Kira Mourao PhD

Kira Mourao


Publications:
2014
  Learning probabilistic planning operators from noisy observations
Mourao, K 2014, Learning probabilistic planning operators from noisy observations. in 31st Workshop of the UK Planning & Scheduling Special Interest Group (PlanSIG 2013).
Building agents which can learn to act autonomously in the world is an important challenge for artificial intelligence. While autonomous agents often have to operate in noisy, uncertain worlds, current methods to learn action models from agents’ experiences typically assume fully deterministic worlds. This paper presents a noise-tolerant approach to learning probabilistic planning operators from experience. Preliminary experiments demonstrate that the approach learns accurate models even if agents’ observations are noisy.
General Information
Organisations: Institute of Language, Cognition and Computation .
Authors: Mourao, Kira.
Number of pages: 2
Publication Date: 2014
Publication Information
Category: Conference contribution
Original Language: English
2013
  Learning Knowledge-Level Domain Dynamics
Mourao, K & Petrick, R 2013, Learning Knowledge-Level Domain Dynamics. in Proceedings of the ICAPS 2013 Workshop on Planning and Learning. pp. 23-31.
The ability to learn relational action models from noisy, incomplete observations is essential to support planning and decision-making in real-world environments. While some methods exist to learn models of STRIPS domains in this setting, these approaches do not support learning of actions at the knowledge level. In contrast, planning at the knowledge level has been explored and in some domains can be more successful than planning at the world level. In this paper we therefore present a method to learn knowledge-level action models. We decompose the learning problem into multiple classification problems, generalising previous decompositional approaches by using a graphical deictic representation. We also develop a similarity measure based on deictic reference which generalises previous STRIPS-based approaches to similarity comparisons of world states. Experiments in a real robot domain demonstrate our approach is effective.
General Information
Organisations: Institute of Language, Cognition and Computation .
Authors: Mourao, Kira & Petrick, Ron.
Number of pages: 9
Pages: 23-31
Publication Date: Jun 2013
Publication Information
Category: Conference contribution
Original Language: English
2012
  Learning STRIPS Operators from Noisy and Incomplete Observations
Mourao, K, Zettlemoyer, L, Petrick, R & Steedman, M 2012, Learning STRIPS Operators from Noisy and Incomplete Observations. in Proceedings of the Twenty Eighth Conference on Uncertainty in Artificial Intelligence (UAI 2012). pp. 614-623.
Agents learning to act autonomously in real world domains must acquire a model of the dynamics of the domain in which they operate. Learning domain dynamics can be challenging, especially where an agent only has partial access to the world state, and/or noisy external sensors. Even in standard STRIPS domains, existing approaches cannot learn from noisy, incomplete observations typical of real-world domains. We propose a method which learns STRIPS action models in such domains, by decomposing the problem into first learning a transition function between states in the form of a set of classifiers, and then deriving explicit STRIPS rules from the classifiers’ parameters. We evaluate our approach on simulated standard planning domains from the International Planning Competition, and show that it learns useful domain descriptions from noisy, incomplete observations.
General Information
Organisations: Neuroinformatics DTC.
Authors: Mourao, Kira, Zettlemoyer, Luke, Petrick, Ron & Steedman, Mark.
Number of pages: 10
Pages: 614-623
Publication Date: 2012
Publication Information
Category: Conference contribution
Original Language: English
2010
  Learning action effects in partially observable domains
Mourao, K, Petrick, R & Steedman, M 2010, Learning action effects in partially observable domains. in H Coelho, R Studer & M Wooldridge (eds), ECAI 2010 - 19th European Conference on Artificial Intelligence, Lisbon, Portugal, August 16-20, 2010, Proceedings.. IOS Press, pp. 973-974. DOI: 10.3233/978-1-60750-606-5-973
We investigate the problem of learning action effects in partially observable STRIPS planning domains. Our approach is based on a voted kernel perceptron learning model, where action and state information is encoded in a compact vector representation as input to the learning mechanism, and resulting state changes are produced as output. Our approach relies on deictic features that assume an attentional mechanism that reduces the size of the representation. We evaluate our approach on a number of partially observable planning domains, and show that it can quickly learn the dynamics of such domains, with low average error rates. We show that our approach handles noisy domains, conditional effects, and that it scales independently of the number of objects in a domain.
General Information
Organisations: Neuroinformatics DTC.
Authors: Mourao, Kira, Petrick, Ron & Steedman, Mark.
Number of pages: 2
Pages: 973-974
Publication Date: 2010
Publication Information
Category: Conference contribution
Original Language: English
DOIs: 10.3233/978-1-60750-606-5-973
2009
  Learning action effects in partially observable domains
Mourao, K, Petrick, R & Steedman, M 2009, Learning action effects in partially observable domains. in Proceedings of the ICAPS 2009 Workshop on Planning and Learning. pp. 15-22.
We investigate the problem of learning action effects in partially observable STRIPS planning domains. Our approach is based on a voted kernel perceptron learning model, where action and state information is encoded in a compact vector representation as input to the learning mechanism, and resulting state changes are produced as output. Our approach relies on deictic features that embody a notion of attention and reduce the size of the representation. We evaluate our approach on a number of partially observable planning domains, adapted from domains used in the International Planning Competition, and show that it can quickly learn the dynamics of such domains, with low average error rates. Furthermore, we show that our approach handles noisy domains, and scales independently of the number of objects in a domain, making it suitable for large planning scenarios.
