Difference between revisions of "Main Page"

From Tamp Benchmarking
Jump to: navigation, search
m (Protected "Main Page" ([Edit=Allow only administrators] (indefinite) [Move=Allow only administrators] (indefinite)))
(About this Wiki)
 
(88 intermediate revisions by 3 users not shown)
Line 1: Line 1:
 
== About this Wiki ==
 
== About this Wiki ==
  
Following the discussions at the TAMP workshop in RSS2016, a document was sketched for initiating thinking together about benchmarking for CTAMP planners. This wiki was created based on this document : we need more space to conduct proper discussions, and we may also have to share downloadable content later on. For now, I simply copy-pasted the contents of the previous document into different pages. Participants are invited to comment, correct, discuss, and bring their ideas.
+
Following the RSS2016 TAMP workshop, a discussion group (see ''Contributors'' below) was created to come up with a methodology and a set of benchmark problems for TAMP. Our discussions and the actual benchmarks can be found in the '''TAMP FO-DA''' section in the ''Contents'' menu. More details can be found in [https://ieeexplore.ieee.org/document/8411475/ this paper]. Our approach was to start with a set of problems which can be handled by the largest number of existing systems, by identifying the least common denominator for requirements and problem specification, at logical and geometric levels.
  
== Aims ==
+
However, since TAMP covers a wide range of problems, it was not possible to come up with a set of benchmarks for '''all''' types of TAMP problems. Therefore the ''Contents'' of the wiki is organized according to '''subclasses''' of TAMP. If you find the assumptions and requirements for ''TAMP FO-DA'' incompatible with the type of problems you are interested in, please contribute by creating your own benchmark set in the ''Contents'' section. Candidate subclasses are suggested:
  
My perception of the current state of the field is that different groups are still exploring “variations” of the CTAMP problem, based on their own research interests, tools and experience. For this reason, defining a common language to specify CTAMP problems is, in my opinion, premature at this stage (or rather: it has to be done sooner or later, but this document does not address this issue).
+
== Contents ==
 +
* [[TAMP_FO-DA|'''TAMP FO-DA''': Fullly Observable, Deterministic Actions]]
 +
* [[TAMP_PO-NDA|'''TAMP PO-NDA''': Partially Observable, Non Deterministic Actions]]  - '''TODO'''
 +
* [[TAMP_MP|'''TAMP MP''': Manipulation Planning (no symbolic planning)]]  - '''TODO'''
 +
* [[TAMP_PH|'''TAMP Physics''': Handling frictions (push-pull), gravity (building stable structures), dynamics]]  - '''TODO'''
 +
* Create your own...
  
My proposal is to '''start with a set of simple benchmark problems, which can be handled by the largest number of existing systems.''' This requires to identify the least common denominator, in terms of requirements and problem specification, at logical and geometric levels.
+
== How to use this wiki ==
  
These requirements will be extended later on, but in order to reach a “critical mass” so that enough people start using these benchmarks and disseminate them through their publications, we need to begin with something simple and accessible.
+
* This page is powered by MediaWiki, you can find instructions for editing pages [https://www.mediawiki.org/wiki/Help:Editing_pages here].
 
+
* Do not modify a page for which you have not taken part in discussions.
== Contents ==
+
* Do not create a benchmark set on your own: this has to be a collaborative work involving other researchers.
 +
* For discussions, use the wiki "Discussion" tab, or set up a discussion thread, by creating a page and adding the following code at the top: <nowiki>{{#useliquidthreads:1}}</nowiki> (see for example [[Comments_from_Neil|this page]])
 +
* Note that there is an Atom feed for each discussion page (left menu) you can subscribe to for getting updated with new replies. You can adjust your preferences (top right menu) for getting notified by email.
  
