Towards energy-efficient limit-cycle walking in biped service robots: design analysis, modeling and experimental study of biped robot actuated by linear motors

Abstract

Researchers have been studying biped robots for many years, and, while many advances in the field have been accomplished, there still remain the challenge to transfer the existing solutions into real applications. The main issues are related to mobility and autonomy. In mobility, biped robots have evolved greatly, nevertheless they are still far from what a human can do in the work-site. Similarly, autonomy of biped platforms has been tackled on several different grounds, but its core problem still remains, and it is associated to energy issues. Because of these energy issues, lately the main attention has been redirected to the long term autonomy of the biped robotics platforms. For that, much effort has been made to develop new more energy-efficient biped robots.

The GIMBiped project in Aalto University was established to tackle the previous issues in energy efficiency and mobility, through the study and implementation of dynamic and energy-efficient bipedal robotic waking. This thesis falls into the first studies needed to achieve the previous goal using the GIMBiped testbed, starting with a detailed analysis of the nonlinear dynamics of the target system, using a modeling and simulation tools. This work also presents an assessment of different control techniques based on Limit Cycle walking, carried out on a two-dimensional kneed bipedal simulator.

Furthermore, a numerical continuation analysis of the mechanical parameters of the first GIMBiped prototype was performed, using the same approximated planar kneed biped model. This study is done to analyze the effect that such variations in the mechanical design parameters produce in the stability and energy-efficiency of the system.Finally, experiments were performed in the GIMBiped testbed. These experiments show the results of a hybrid control technique proposed by the author, which combines traditional ZMP-based walking approach with a Limit Cycle trajectory-following control. Furthermore the results of a pure ZMP-based type of control are also presented.