Combining dynamic abstractions in large MDPs
One of the reasons that it is difficult to plan and act in real-worlddomains is that they are very large. Existing research generallydeals with the large domain size using a static representation andexploiting a single type of domain structure. In this paper, wecreate a framework that encapsulates...
| Main Authors: | , |
|---|---|
| Language: | en_US |
| Published: |
2005
|
| Subjects: | |
| Online Access: | http://hdl.handle.net/1721.1/30496 |
| _version_ | 1811079087361359872 |
|---|---|
| author | Steinkraus, Kurt Kaelbling, Leslie Pack |
| author_facet | Steinkraus, Kurt Kaelbling, Leslie Pack |
| author_sort | Steinkraus, Kurt |
| collection | MIT |
| description | One of the reasons that it is difficult to plan and act in real-worlddomains is that they are very large. Existing research generallydeals with the large domain size using a static representation andexploiting a single type of domain structure. In this paper, wecreate a framework that encapsulates existing and new abstraction andapproximation methods into modules, and combines arbitrary modulesinto a system that allows for dynamic representation changes. We showthat the dynamic changes of representation allow our framework tosolve larger and more interesting domains than were previouslypossible, and while there are no optimality guarantees, suitablemodule choices gain tractability at little cost to optimality. |
| first_indexed | 2024-09-23T11:09:32Z |
| id | mit-1721.1/30496 |
| institution | Massachusetts Institute of Technology |
| language | en_US |
| last_indexed | 2024-09-23T11:09:32Z |
| publishDate | 2005 |
| record_format | dspace |
| spelling | mit-1721.1/304962019-04-12T08:37:51Z Combining dynamic abstractions in large MDPs Steinkraus, Kurt Kaelbling, Leslie Pack AI One of the reasons that it is difficult to plan and act in real-worlddomains is that they are very large. Existing research generallydeals with the large domain size using a static representation andexploiting a single type of domain structure. In this paper, wecreate a framework that encapsulates existing and new abstraction andapproximation methods into modules, and combines arbitrary modulesinto a system that allows for dynamic representation changes. We showthat the dynamic changes of representation allow our framework tosolve larger and more interesting domains than were previouslypossible, and while there are no optimality guarantees, suitablemodule choices gain tractability at little cost to optimality. 2005-12-22T01:41:41Z 2005-12-22T01:41:41Z 2004-10-21 MIT-CSAIL-TR-2004-065 AIM-2004-023 http://hdl.handle.net/1721.1/30496 en_US Massachusetts Institute of Technology Computer Science and Artificial Intelligence Laboratory 12 p. 9975204 bytes 424481 bytes application/postscript application/pdf application/postscript application/pdf |
| spellingShingle | AI Steinkraus, Kurt Kaelbling, Leslie Pack Combining dynamic abstractions in large MDPs |
| title | Combining dynamic abstractions in large MDPs |
| title_full | Combining dynamic abstractions in large MDPs |
| title_fullStr | Combining dynamic abstractions in large MDPs |
| title_full_unstemmed | Combining dynamic abstractions in large MDPs |
| title_short | Combining dynamic abstractions in large MDPs |
| title_sort | combining dynamic abstractions in large mdps |
| topic | AI |
| url | http://hdl.handle.net/1721.1/30496 |
| work_keys_str_mv | AT steinkrauskurt combiningdynamicabstractionsinlargemdps AT kaelblinglesliepack combiningdynamicabstractionsinlargemdps |