* Requirements (based on the article by Neil<ref>Neil T. Dantam, Zachary K. Kingston, Swarat Chaudhuri, and Lydia E. Kavraki, ''Incremental Task and Motion Planning: A Constraint-Based Approach'', (TAMP Workshop, RSS2016).</ref>)
+
== Contact ==
** [[Task Planning Requirements]]
+
If you want to contribute to this wiki, contact me at fabien[dot]lagriffoul[at]oru[dot]se and I'll create an account for you.
** [[Motion Planning Requirements]]
 
** [[Task-Motion Interface Requirements]]
 
* [[General Assumptions]]
 
* [[Metrics]]
 
* [[Problems]]
 
* [[Practical Issues]]
 
  
== Discussions ==
+
== Contributors ==
 +
* '''TAMP FO-DA''':
 +
''Fabien Lagriffoul'' (Orebro University, Sweden.) <br />
 +
''Neil T. Dantam'' (Colorado School of Mines, USA) <br />
 +
''Caelan Garrett'' (Massachusetts Institute of Technology, USA) <br />
 +
''Ali Akbari'' (Tech. U. of Catalonia, Spain) <br />
 +
''Siddharth Srivastava'' (Arizona State University, USA) <br />
  
* [[Comments from Neil]]
+
== Support ==
* [[Comments from Calean]]
+
This initiative is supported by the Swedish Knowledge Foundation (KKS) project [https://www.oru.se/english/research/research-projects/rp/?rdb=p1152 Semantic Robots]. It is also supported by NSF grants IIS 1317849 and CCF-1514372 to the Kavraki Lab at Rice University, USA.
  
== References ==
+
== RSS Workshops ==
 +
* 2018: [http://www.neil.dantam.name/2018/rss-tmp-workshop/ Exhibition and Benchmarking of Task and Motion Planners RSS 2018]
 +
* 2017: [http://www.kavrakilab.org/2017-rss-workshop/ Task and Motion Planning RSS 2017]
 +
* 2016: [http://www.kavrakilab.org/2016-rss-workshop/ Task and Motion Planning RSS 2016]

Latest revision as of 09:44, 18 July 2018

About this Wiki

Following the RSS2016 TAMP workshop, a discussion group (see Contributors below) was created to come up with a methodology and a set of benchmark problems for TAMP. Our discussions and the actual benchmarks can be found in the TAMP FO-DA section in the Contents menu. More details can be found in this paper. Our approach was to start with a set of problems which can be handled by the largest number of existing systems, by identifying the least common denominator for requirements and problem specification, at logical and geometric levels.

However, since TAMP covers a wide range of problems, it was not possible to come up with a set of benchmarks for all types of TAMP problems. Therefore the Contents of the wiki is organized according to subclasses of TAMP. If you find the assumptions and requirements for TAMP FO-DA incompatible with the type of problems you are interested in, please contribute by creating your own benchmark set in the Contents section. Candidate subclasses are suggested:

Contents

How to use this wiki

  • This page is powered by MediaWiki, you can find instructions for editing pages here.
  • Do not modify a page for which you have not taken part in discussions.
  • Do not create a benchmark set on your own: this has to be a collaborative work involving other researchers.
  • For discussions, use the wiki "Discussion" tab, or set up a discussion thread, by creating a page and adding the following code at the top: {{#useliquidthreads:1}} (see for example this page)
  • Note that there is an Atom feed for each discussion page (left menu) you can subscribe to for getting updated with new replies. You can adjust your preferences (top right menu) for getting notified by email.

Contact

If you want to contribute to this wiki, contact me at fabien[dot]lagriffoul[at]oru[dot]se and I'll create an account for you.

Contributors

  • TAMP FO-DA:

Fabien Lagriffoul (Orebro University, Sweden.)
Neil T. Dantam (Colorado School of Mines, USA)
Caelan Garrett (Massachusetts Institute of Technology, USA)
Ali Akbari (Tech. U. of Catalonia, Spain)
Siddharth Srivastava (Arizona State University, USA)

Support

This initiative is supported by the Swedish Knowledge Foundation (KKS) project Semantic Robots. It is also supported by NSF grants IIS 1317849 and CCF-1514372 to the Kavraki Lab at Rice University, USA.

RSS Workshops