What Can I Not Do? Towards an Architecture for Reasoning about and Learning Affordances

Mohan Sridharan; Ben Meadows; Rocio Gomez

What Can I Not Do? Towards an Architecture for Reasoning about and Learning Affordances

Mohan Sridharan, Ben Meadows, Rocio Gomez

Computer Science

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

31 Downloads (Pure)

Abstract

This paper describes an architecture for an agent to learn and reason about affordances. In this architecture, Answer Set Prolog, a declarative language, is used to represent and reason with incomplete domain knowledge that includes a representation of affordances as relations defined jointly over objects and actions. Reinforcement learning and decision-tree induction based on this relational representation and observations of action outcomes, are used to interactively and cumulatively (a) acquire knowledge of affordances of specific
objects being operated upon by specific agents; and (b) generalize from these specific learned instances. The capabilities of this architecture are illustrated and evaluated in two simulated domains, a variant of the classic Blocks World domain, and a robot assisting humans in an office environment.

Original language	English
Title of host publication	Proceedings of the Twenty-Seventh International Conference on Automated Planning and Scheduling (ICAPS 2017)
Editors	Laura Barbalescu, Jeremy Frank, Mausam, Stephen F. Smith
Publisher	AAAI Press
Number of pages	9
Publication status	Published - 18 Jun 2017
Event	Twenty-Seventh International Conference on Automated Planning and Scheduling (ICAPS 2017) - Pittsburgh, United States Duration: 18 Jun 2017 → 23 Jun 2017

Conference

Conference	Twenty-Seventh International Conference on Automated Planning and Scheduling (ICAPS 2017)
Country/Territory	United States
City	Pittsburgh
Period	18/06/17 → 23/06/17

Access to Document

What can I not do
Checked for eligibility 11/06/2018
Accepted author manuscript, 160 KBLicence: None: All rights reserved

https://www.aaai.org/Press/Proceedings/icaps17.phpLicence: None: All rights reserved

Cite this

@inproceedings{80d38078328e46489d9faef9c0595fb6,

title = "What Can I Not Do? Towards an Architecture for Reasoning about and Learning Affordances",

abstract = "This paper describes an architecture for an agent to learn and reason about affordances. In this architecture, Answer Set Prolog, a declarative language, is used to represent and reason with incomplete domain knowledge that includes a representation of affordances as relations defined jointly over objects and actions. Reinforcement learning and decision-tree induction based on this relational representation and observations of action outcomes, are used to interactively and cumulatively (a) acquire knowledge of affordances of specific objects being operated upon by specific agents; and (b) generalize from these specific learned instances. The capabilities of this architecture are illustrated and evaluated in two simulated domains, a variant of the classic Blocks World domain, and a robot assisting humans in an office environment.",

author = "Mohan Sridharan and Ben Meadows and Rocio Gomez",

year = "2017",

month = jun,

day = "18",

language = "English",

editor = "Laura Barbalescu and Jeremy Frank and Mausam and Smith, {Stephen F.}",

booktitle = "Proceedings of the Twenty-Seventh International Conference on Automated Planning and Scheduling (ICAPS 2017)",

publisher = "AAAI Press",

note = "Twenty-Seventh International Conference on Automated Planning and Scheduling (ICAPS 2017)<br/> ; Conference date: 18-06-2017 Through 23-06-2017",

}

Sridharan, M, Meadows, B & Gomez, R 2017, What Can I Not Do? Towards an Architecture for Reasoning about and Learning Affordances. in L Barbalescu, J Frank, Mausam & SF Smith (eds), Proceedings of the Twenty-Seventh International Conference on Automated Planning and Scheduling (ICAPS 2017). AAAI Press, Twenty-Seventh International Conference on Automated Planning and Scheduling (ICAPS 2017)
, Pittsburgh, United States, 18/06/17. <https://www.aaai.org/Press/Proceedings/icaps17.php>

What Can I Not Do? Towards an Architecture for Reasoning about and Learning Affordances. / Sridharan, Mohan; Meadows, Ben; Gomez, Rocio.
Proceedings of the Twenty-Seventh International Conference on Automated Planning and Scheduling (ICAPS 2017). ed. / Laura Barbalescu; Jeremy Frank; Mausam; Stephen F. Smith. AAAI Press, 2017.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - What Can I Not Do? Towards an Architecture for Reasoning about and Learning Affordances

AU - Sridharan, Mohan

AU - Meadows, Ben

AU - Gomez, Rocio

PY - 2017/6/18

Y1 - 2017/6/18

N2 - This paper describes an architecture for an agent to learn and reason about affordances. In this architecture, Answer Set Prolog, a declarative language, is used to represent and reason with incomplete domain knowledge that includes a representation of affordances as relations defined jointly over objects and actions. Reinforcement learning and decision-tree induction based on this relational representation and observations of action outcomes, are used to interactively and cumulatively (a) acquire knowledge of affordances of specific objects being operated upon by specific agents; and (b) generalize from these specific learned instances. The capabilities of this architecture are illustrated and evaluated in two simulated domains, a variant of the classic Blocks World domain, and a robot assisting humans in an office environment.

AB - This paper describes an architecture for an agent to learn and reason about affordances. In this architecture, Answer Set Prolog, a declarative language, is used to represent and reason with incomplete domain knowledge that includes a representation of affordances as relations defined jointly over objects and actions. Reinforcement learning and decision-tree induction based on this relational representation and observations of action outcomes, are used to interactively and cumulatively (a) acquire knowledge of affordances of specific objects being operated upon by specific agents; and (b) generalize from these specific learned instances. The capabilities of this architecture are illustrated and evaluated in two simulated domains, a variant of the classic Blocks World domain, and a robot assisting humans in an office environment.

M3 - Conference contribution

BT - Proceedings of the Twenty-Seventh International Conference on Automated Planning and Scheduling (ICAPS 2017)

A2 - Barbalescu, Laura

A2 - Frank, Jeremy

A2 - Mausam, null

A2 - Smith, Stephen F.

PB - AAAI Press

T2 - Twenty-Seventh International Conference on Automated Planning and Scheduling (ICAPS 2017)<br/>

Y2 - 18 June 2017 through 23 June 2017

ER -

What Can I Not Do? Towards an Architecture for Reasoning about and Learning Affordances

Abstract

Conference

Access to Document

Fingerprint

Cite this