Humanoid robots represent complex technical systems with a high number of degrees of freedom which must provide human-like capabilities to perform everyday tasks. Mechanical designers must therefore carefully determine the robot’s kinematic structure by determining a joint arrangement which facilitates a suitable workspace for these task applications. Research has provided several approaches to optimize a kinematic structure numerically. However, the proposed objective functions are mostly simple and do not provide the incorporation of task properties except for the reachability of predefined task poses. This thesis derives a method- ology for a task-oriented structure optimization for humanoid robots. It allows to formulate an objective function which statically evaluates the robot’s kinematic dexterities within pre- defined task areas regarding position and orientation properties of the task descriptions. Additionally it is considered, that the robot can execute the task with a single end effector or with both hands in a coupled state. The thesis finally provides a software framework which takes an initial structure topology and finds a good or even optimal set of dimensional param- eters based on the formulated objective funtion, using meta-heuristic numerical optimization methods. The applicability of the optimizers and the methodology in general is shown with three application examples.
Kinematic Structure Optimization for Humanoid Robots
09.04.2020
Aufsatz (Zeitschrift)
Elektronische Ressource
Englisch
Compliant intervertebral mechanism for humanoid backbone: Kinematic modeling and optimization
Online Contents | 2013
|Statistical Learning for Humanoid Robots
British Library Online Contents | 2002
|