Optimization-Based Motion Planning for Dynamic Robotic Manipulation of Physical Systems


Loading...

Author / Producer

Date

2022

Publication Type

Doctoral Thesis

ETH Bibliography

yes

Citations

Altmetric

Data

Abstract

Thanks to their high precision and reliability, manipulator robots have traditionally been employed in industrial production halls to perform mostly simple and repetitive tasks. However, in recent years, roboticists from both research and industry have conducted extensive investigations into enabling robots to carry out chores with increasing complexity. Sharing their vision, we are working towards a future where robots are no longer exclusively used as industrial machines but can assist us in our daily lives at work and at home. To be valuable assets in areas like construction, medical care, or housekeeping, robots need to be able to manipulate complex physical systems in a dynamic and efficient manner. In this thesis, we tackle this long-term goal by investigating how to endow commercially available manipulator robots with the capabilities to dynamically manipulate a variety of real-world systems. To this end, we present a generic, computational framework for robotic motion planning. This framework combines trajectory optimization methods with model-based techniques to generate smooth, direct, and agile robotic motions for different manipulation tasks. Specifically, we investigate skillful manipulation of complex dynamical systems, dynamic grasping maneuvers for static objects, and simultaneous grasp and motion planning for assemblies. We solve the derived trajectory optimization methods with numerical techniques that leverage the differentiability of all models, constraints, and objectives. We evaluate the efficacy of these methodologies in a variety of simulated and real-world experiments. As our framework outputs optimal nominal trajectories directly in the configuration space of the robots, the simulated motions can seamlessly be transferred to their physical counterparts.

Publication status

published

Editor

Contributors

Examiner : Coros, Stelian
Examiner : Billard, Aude
Examiner : Poranne, Roi
Examiner : Siegwart, Roland

Book title

Journal / series

Volume

Pages / Article No.

Publisher

ETH Zurich

Event

Edition / version

Methods

Software

Geographic location

Date collected

Date created

Subject

Trajectory optimization; Robotics; Manipulation; Motion planning; Optimal control; Model-based control; Dynamical systems

Organisational unit

09620 - Coros, Stelian / Coros, Stelian check_circle

Notes

Funding

Related publications and datasets