General Information
Organisations: Institute of Language, Cognition and Computation .
Authors: Mourao, Kira, Petrick, Ron & Steedman, Mark.
Number of pages: 8
Pages: 15-22
Publication Date: 2009
Publication Information
Category: Conference contribution
Original Language: English
2008
  Using Kernel Perceptrons to Learn Action Effects for Planning
Mourao, K, Petrick, R & Steedman, M 2008, Using Kernel Perceptrons to Learn Action Effects for Planning. in Proceedings of the International Conference on Cognitive Systems (CogSys 2008). pp. 45-50.
We investigate the problem of learning action effects in STRIPS and ADL planning domains. Our approach is based on a kernel perceptron learning model, where action and state information is encoded in a compact vector representation as input to the learning mechanism, and resulting state changes are produced as output. Empirical results of our approach indicate efficient training and prediction times, with low average error rates (< 3%) when tested on STRIPS and ADL versions of an object manipulation scenario. This work is part of a project to integrate machine learning techniques with a planning system, as part of a larger cognitive architecture linking a high-level
reasoning component with a low-level robot/vision system.
General Information
Organisations: Neuroinformatics DTC.
Authors: Mourao, Kira, Petrick, Ron & Steedman, Mark.
Number of pages: 6
Pages: 45-50
Publication Date: Apr 2008
Publication Information
Category: Conference contribution
Original Language: English
  Representation and Integration: Combining Robot Control, High-Level Planning, and Action Learning
Petrick, R, Kraft, D, Mourao, K, Geib, C, Pugeault, N, Krüger, N & Steedman, M 2008, Representation and Integration: Combining Robot Control, High-Level Planning, and Action Learning. in Proceedings of the International Cognitive Robotics Workshop (CogRob 2008) at ECAI 2008. pp. 32-41.
We describe an approach to integrated robot control, high-level planning, and action effect learning that attempts to overcome the representational difficulties that exist between these diverse areas. Our approach combines ideas from robot vision, knowledge-level planning, and connectionist machine learning, and focuses on the representational needs of these components. We also make use of a simple representational unit called an instantiated state transition fragment (ISTF) and a related structure called an object-action complex (OAC). The goal of this work is a general approach for inducing high-level action specifications, suitable for planning, from a robot’s interactions with the world. We present a detailed overview of our approach and show how it supports the learning of certain aspects of a high-level representation from low-level world state information.
General Information
Organisations: Neuroinformatics DTC.
Authors: Petrick, Ron, Kraft, Dirk, Mourao, Kira, Geib, Christopher, Pugeault, Nico, Krüger, Norbert & Steedman, Mark.
Number of pages: 10
Pages: 32-41
Publication Date: Jul 2008
Publication Information
Category: Conference contribution
Original Language: English
2007
  Setting the tone: An ERP study of the effects of intonation on sentence processing
Mourao, K, Steedman, M & Donaldson, DI 2007, 'Setting the tone: An ERP study of the effects of intonation on sentence processing'.
General Information
Organisations: Royal (Dick) School of Veterinary Studies.
Authors: Mourao, Kira, Steedman, Mark & Donaldson, David I..
Publication Date: 2007
Publication Information
Category: Poster
Original Language: English
  Effects of sentence structure on sentence processing - an ERP study
Mourao, K, Steedman, M & Donaldson, DI 2007, 'Effects of sentence structure on sentence processing - an ERP study'.
General Information
Organisations: Royal (Dick) School of Veterinary Studies.
Authors: Mourao, Kira, Steedman, Mark & Donaldson, David I..
Publication Date: 2007
Publication Information
Category: Poster
Original Language: English
2006
  From Intonation to Information - an ERP study
Mourao, K, Steedman, M & Donaldson, DI 2006, 'From Intonation to Information - an ERP study'.
General Information
Organisations: Royal (Dick) School of Veterinary Studies.
Authors: Mourao, Kira, Steedman, Mark & Donaldson, David I..
Publication Date: 2006
Publication Information
Category: Poster
Original Language: English
  Object Action Complexes as an Interface for Planning and Robot Control
Geib, C, Mourao, K, Petrick, R, Pugeault, N, Steedman, M, Krüger, N & Wörgötter, F 2006, Object Action Complexes as an Interface for Planning and Robot Control. in IEEE-RAS Humanoids-06 Workshop: Towards Cognitive Humanoid Robots.
Much prior work in integrating high-level artificial intelligence planning technology with low-level robotic control has foundered on the significant representational differences between these two areas of research. We discuss a proposed solution to this representational discontinuity in the form of object-action complexes (OACs). The pairing of actions and objects in a single interface representation captures the needs of both reasoning levels, and will enable machine learning of high-level action representations from low-level control representations.
General Information
Organisations: Neuroinformatics DTC.
Authors: Geib, Christopher, Mourao, Kira, Petrick, Ron, Pugeault, Nico, Steedman, Mark, Krüger, Norbert & Wörgötter, Florentin.
Number of pages: 6
Publication Date: 2006
Publication Information
Category: Conference contribution
Original Language: English

Projects:
Actions speak louder than words: representing the world in robot actions (PhD)