Twickel, Arndt von (2011). Embodied Modular Neural Control of Walking in Stick Insects -- From Biological Models to Evolutionary Robotics. PhD thesis, Universität zu Köln.
|
PDF (Complete Thesis)
von_Twickel,_Arndt,_Dissertation,_Maerz_2012.pdf - Accepted Version Download (16MB) |
Abstract
Walking requires the flexible co-ordination of many degrees of freedom. Biomechanical and nervous systems have to interact with the environment to fulfill this challenging task. Due to the complexity of interactions, important questions, especially regarding the neural control of walking, remain unanswered. Biological models are promising tools to integrate available data and to generate testable hypotheses, but they face the problem of a huge parameter space. Therefore, this thesis combines (neuro-)biological models of stick insect walking with the complementary approach of evolutionary robotics. On the one hand, extremely simple single-leg controllers are developed by artificial evolution that exploit properties of the body and the environment. General principles of sensori-motor couplings are discovered and the importance of hysteresis in neural walking control is demonstrated. On the other hand, neuro-biological models of single-leg stepping control in stick insects are thoroughly tested under multiple perturbing conditions in stick insect as well as in robotic models. Their robustness and behavioral adaptability is demonstrated, suggesting that they are suitable to work as modules of hexapod controllers. Subsequently, the impact of muscles on a set of neural control structures and behavior in robotic as well as in stick insect models is investigated using a muscle model derived from the stick insect extensor muscle. Employing evolutionary parameter optimization of neural controllers it is shown that muscle properties reduce the requirement for neural intra-joint feedback and that behaviors become more robust under perturbing conditions. Furthermore, it is demonstrated that diverse and simple neural mechanisms may be used to compensate for the slowness of muscles observed in stick insects. Finally, a modular approach to hexapod controller development is taken, integrating the evolved and neuro-biologically inspired single-leg controller structures with additional sensori-motor couplings. Coupling structures are either derived from behavioral based biological data or by artificial evolution. Robust hexapod walking is demonstrated for robotic and stick insect models with and without muscle properties. The results show that the integrative approach presented in this thesis allows to develop robust control mechanisms for walking machines and to provide testable hypotheses about the neural basis of inter-leg coupling mechanisms.
Item Type: | Thesis (PhD thesis) | ||||||||
Translated abstract: |
|
||||||||
Creators: |
|
||||||||
URN: | urn:nbn:de:hbz:38-45865 | ||||||||
Date: | 2 March 2011 | ||||||||
Language: | English | ||||||||
Faculty: | Faculty of Mathematics and Natural Sciences | ||||||||
Divisions: | Faculty of Mathematics and Natural Sciences > Department of Biology > Zoologisches Institut | ||||||||
Subjects: | Data processing Computer science Natural sciences and mathematics |
||||||||
Uncontrolled Keywords: |
|
||||||||
Date of oral exam: | 6 April 2011 | ||||||||
Referee: |
|
||||||||
Refereed: | Yes | ||||||||
URI: | http://kups.ub.uni-koeln.de/id/eprint/4586 |
Downloads
Downloads per month over past year
Export
Actions (login required)
View Item |