examples include the seam sealing of car bodies with robots and laser welding. In these applications, it is not only high static accuracy that is important, but also overall accuracy, which includes tracking errors of the control system, as well as dynamic effects of the mechanical structure. Figure 1 shows the tracking error in laser marking.

A major cause for the deviation from the ideal behavior shown above lies in the drive trains. Here, the gearboxes have the most significant influence. Compared to the ideal behavior, gearboxes exhibit the following behavior:

  • Non-uniform friction: due to the varying engagement of the gear teeth, the friction is position-dependent and load-dependent.
  • Non-constant gear ratio: due to imperfect tooth flanks, small deviations occur in the endeffector position. For robots 1-3 meters arm length, this can give rise for TCP (Tool Center Point) errors in the range of a few tenths of a millimeter.
  • Finite and variable stiffness: due to the compliance of the gearbox, vibrations are excited in interaction with friction and imperfect transmission behavior. If the frequency of the variable stiffness coincides with the resonant frequency of the robot axes, this can lead to very high dynamic errors.

The gear effects are complex, especially in their dynamic interaction. Measurement methods usually require high-priced measuring equipment. Each robot for which compensation is to be used must be measured individually.

The aim of the project is to develop suitable robot models and state observers that incorporate the gear effects described above. Based on this, new control strategies are to be designed with which the growing accuracy requirements can be achieved. The main goal is to combine mechatronic, physical models in an optimal way with the processing of unstructured data in order to achieve the highest possible efficiency and ease of use. Furthermore, to address the availability of highly accurate measurement systems in practice, inexpensive available sensors will be used to guarantee a broad applicability of the developed methods. The measurement data obtained in this way will be used in the course of an iterative learning control to ultimately achieve the desired accuracy, not only in positioning tasks, but also while traversing paths.

Selected Publications

  • B. Bischof, T. Glück, M. Böck, and A. Kugi, Path Following Control for Elastic Joint Robots, in Proceedings of the 20th IFAC World Congress, Toulouse, France, 2017, p. 4806–4811.
    author = {Bischof, B. and Gl\"uck, T. and B\"ock, M. and Kugi, A.},
    title = {Path Following Control for Elastic Joint Robots},
    booktitle = {Proceedings of the 20th IFAC World Congress},
    year = {2017},
    volume = {50},
    number = {1},
    month = {7},
    pages = {4806--4811},
    doi = {10.1016/j.ifacol.2017.08.965},
    address = {Toulouse, France},
    issn = {2405-8963},
  • B. Bischof, Path and Surface Following Control for Industrial Robotic Applications, A. Kugi and K. Schlacher, Eds., Aachen: Shaker Verlag, 2020, vol. 47.
    author = {Bischof, Bernhard},
    title = {{Path and Surface Following Control for Industrial Robotic Applications}},
    year = {2020},
    editor = {A. Kugi and K. Schlacher},
    volume = {47},
    series = {Modellierung und Regelung komplexer dynamischer Systeme},
    publisher = {Shaker Verlag},
    isbn = {978-3-8440-7200-6},
    address = {Aachen},
    organization = {Institute f{\"u}r Automatisierungs- und Regelungstechnik (TU Wien) und Regelungstechnik und Prozessautomatisierung (JKU Linz)},


Univ.-Prof. Dr.techn. Andreas Kugi


2018 - 2021