This dissertation deals with various aspects of biped locomotion. First, the planar walking problem is treated using a 5 dof’s biped model without actuators at the feet. An input-output linearization is applied, and the walking controller exploits the resulting zero dynamics. The objective of the controller is to produce a stable walking and to regulate the average speed of locomotion. Results are shown through simulation sessions. Some of the ideas used in the planar case are later extended and included in a three-dimensional walking controller based on the linear inverted pendulum approx- imation. The presence of actuated feet is neglected, so that the resulting pendulum model is underactuated. The walking generator and controller exploits the passive dynamics of the approximated system, and generates the walking motion relying on both the double support and the single support phase. The former is used to control the evolution of the inverse pendulum, and it also improves the robot stability. The method is parametric: the reference for the support foot can be arbitrary, so that walking can be achieved with different step lenghts. The above walking generator and controller has been implemented and tested on the humanoid robot UT-Theta, developed at the Nakamura&Yamane Lab. of the University of Tokyo. A strategy for improving the robot stability has also been developed, which works in parallel mode with the walking generator and controller. It is based on data from the foot force sensors, from which the ZMP location can be obtained. This data is used to correct the swing foot reference for the next step, so that external as well as internal disturbances can be rejected during walking. The effectiveness of the overall controller is shown through various walking experiments. In the latest experiments the robot was able to walk for a long time, in a completely autonomous mode.
Mendeley saves you time finding and organizing research
There are no full text links
Choose a citation style from the tabs below