Directly to content
  1. Publishing |
  2. Search |
  3. Browse |
  4. Recent items rss |
  5. Open Access |
  6. Jur. Issues |
  7. DeutschClear Cookie - decide language by browser settings

Using Model-based Optimal Control for Conceptional Motion Generation for the Humannoid Robot HRP-2 14 and Design Investigations for Exo-Skeletons

Koch, Kai Henning

[img]
Preview
PDF, English
Download (39MB) | Terms of use

Citation of documents: Please do not cite the URL that is displayed in your browser location input, instead use the DOI, URN or the persistent URL below, as we can guarantee their long-time accessibility.

Abstract

The research field of bipedal locomotion has been active since a few decades now. At one hand, the legged locomotion principle comprises highly flexible and robust mobility for technical applications. At the other hand, a thorough technical understanding of bipedalism supports efforts of clinicians and engineers to help people, suffering from reduced locomotion capabilities caused by fatal incidents.

Since the technology enabled the construction of numerous robotic devices, among them: various humanoids, researchers started to investigate bipedalism by abstraction and adoption for technical applications. Findings from humanoid robotics are further exploited for the construction of devices for human performance augmentation and mobility support or gait rehabilitation, among them: orthosis and exo-skeletons.

Although this research continuously progresses, the motion capacities of humanoid robots still lack far behind those of humans in terms of forward velocity, robustness and appearance of the overall motion. Generally, it is claimed that the difference of performance between humans and robotics is not only due to the limiting characteristics of the employed technology, e.g. constructive lack of specific determinants of gait for bipedalism or dynamic limits of the actuation system, but as well to the adopted methods for motion generation and control.

For humanoid robotics, methods for motion generation are classified into optimization-based methods and those that employ heuristics, that are mostly distinguished based on the problem complexity (computation time) and the resulting dynamic error between the generated motion and the dynamics of the real robot. The implementation of the dynamic motion on the robotic platform is usually comprised with an on-line stabilizing control system. This control system must then identify and resolve instantaneously the dynamic error to maintain a continuously stable operation of the device. A large dynamic error and breach of the dynamic limits of the actuation system can quickly lead to a fatal destabilization of the device.

This work proposes a contribution to the model computation and the strategy of the problem formulation of direct multiple-shooting based optimal control (Bock et. al.) for dynamically stable optimization-based motion generation. The computation of the whole-body dynamic model inside the optimization relies either on forward or inverse dynamics approach. As the inverse dynamics approach has frequently been perceived as less resource intensive than the forward dynamics approach, a new generic algorithm for insufficiently constrained, under-actuated dynamic systems has been developed and thoroughly tested to comply with all numerical restrictions of the enveloping optimization algorithm.

Based on this contribution, various optimal control problems for the humanoid platform HRP-2 14 have been formulated to assess the influence of different biologically inspired optimization criteria on the final motion characteristics of walking motions. From thorough bibliographic researches a dynamically more accurate model was comprised, by taking into account the impact absorbing element in the ankle joint complex. Based on the experiences of the previous study, a problem formulation for the limiting case of, dynamically overstepping an obstacle of 20cm x 11cm (height x width) with only two steps, while maintaining its stable operation was accomplished. This is a new record for this platform.

In a further part, this work proposes an iterative comprehensive model-based optimal control approach for the conception of a lower limb exo-skeleton that respects the integrated nature of such a mechatronic device. In this contribution, a human effectively wearing such a lower limb exo-skeleton is modeled. The approach then substantiates all system components in an iterative procedure, based on the complete system model, effectively resolving all complex inter-dependencies between the different components of the system. The study in this work is conducted on an important benchmark motion, walking, of a healthy human being. From this study the limiting characteristics of the system are determined and substantial propositions to the realization of various system components are formulated.

Item Type: Dissertation
Supervisor: Mombaur, Prof. Dr. Katja
Date of thesis defense: 3 June 2015
Date Deposited: 27 Jul 2015 08:18
Date: 2015
Faculties / Institutes: The Faculty of Mathematics and Computer Science > Department of Applied Mathematics
About | FAQ | Contact | Imprint |
OA-LogoDINI certificate 2013Logo der Open-Archives-Initiative