Can I Do That? Discovering Domain Axioms Using Declarative Programming and Relational Reinforcement Learning

Mohan Sridharan; Prashanth Devarakonda; Rashmica Gupta

doi:10.1007/978-3-319-46840-2_3

Can I Do That? Discovering Domain Axioms Using Declarative Programming and Relational Reinforcement Learning

Mohan Sridharan, Prashanth Devarakonda, Rashmica Gupta

Computer Science

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

Abstract

Robots deployed to assist humans in complex, dynamic domains need the ability to represent, reason with, and learn from, different descriptions of incomplete domain knowledge and uncertainty. This paper presents an architecture that integrates declarative programming and relational reinforcement learning to support cumulative and interactive discovery of previously unknown axioms governing domain dynamics. Specifically, Answer Set Prolog (ASP), a declarative programming paradigm, is used to represent and reason with incomplete commonsense domain knowledge. For any given goal, unexplained failure of plans created by inference in the ASP program is taken to indicate the existence
of unknown domain axioms. The task of learning these axioms is formulated
as a Reinforcement Learning problem, and decision-tree regression with a relational representation is used to generalize from specific axioms identified over time. The new axioms are added to the ASP-based representation
for subsequent inference. We demonstrate and evaluate the capabilities of our architecture in two simulated domains: Blocks World and Simple Mario.

Original language	English
Title of host publication	Autonomous Agents and Multiagent Systems
Subtitle of host publication	AAMAS 2016 Workshops, Visionary Papers, Singapore, May 9-10, 2016, Revised Selected Papers
Editors	Nardine Osman, Carles Sierra
Publisher	Springer
Chapter	3
Pages	34-49
ISBN (Electronic)	9783319468402
ISBN (Print)	9783319468396
DOIs	https://doi.org/10.1007/978-3-319-46840-2_3
Publication status	Published - 24 Sept 2016
Event	AAMAS 2016 Workshops, - , Singapore Duration: 9 May 2016 → 10 May 2016

Publication series

Name	Lecture Notes in Computer Science
Publisher	Springer
Volume	10003
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	AAMAS 2016 Workshops,
Country/Territory	Singapore
Period	9/05/16 → 10/05/16

Access to Document

10.1007/978-3-319-46840-2_3Licence: None: All rights reserved

http://link.springer.com/10.1007/978-3-319-46840-2_3Licence: None: All rights reserved

Cite this

Sridharan, M., Devarakonda, P., & Gupta, R. (2016). Can I Do That? Discovering Domain Axioms Using Declarative Programming and Relational Reinforcement Learning. In N. Osman, & C. Sierra (Eds.), Autonomous Agents and Multiagent Systems: AAMAS 2016 Workshops, Visionary Papers, Singapore, May 9-10, 2016, Revised Selected Papers (pp. 34-49). (Lecture Notes in Computer Science; Vol. 10003). Springer. https://doi.org/10.1007/978-3-319-46840-2_3

Sridharan, Mohan ; Devarakonda, Prashanth ; Gupta, Rashmica. / Can I Do That? Discovering Domain Axioms Using Declarative Programming and Relational Reinforcement Learning. Autonomous Agents and Multiagent Systems: AAMAS 2016 Workshops, Visionary Papers, Singapore, May 9-10, 2016, Revised Selected Papers. editor / Nardine Osman ; Carles Sierra. Springer, 2016. pp. 34-49 (Lecture Notes in Computer Science).

@inproceedings{37c9ad3bd55345a98333b14664a9f119,

title = "Can I Do That? Discovering Domain Axioms Using Declarative Programming and Relational Reinforcement Learning",

abstract = "Robots deployed to assist humans in complex, dynamic domains need the ability to represent, reason with, and learn from, different descriptions of incomplete domain knowledge and uncertainty. This paper presents an architecture that integrates declarative programming and relational reinforcement learning to support cumulative and interactive discovery of previously unknown axioms governing domain dynamics. Specifically, Answer Set Prolog (ASP), a declarative programming paradigm, is used to represent and reason with incomplete commonsense domain knowledge. For any given goal, unexplained failure of plans created by inference in the ASP program is taken to indicate the existenceof unknown domain axioms. The task of learning these axioms is formulatedas a Reinforcement Learning problem, and decision-tree regression with a relational representation is used to generalize from specific axioms identified over time. The new axioms are added to the ASP-based representationfor subsequent inference. We demonstrate and evaluate the capabilities of our architecture in two simulated domains: Blocks World and Simple Mario.",

