- (Goertzel & Garis, 2008) ⇒ Ben Goertzel, and Hugo de Garis. (2008). “XIA-MAN: An Extensible, Integrative Architecture for Intelligent Humanoid Robotics.”
The XIA-MAN architecture for intelligent humanoid robot control is proposed, a novel design in which perception and action are achieved via a combination of GA-evolved neural-net modules with existing open-source software packages; and cognition is achieved via the OpenCog Prime framework. XML is used to communicate between components, enabling simple pluggability of additional or modified components, and leading to the name XIA-MAN (eXtensible Integrative Artificial Man). XIA-MANâs neural net modules are used to convert between high-dimensional numerical representations denoting perceptions and actions, and probabilistic symbolic representations suitable for cognitive manipulation. XIA-MANâs Cognition Engine is used, at each time cycle, to choose a high-level behavioral procedure that is believes is likely to enable the robot to achieve its goals in the current context. This provides a pragmatic approach to achieving intelligent robot functionality given currently available technologies; and the architecture is conceptually reminiscent of the complexly interconnected multimodular architecture of the brain. Initial work involves an incarnation of the XIA-MAN architecture using the Nao humanoid robot, to create an intelligent humanoid called XiaoNao.
The XIA-MAN Architecture
A Two-Phase Development Plan
Examples of Desired XIA-MAN Behavior s
In What Sense Is XIA-MAN BiologicallyInspired?
Evolving Neural Nets Carrying Out Perceptual and Motor Functions
Genetic Algorithm Evolution of Neural Net Modules in Software and Hardware, Using the Parcone Model
Evolving Neural Modules for Object Recognition
Evolving Neural Modules for Movement Control
Evolving Neural Modules for Emotion and Modality Recognition
From Supervised to Unsupervised Learning
The Cognition Engine
Novamente Pet Brain
The RelEx System for Natural Language Comprehension and Generation
OpenCog also contains a significant piece of software, donated by Novamente LLC, which is not strictly a part of the NCE although it interoperates with the NCE: this is the RelEx engine for natural language comprehension and generation. The comprehension aspect of RelEx is more mature and has been briefly described in (Goertzel et al, 2006) in the context of its application to interpreting biomedical research abstracts; the generation aspect is still in prototype phase, but has already been shown to work on simple sentences.
RelEx, in its comprehension aspect, takes English text and maps that text into abstract logical relations, in the Atom format utilized internally by the NCE and OpenCog. Generally speaking it produces multiple interpretations (logical relation sets) for each sentence it processes, and the task of selecting the contextually appropriate interpretation is left to the cognition engine itself. Also, the cognition engine is relied upon to correct errors RelEx may make in areas such as word sense disambiguation and reference resolution. It is anticipated that the sensory data gathered by a robot, regarding the physical and social context of instances of linguistic usages it produces or hears, may provide data helpful to the cognition engine in executing the linguistic tasks of interpretation-selection, reference resolution and sense disambiguation.
Next, in its generation aspect, RelEx maps logical relation sets (Atom sets) into sets of English sentences. Note that RelEx does not contain any facility for discourse management: this is assumed to be handled within the cognition engine. A design exists for controlling dialogue within OpenCog utilizing a probabilistic implementation of ideas from Rhetorical Structure Theory (Mann and Thompson, 1988), but this still awaits implementation.
RelEx also does not contain any facility for converting speech to text or vice versa. In the proposed integrated architecture for robot control, these conversions will be carried out by existing open-source software and RelEx will be used (together with the Cognition Engine) to select between the multiple outputs of speech-to-text software in an intelligent way.
Conclusion and Further Work
|2008 XIAMANAnExtensibleIntegrativeAr||Ben Goertzel|
Hugo de Garis
|XIA-MAN: An Extensible, Integrative Architecture for Intelligent Humanoid Robotics||2008|