Difference between revisions of "The CINDY Cognitive Architecture"

From David Vernon's Wiki
Jump to: navigation, search
(Resources)
 
(8 intermediate revisions by the same user not shown)
Line 1: Line 1:
 
== Long-term Goal ==
 
== Long-term Goal ==
  
CINDY stands for '''Ci'''rcular '''N'''etwork '''Dy'''namics. As a project in developing a cognitive system, it is just beginning. It  is based on several other cognitive architectures, especially the iCub cognitive architecture.   
+
CINDY stands for '''Ci'''rcular '''N'''etwork '''Dy'''namics. As a project in developing a complete cognitive system, it is just getting started. It  is based on several other cognitive architectures, especially the iCub cognitive architecture [Vernon et al. 2007; Vernon et al. 2011].   
  
CINDY is a very early work-in-progress but the long-term goal is to design and implement of an emergent artificial cognitive system that builds on the principles of growth and development through circular causality, sharing many of the principles of the field of Systems Dynamics, Dynamical Systems Theory, and Enaction (e.g. positive feedback, circular causality, non-linearity, far-from equilibrium stability, endogenous change, structurally-determined dynamic patterns).
+
While CINDY is a work-in-progress the long-term goal is to design and implement of an emergent artificial cognitive system that builds on the principles of growth and development through circular causality, sharing many of the principles of the field of Systems Dynamics, Second-order Cybernetics, and Enaction (e.g. positive feedback, circular causality, non-linearity, far-from-equilibrium stability, endogenous change, structurally-determined dynamic patterns).
 
   
 
   
In the long-term, CINDY will feature:
+
At present, CINDY focusses mainly on visual attention because the attention problem shares many of the features of cognition in general.  Specifically, it has components dedicated to
  
* exogenous and endogenous attention (both overt and covert)
+
* exogenous attention (both overt and covert)
* joint perceptuo-motor representations (e.g. joint episodic-procedural memory)
+
* joint perceptuo-motor representations (i.e. joint episodic-procedural memory)
* value-system driven development
+
* action selection based on internal simulation
+
  
== Short-term Relevance ==
+
I am currently implementing
  
Right now, though, apart from some utility functions, it comprises just two prototype components that serve simply as examples of how a typical component in the architecture should be implemented.  Over the coming months, I will add more components,  some based on the original iCub cognitive architecture and some completely new. These will provide to the functionality expected in a cognitive architecture.
+
* endogenous attention (both overt and covert)
 +
* value-system driven development
  
'''These prototype components can also be used as the basis for developing applications using the component-port-connector model of component-based software engineering.'''  One of the prototypes features a graphic user interface (GUI) implemented using FLTK, while the other does not.
+
Eventually, CINDY will feature predictive action selection based on internal simulation
 +
 
 +
Currently, the attention system uses Bruce's Attention based Information Maximization  (AIM) saliency filter [Bruce and Tsotsos 2009], Tsotsos's Selective Tuning Model [Tstotsos et al 1995] for winner-take-all selection of fixation points, and a custom implementation of inhibition of return (IOR) and habituation using decaying and growing Gaussian function, respectively, the parameters of which are user selectable through a dedicated GUI.  Log-polar images of fixation points are stored in episodic memory and saccade gaze control commands are stored in procedural memory. These are combined as a network of image and gaze identification indices and stores in joint episodic-procedural memory.
 +
 +
CINDY, as a software system, also has two prototype components that serve as examples of how a typical component in the architecture should be implemented.  These are included to show students how to develop components for CINDY, in particular, but more generally '''these prototype components can also be used as the basis for developing applications using the component-port-connector model of component-based software engineering''' [Brugali and Scandurra 2009; Brugali and Shakhimardanov 2010, and Bruyninckx et al. 2013]. One of the prototypes features a graphic user interface (GUI) implemented using FLTK, while the other does not.
  
 
== What to Do Next ==  
 
== What to Do Next ==  
Line 24: Line 27:
 
There are also a set of software engineering standards which have been adopted to make sure the software can be maintained in the long term.  Please respect these standards if you are developing code that you intend to be integrated into the architecture.  
 
There are also a set of software engineering standards which have been adopted to make sure the software can be maintained in the long term.  Please respect these standards if you are developing code that you intend to be integrated into the architecture.  
  
These guides and standards were originally developed for the iCub in the RobotCub project and subsequently adapted for the DREAM project (see my website home page for links).
+
These guides and standards were originally developed for the [http://www.icub.org iCub] in the [http://www.robotcub.eu RobotCub] project and subsequently adapted for the [http://www.dream2020.eu DREAM] project.
  