author = "Mohan Sridharan and Prashanth Devarakonda and Rashmica Gupta",

year = "2016",

month = sep,

day = "24",

doi = "10.1007/978-3-319-46840-2_3",

language = "English",

isbn = "9783319468396",

series = "Lecture Notes in Computer Science",

publisher = "Springer",

pages = "34--49",

editor = "Osman, {Nardine } and Carles Sierra",

booktitle = "Autonomous Agents and Multiagent Systems",

note = "AAMAS 2016 Workshops, ; Conference date: 09-05-2016 Through 10-05-2016",

}

Sridharan, M, Devarakonda, P & Gupta, R 2016, Can I Do That? Discovering Domain Axioms Using Declarative Programming and Relational Reinforcement Learning. in N Osman & C Sierra (eds), Autonomous Agents and Multiagent Systems: AAMAS 2016 Workshops, Visionary Papers, Singapore, May 9-10, 2016, Revised Selected Papers. Lecture Notes in Computer Science, vol. 10003, Springer, pp. 34-49, AAMAS 2016 Workshops, , Singapore, 9/05/16. https://doi.org/10.1007/978-3-319-46840-2_3

Can I Do That? Discovering Domain Axioms Using Declarative Programming and Relational Reinforcement Learning. / Sridharan, Mohan; Devarakonda, Prashanth; Gupta, Rashmica.
Autonomous Agents and Multiagent Systems: AAMAS 2016 Workshops, Visionary Papers, Singapore, May 9-10, 2016, Revised Selected Papers. ed. / Nardine Osman; Carles Sierra. Springer, 2016. p. 34-49 (Lecture Notes in Computer Science; Vol. 10003).

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

TY - GEN

T1 - Can I Do That? Discovering Domain Axioms Using Declarative Programming and Relational Reinforcement Learning

AU - Sridharan, Mohan

AU - Devarakonda, Prashanth

AU - Gupta, Rashmica

PY - 2016/9/24

Y1 - 2016/9/24

N2 - Robots deployed to assist humans in complex, dynamic domains need the ability to represent, reason with, and learn from, different descriptions of incomplete domain knowledge and uncertainty. This paper presents an architecture that integrates declarative programming and relational reinforcement learning to support cumulative and interactive discovery of previously unknown axioms governing domain dynamics. Specifically, Answer Set Prolog (ASP), a declarative programming paradigm, is used to represent and reason with incomplete commonsense domain knowledge. For any given goal, unexplained failure of plans created by inference in the ASP program is taken to indicate the existenceof unknown domain axioms. The task of learning these axioms is formulatedas a Reinforcement Learning problem, and decision-tree regression with a relational representation is used to generalize from specific axioms identified over time. The new axioms are added to the ASP-based representationfor subsequent inference. We demonstrate and evaluate the capabilities of our architecture in two simulated domains: Blocks World and Simple Mario.

AB - Robots deployed to assist humans in complex, dynamic domains need the ability to represent, reason with, and learn from, different descriptions of incomplete domain knowledge and uncertainty. This paper presents an architecture that integrates declarative programming and relational reinforcement learning to support cumulative and interactive discovery of previously unknown axioms governing domain dynamics. Specifically, Answer Set Prolog (ASP), a declarative programming paradigm, is used to represent and reason with incomplete commonsense domain knowledge. For any given goal, unexplained failure of plans created by inference in the ASP program is taken to indicate the existenceof unknown domain axioms. The task of learning these axioms is formulatedas a Reinforcement Learning problem, and decision-tree regression with a relational representation is used to generalize from specific axioms identified over time. The new axioms are added to the ASP-based representationfor subsequent inference. We demonstrate and evaluate the capabilities of our architecture in two simulated domains: Blocks World and Simple Mario.

U2 - 10.1007/978-3-319-46840-2_3

DO - 10.1007/978-3-319-46840-2_3

M3 - Conference contribution

SN - 9783319468396

T3 - Lecture Notes in Computer Science

SP - 34

EP - 49

BT - Autonomous Agents and Multiagent Systems

A2 - Osman, Nardine

A2 - Sierra, Carles

PB - Springer

T2 - AAMAS 2016 Workshops,

Y2 - 9 May 2016 through 10 May 2016

ER -

Sridharan M, Devarakonda P, Gupta R. Can I Do That? Discovering Domain Axioms Using Declarative Programming and Relational Reinforcement Learning. In Osman N, Sierra C, editors, Autonomous Agents and Multiagent Systems: AAMAS 2016 Workshops, Visionary Papers, Singapore, May 9-10, 2016, Revised Selected Papers. Springer. 2016. p. 34-49. (Lecture Notes in Computer Science). doi: 10.1007/978-3-319-46840-2_3

Can I Do That? Discovering Domain Axioms Using Declarative Programming and Relational Reinforcement Learning

Abstract

Publication series

Conference

Access to Document

Fingerprint

Cite this