TU Darmstadt / ULB / TUbiblio

Model-Based Optimization of Whole-Body-Poses for Humanoid Robots in Challenging Environments

Zelch, Christoph (2017):
Model-Based Optimization of Whole-Body-Poses for Humanoid Robots in Challenging Environments.
Darmstadt, Technische Universitaet Darmstadt, Department of Computer Science (SIM), [Master Thesis]

Abstract

The control of complex multi-body systems requires effective and flexible methods to find and reach optimal joint states for a goal posture with some given constraints like static stability. In this thesis, a framework has been developed to optimize general whole-body poses of serial-chain multi-body systems. Specifications of the position and orientation constraints, of the manipulability, of static stability or of self-collision avoidance can be added to a nonlinear optimization problem as constraints or as addend of its objective function. This optimization problem is solved by the commercial solver SNOPT that implements a SQP method. The framework allows an easy integration of new specifications. Further, it is built to be extended by a dynamic robot model, which allows more intricate specifications of postures. A trajectory generator has been implemented that allows a robot to advance safely from a start to a goal posture, since intermediate poses are generated that have to comply with some given constraints. Multiple test scenarios have been created to test the framework using different requirements and considering a majority of the framework’s implemented functionality.

Item Type: Master Thesis
Erschienen: 2017
Creators: Zelch, Christoph
Title: Model-Based Optimization of Whole-Body-Poses for Humanoid Robots in Challenging Environments
Language: English
Abstract:

The control of complex multi-body systems requires effective and flexible methods to find and reach optimal joint states for a goal posture with some given constraints like static stability. In this thesis, a framework has been developed to optimize general whole-body poses of serial-chain multi-body systems. Specifications of the position and orientation constraints, of the manipulability, of static stability or of self-collision avoidance can be added to a nonlinear optimization problem as constraints or as addend of its objective function. This optimization problem is solved by the commercial solver SNOPT that implements a SQP method. The framework allows an easy integration of new specifications. Further, it is built to be extended by a dynamic robot model, which allows more intricate specifications of postures. A trajectory generator has been implemented that allows a robot to advance safely from a start to a goal posture, since intermediate poses are generated that have to comply with some given constraints. Multiple test scenarios have been created to test the framework using different requirements and considering a majority of the framework’s implemented functionality.

Place of Publication: Darmstadt
Uncontrolled Keywords: Humanoid Optimization Control Planning
Divisions: 20 Department of Computer Science
20 Department of Computer Science > Simulation, Systems Optimization and Robotics Group
Date Deposited: 06 Jun 2019 06:09
Export:
Suche nach Titel in: TUfind oder in Google

Optionen (nur für Redakteure)

View Item View Item