 
== Developer Guides ==
 
== Developer Guides ==
Line 40: Line 43:
 
* [[Standards for Programming Practice]]
 
* [[Standards for Programming Practice]]
  
== Resources ==
+
== References ==
  
[http://vernon.eu/publications/07_Vernon_Metta_Sandini_ICDL.pdf D. Vernon, G. Metta, and G. Sandini, "The iCub Cognitive Architecture: Interactive Development in a Humanoid Robot", IEEE International Conference on Development and Learning, Imperial College, London, July 2007.]
+
N. D. B. Bruce and J. K. Tsotsos. Saliency, attention, and visual search: An information theoretic approach. Journal of Vision, 9(3):1–24, 2009.
 
+
[http://vernon.eu/publications/10_Vernon_vonHofsten_Fadiga_COSMOS_Extracts.pdf D. Vernon, C. von Hofsten, and L. Fadiga. "A Roadmap for Cognitive Development in Humanoid Robots", Cognitive Systems Monographs (COSMOS), Vol. 11, Springer, ISBN 978-3-642-16903-8, 2011.  Book Extracts.]
+
  
 
D. Brugali and P. Scandurra. Component-Based Robotic Engineering (Part I). IEEE Robotics and Automation Magazine, pages 84–96, December 2009.
 
D. Brugali and P. Scandurra. Component-Based Robotic Engineering (Part I). IEEE Robotics and Automation Magazine, pages 84–96, December 2009.
Line 50: Line 51:
 
D. Brugali and A. Shakhimardanov. Component-Based Robotic Engineering (Part II). IEEE Robotics and Automation Magazine, pages 100–112, March 2010.
 
D. Brugali and A. Shakhimardanov. Component-Based Robotic Engineering (Part II). IEEE Robotics and Automation Magazine, pages 100–112, March 2010.
  
H. Bruyninckx, M. Klotzbu ̈cher, N. Hochgeschwender, G. Kraetzschmar, L. Gherardi, and D. Brugali. The BRICS component model: A model-based development paradigm for complex robotics software systems. In Proceedings of the 28th Annual ACM Symposium on Applied Computing, SAC ‘13, pages 1758–1764, New York, NY, USA, 2013. ACM.
+
H. Bruyninckx, M. Klotzbücher, N. Hochgeschwender, G. Kraetzschmar, L. Gherardi, and D. Brugali. The BRICS component model: A model-based development paradigm for complex robotics software systems. In Proceedings of the 28th Annual ACM Symposium on Applied Computing, SAC ‘13, pages 1758–1764, New York, NY, USA, 2013. ACM.
 +
 
 +
J. K. Tsotsos, S. Culhane, W. Wai, Y. Lai, N. David, and F. Nuflo.    Modeling visual attention via selective tuning. Artificial Intelligence, 78:507–547, 1995.
 +
 
 +
[http://vernon.eu/publications/07_Vernon_Metta_Sandini_ICDL.pdf D. Vernon, G. Metta, and G. Sandini, "The iCub Cognitive Architecture: Interactive Development in a Humanoid Robot", IEEE International Conference on Development and Learning, Imperial College, London, July 2007.]
 +
 
 +
[http://vernon.eu/publications/10_Vernon_vonHofsten_Fadiga_COSMOS_Extracts.pdf D. Vernon, C. von Hofsten, and L. Fadiga. "A Roadmap for Cognitive Development in Humanoid Robots", Cognitive Systems Monographs (COSMOS), Vol. 11, Springer, ISBN 978-3-642-16903-8, 2011.  Book Extracts.]
 +
 
 +
[http://vernon.eu/publications/15_Vernon_et_al_Frontiers.pdf D. Vernon, R. Lowe, S. Thill, and  T. Ziemke, "Embodied Cognition and Circular Causality: On the Role of Constitutive Autonomy in the Reciprocal Coupling of Perception and Action"], [http://journal.frontiersin.org/article/10.3389/fpsyg.2015.01660/full Frontiers in Psychology, Vol. 6, No. 1660, 2015]
 +
 
 +
[http://vernon.eu/15_Vernon_Beetz_Sandini_Frontiers.pdf D. Vernon, M. Beetz, and G. Sandini, "Prospection in cognition: the case for joint episodic-procedural memory"], [http://journal.frontiersin.org/article/10.3389/frobt.2015.00019/full Frontiers in AI and Robotics, Vol. 2, No. 19, 2015.]

