Long horizon planning in the real world
Abstract/Contents
- Abstract
- Enabling robots to perform everyday tasks such as cooking a meal or doing laundry requires giving them the ability to plan future actions over long horizons. Task and Motion Planning (TAMP) seeks to achieve this by combining high-level symbolic reasoning with low-level geometric reasoning to produce long horizon plans that are grounded with actionable motor commands. However, prior TAMP works have largely been limited to simulation due to their computational complexity and brittleness to perception and control noise. This thesis presents the first full-stack TAMP system designed to handle the unpredictability of the real world in real time. We demonstrate the ability to perform TAMP on a real robot with real-time reactive behavior and planning times no longer than several seconds. This system includes a symbolic perception pipeline that enables robust closed-loop task planning, efficient TAMP algorithms that are faster than previous state-of-the-art by an order of magnitude, and a method for integrating TAMP planning with reactive controllers that can adapt to unexpected environmental changes in real time. This thesis also addresses the challenge of scaling TAMP to large domains by proposing an alternative framework that can perform TAMP with a library of learned skills.
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 | 2023; ©2023 |
| Publication date | 2023; 2023 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Author | Migimatsu, Takatoki |
|---|---|
| Degree supervisor | Bohg, Jeannette, 1981- |
| Thesis advisor | Bohg, Jeannette, 1981- |
| Thesis advisor | Khatib, Oussama |
| Thesis advisor | Sadigh, Dorsa |
| Degree committee member | Khatib, Oussama |
| Degree committee member | Sadigh, Dorsa |
| Associated with | Stanford University, School of Engineering |
| Associated with | Stanford University, Computer Science Department |
Subjects
| Genre | Theses |
|---|---|
| Genre | Text |
Bibliographic information
| Statement of responsibility | Takatoki Migimatsu. |
|---|---|
| Note | Submitted to the Computer Science Department. |
| Thesis | Thesis Ph.D. Stanford University 2023. |
| Location | https://purl.stanford.edu/gg626sc8638 |
Access conditions
- Copyright
- © 2023 by Takatoki Migimatsu
- 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...