Towards understanding human strategies for encoding robot manipulation skills
Abstract/Contents
- Abstract
- Enabling robots to perform tasks in unstructured, human environments has been a long standing vision of roboticists for decades. Some mobile platforms are already operating in these environments, such as autonomous robot vacuums and telepresence robots. In industrial settings, robots perform many manipulation tasks such as welding and handling and transporting objects, however, in these settings, the environment is very structured and the position and orientation of obstacles and work pieces are known with high accuracy. In unstructured human environments, few practical applications of robots performing manipulation tasks exist yet, especially for constrained motion, assembly-type tasks that require forming contact relationships to perform the task. Due to uncertainty in object poses and geometry in unstructured environments, it is essential that the robot complies with object constraints to perform the task and ensure safe operation. This thesis presents work towards developing compliant, human-inspired manipulation skills in task space for robots operating in unstructured environments that are able to accommodate uncertainty and adapt to environmental perturbations. To develop a framework for programming compliant manipulation primitives, we study how humans use contact relationships and compliance while performing insertion-type tasks in a haptic simulation environment. The analysis from the haptic simulation experiments motivates subsequent studies in the physical environment of how humans use contact force information to perform a basic constrained task and combine several primitives for a more complex task. We use insights gleaned from the human experiments to develop robot contact primitives to perform basic tasks and networks of contact primitives to encode more complex skills. The robot skills are encoded at the task level, which inherently enables robot skills to be transferable between robot platforms. Additionally, we use the operational space whole body control architecture which enables handling task and whole body motion coordination while in contact under various constraints within a unified and general form. The contact primitives are developed for tasks where the robot has already firmly grasped an object and uses contact sensing, not relying on any visual information. This thesis also presents work on developing perception primitives using vision-based algorithms to grasp objects and identify visual features to parameterize motion primitives for manipulating simple constrained objects.
Description
| Type of resource | text |
|---|---|
| Form | electronic resource; remote; computer; online resource |
| Extent | 1 online resource. |
| Place | California |
| Place | [Stanford, California] |
| Publisher | [Stanford University] |
| Copyright date | 2018; ©2018 |
| Publication date | 2018; 2018 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Author | Klingbeil, Ellen |
|---|---|
| Degree supervisor | Khatib, Oussama |
| Thesis advisor | Khatib, Oussama |
| Thesis advisor | Rock, Stephen |
| Thesis advisor | Schwager, Mac |
| Degree committee member | Rock, Stephen |
| Degree committee member | Schwager, Mac |
| Associated with | Stanford University, Department of Aeronautics and Astronautics. |
Subjects
| Genre | Theses |
|---|---|
| Genre | Text |
Bibliographic information
| Statement of responsibility | Ellen Klingbeil. |
|---|---|
| Note | Submitted to the Department of Aeronautics and Astronautics. |
| Thesis | Thesis Ph.D. Stanford University 2018. |
| Location | electronic resource |
Access conditions
- Copyright
- © 2018 by Ellen Renae Klingbeil
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...