"This master thesis work presents the theoretical fundamentals, development strategies and implementation results of a walking pattern generator for humanoid robots, which is suitable for visually-guided tasks. In first place, a general introduction about the state-of-the-art related to research on humanoid locomotion solutions is introduced, as well as a general overview about humanoid robots, making emphasis on the main characteristics that identify this type of robots from the rest of the existing platforms. Furthermore, a method to construct the kinematic model for robots based on a tree-like structure is proposed. Subsequently, several algorithms and relations necessary to compute the model forward kinematics and to calculate the different versions of Jacobian matrices are introduced. Moreover, this work presents and explains two powerful techniques to compute the model inverse kinematics. These methods take advantage of the inherent redundancy of humanoid robots to simultaneously solve a determined amount of kinematic tasks, based on a prioritized scheme. Later, all important fundamentals related to the humanoid dynamics are introduced. Particularly, the concept of the Zero-Moment Point (ZMP), which is a key parameter on humanoid locomotion, is presented in this thesis work and besides, some methods to determine the position of this parameter are discussed as well. This thesis also presents some schemes to model the process of biped locomotion, focusing on a method to determine the optimal trajectory of the robot Center of Mass (CoM), based on the preview control of the reference ZMP position. Besides, all fields that are studied in this thesis are taken together into a global architecture for walking pattern generation, consisting on two stages of control, each of them containing the processes of linear-quadratic regulation of the reference ZMP, Prioritized Inverse Kinematics (PIK) and approximation of the real position of the ZMP based on the robot movements. A genuine verification of the functionality of the designed walking pattern generator is exposed at the end of this thesis work, by presenting the results of the performed simulations on a virtual humanoid model, as well as the experimentations on a physical humanoid robot."
Humanoid locomotion planning for visually-guided task
2011-09-01
Theses
Electronic Resource
English
DDC: | 629 |
Balance control and locomotion planning for humanoid robots using nonlinear centroidal models
BASE | 2020
|