Bioinspired template-based control of legged locomotion

cient and robust locomotion is a crucial condition for the more extensive use of legged robots in real world applications. In that respect, robots can learn from animals, if the principles underlying locomotion in biological legged systems can be transferred to their artificial counterparts. However, legged locomotion in biological systems is a complex and not fully understood problem. A great progress to simplify understanding locomotion dynamics and control was made by introducing simple models, coined ``templates'', able to represent the overall dynamics of animal (including human) gaits. One of the most recognized models is the spring-loaded inverted pendulum (SLIP) which consists of a point mass atop a massless spring. This model provides a good description of human gaits, such as walking, hopping and running. Despite its high level of abstraction, it supported and inspired the development of successful legged robots and was used as explicit targets for control, over the years. Inspired from template models explaining biological locomotory systems and Raibert's pioneering legged robots, locomotion can be realized by basic subfunctions: (i) stance leg function, (ii) leg swinging and (iii) balancing. Combinations of these three subfunctions can generate different gaits with diverse properties. Using the template models, we investigate how locomotor subfunctions contribute to stabilize different gaits (hopping, running and walking) in different conditions (e.g., speeds). We show that such basic analysis on human locomotion using conceptual models can result in developing new methods in design and control of legged systems like humanoid robots and assistive devices (exoskeletons, orthoses and prostheses). This thesis comprises research in different disciplines: biomechanics, robotics and control. These disciplines are required to do human experiments and data analysis, modeling of locomotory systems, and implementation on robots and an exoskeleton. We benefited from facilities and experiments performed in the Lauflabor locomotion laboratory. Modeling includes two categories: conceptual (template-based, e.g. SLIP) models and detailed models (with segmented legs, masses/inertias). Using the BioBiped series of robots (and the detailed BioBiped MBS models; MBS stands for Multi-Body-System), we have implemented newly-developed design and control methods related to the concept of locomotor subfunctions on either MBS models or on the robot directly. In addition, with involvement in BALANCE project (\url{http://balance-fp7.eu/}), we implemented balance-related control approaches on an exoskeleton to demonstrate their performance in human walking. The outcomes of this research includes developing new conceptual models of legged locomotion, analysis of human locomotion based on the newly developed models following the locomotor subfunction trilogy, developing methods to benefit from the models in design and control of robots and exoskeletons. The main contribution of this work is providing a novel approach for modular control of legged locomotion. With this approach we can identify the relation between different locomotor subfunctions e.g., between balance and stance (using stance force for tuning balance control) or balance and swing (two joint hip muscles can support the swing leg control relating it to the upper body posture) and implement the concept of modular control based on locomotor subfunctions with a limited exchange of sensory information on several hardware platforms (legged robots, exoskeleton).