Latest revision as of 04:35, 17 July 2016

Long-term Goal

CINDY stands for Circular Network Dynamics. As a project in developing a complete cognitive system, it is just getting started. It is based on several other cognitive architectures, especially the iCub cognitive architecture [Vernon et al. 2007; Vernon et al. 2011].

While CINDY is a work-in-progress the long-term goal is to design and implement of an emergent artificial cognitive system that builds on the principles of growth and development through circular causality, sharing many of the principles of the field of Systems Dynamics, Second-order Cybernetics, and Enaction (e.g. positive feedback, circular causality, non-linearity, far-from-equilibrium stability, endogenous change, structurally-determined dynamic patterns).

At present, CINDY focusses mainly on visual attention because the attention problem shares many of the features of cognition in general. Specifically, it has components dedicated to

  • exogenous attention (both overt and covert)
  • joint perceptuo-motor representations (i.e. joint episodic-procedural memory)

I am currently implementing

  • endogenous attention (both overt and covert)
  • value-system driven development

Eventually, CINDY will feature predictive action selection based on internal simulation

Currently, the attention system uses Bruce's Attention based Information Maximization (AIM) saliency filter [Bruce and Tsotsos 2009], Tsotsos's Selective Tuning Model [Tstotsos et al 1995] for winner-take-all selection of fixation points, and a custom implementation of inhibition of return (IOR) and habituation using decaying and growing Gaussian function, respectively, the parameters of which are user selectable through a dedicated GUI. Log-polar images of fixation points are stored in episodic memory and saccade gaze control commands are stored in procedural memory. These are combined as a network of image and gaze identification indices and stores in joint episodic-procedural memory.

CINDY, as a software system, also has two prototype components that serve as examples of how a typical component in the architecture should be implemented. These are included to show students how to develop components for CINDY, in particular, but more generally these prototype components can also be used as the basis for developing applications using the component-port-connector model of component-based software engineering [Brugali and Scandurra 2009; Brugali and Shakhimardanov 2010, and Bruyninckx et al. 2013]. One of the prototypes features a graphic user interface (GUI) implemented using FLTK, while the other does not.

What to Do Next

The following sections will help you install the current implementation and run some test applications. It also provides a comprehensive overview of how to develop a new component for the architecture.

There are also a set of software engineering standards which have been adopted to make sure the software can be maintained in the long term. Please respect these standards if you are developing code that you intend to be integrated into the architecture.

These guides and standards were originally developed for the iCub in the RobotCub project and subsequently adapted for the DREAM project.

Developer Guides

Software Engineering Standards

References

N. D. B. Bruce and J. K. Tsotsos. Saliency, attention, and visual search: An information theoretic approach. Journal of Vision, 9(3):1–24, 2009.

D. Brugali and P. Scandurra. Component-Based Robotic Engineering (Part I). IEEE Robotics and Automation Magazine, pages 84–96, December 2009.

D. Brugali and A. Shakhimardanov. Component-Based Robotic Engineering (Part II). IEEE Robotics and Automation Magazine, pages 100–112, March 2010.

H. Bruyninckx, M. Klotzbücher, N. Hochgeschwender, G. Kraetzschmar, L. Gherardi, and D. Brugali. The BRICS component model: A model-based development paradigm for complex robotics software systems. In Proceedings of the 28th Annual ACM Symposium on Applied Computing, SAC ‘13, pages 1758–1764, New York, NY, USA, 2013. ACM.

J. K. Tsotsos, S. Culhane, W. Wai, Y. Lai, N. David, and F. Nuflo. Modeling visual attention via selective tuning. Artificial Intelligence, 78:507–547, 1995.

D. Vernon, G. Metta, and G. Sandini, "The iCub Cognitive Architecture: Interactive Development in a Humanoid Robot", IEEE International Conference on Development and Learning, Imperial College, London, July 2007.

D. Vernon, C. von Hofsten, and L. Fadiga. "A Roadmap for Cognitive Development in Humanoid Robots", Cognitive Systems Monographs (COSMOS), Vol. 11, Springer, ISBN 978-3-642-16903-8, 2011. Book Extracts.

D. Vernon, R. Lowe, S. Thill, and T. Ziemke, "Embodied Cognition and Circular Causality: On the Role of Constitutive Autonomy in the Reciprocal Coupling of Perception and Action", Frontiers in Psychology, Vol. 6, No. 1660, 2015

D. Vernon, M. Beetz, and G. Sandini, "Prospection in cognition: the case for joint episodic-procedural memory", Frontiers in AI and Robotics, Vol. 2, No. 19, 2015.