Machine Learning for Efficient Sampling-Based Algorithms in Robust Multi-Agent Planning Under Uncertainty
Robust multi-agent planning algorithms have been developed to assign tasks to cooperative teams of robots operating under various uncertainties. Often, it is difficult to evaluate the robustness of potential task assignments analytically, so sampling-based approximations are used instead. In many ap...
| Main Authors: | , |
|---|---|
| Other Authors: | |
| Format: | Article |
| Published: |
American Institute of Aeronautics and Astronautics (AIAA)
2018
|
| Online Access: | http://hdl.handle.net/1721.1/114295 https://orcid.org/0000-0002-0464-4108 https://orcid.org/0000-0001-8576-1930 |
| _version_ | 1826200303624519680 |
|---|---|
| author | Quindlen, John Francis How, Jonathan P |
| author2 | Massachusetts Institute of Technology. Department of Aeronautics and Astronautics |
| author_facet | Massachusetts Institute of Technology. Department of Aeronautics and Astronautics Quindlen, John Francis How, Jonathan P |
| author_sort | Quindlen, John Francis |
| collection | MIT |
| description | Robust multi-agent planning algorithms have been developed to assign tasks to cooperative teams of robots operating under various uncertainties. Often, it is difficult to evaluate the robustness of potential task assignments analytically, so sampling-based approximations are used instead. In many applications, not only are sampling-based approximations the only solution, but these samples are computationally-burdensome to obtain. This paper presents a machine learning procedure for sampling-based approximations that actively selects samples in order to maximize the accuracy of the approximation with a limited number of samples. Gaussian process regression models are constructed from a small set of training samples and used to approximate the robustness evaluation. Active learning is then used to iteratively select samples that most improve this evaluation. Three example problems demonstrate that the new procedure achieves a similar level of accuracy as the existing sample-inefficient procedures, but with a significant reduction in the number of samples. |
| first_indexed | 2024-09-23T11:34:25Z |
| format | Article |
| id | mit-1721.1/114295 |
| institution | Massachusetts Institute of Technology |
| last_indexed | 2024-09-23T11:34:25Z |
| publishDate | 2018 |
| publisher | American Institute of Aeronautics and Astronautics (AIAA) |
| record_format | dspace |
| spelling | mit-1721.1/1142952022-10-01T04:32:31Z Machine Learning for Efficient Sampling-Based Algorithms in Robust Multi-Agent Planning Under Uncertainty Quindlen, John Francis How, Jonathan P Massachusetts Institute of Technology. Department of Aeronautics and Astronautics Quindlen, John Francis How, Jonathan P Robust multi-agent planning algorithms have been developed to assign tasks to cooperative teams of robots operating under various uncertainties. Often, it is difficult to evaluate the robustness of potential task assignments analytically, so sampling-based approximations are used instead. In many applications, not only are sampling-based approximations the only solution, but these samples are computationally-burdensome to obtain. This paper presents a machine learning procedure for sampling-based approximations that actively selects samples in order to maximize the accuracy of the approximation with a limited number of samples. Gaussian process regression models are constructed from a small set of training samples and used to approximate the robustness evaluation. Active learning is then used to iteratively select samples that most improve this evaluation. Three example problems demonstrate that the new procedure achieves a similar level of accuracy as the existing sample-inefficient procedures, but with a significant reduction in the number of samples. 2018-03-26T19:10:47Z 2018-03-26T19:10:47Z 2017-01 2018-03-21T16:52:43Z Article http://purl.org/eprint/type/JournalArticle 978-1-62410-450-3 http://hdl.handle.net/1721.1/114295 Quindlen, John F., and Jonathan P. How. “Machine Learning for Efficient Sampling-Based Algorithms in Robust Multi-Agent Planning Under Uncertainty.” AIAA Guidance, Navigation, and Control Conference (January 2017) © 2017 American Institute of Aeronautics and Astronautics Inc https://orcid.org/0000-0002-0464-4108 https://orcid.org/0000-0001-8576-1930 http://dx.doi.org/10.2514/6.2017-1921 AIAA Guidance, Navigation, and Control Conference Creative Commons Attribution-Noncommercial-Share Alike http://creativecommons.org/licenses/by-nc-sa/4.0/ application/pdf American Institute of Aeronautics and Astronautics (AIAA) MIT Web Domain |
| spellingShingle | Quindlen, John Francis How, Jonathan P Machine Learning for Efficient Sampling-Based Algorithms in Robust Multi-Agent Planning Under Uncertainty |
| title | Machine Learning for Efficient Sampling-Based Algorithms in Robust Multi-Agent Planning Under Uncertainty |
| title_full | Machine Learning for Efficient Sampling-Based Algorithms in Robust Multi-Agent Planning Under Uncertainty |
| title_fullStr | Machine Learning for Efficient Sampling-Based Algorithms in Robust Multi-Agent Planning Under Uncertainty |
| title_full_unstemmed | Machine Learning for Efficient Sampling-Based Algorithms in Robust Multi-Agent Planning Under Uncertainty |
| title_short | Machine Learning for Efficient Sampling-Based Algorithms in Robust Multi-Agent Planning Under Uncertainty |
| title_sort | machine learning for efficient sampling based algorithms in robust multi agent planning under uncertainty |
| url | http://hdl.handle.net/1721.1/114295 https://orcid.org/0000-0002-0464-4108 https://orcid.org/0000-0001-8576-1930 |
| work_keys_str_mv | AT quindlenjohnfrancis machinelearningforefficientsamplingbasedalgorithmsinrobustmultiagentplanningunderuncertainty AT howjonathanp machinelearningforefficientsamplingbasedalgorithmsinrobustmultiagentplanningunderuncertainty |