COBOSLAB has moved!!!

COBOSLAB
COgnitive BOdy Spaces: Learning And Behavior.

The COBOSLAB develops artificial self-organized cognitive learning systems that learn multimodal modular sensorimotor bodyspace representations for effective learning and behavior.

Project goal is the development and implementation of artificial adaptive systems that learn, develop, and behave autonomously based on learning principles derived from cognitive psychology and neuroscience. Meanwhile, we investigate behavioral flexibilities and spatial representations and perceptions to (1) verify or evaluate the developed computational models and to (2) gain further insights on how space is perceived and behavior is controlled.

Funding comes from the EmmyNoether program of the German Research Foundation (DFG).

Bodyspaces are internal representations of the own body and parts thereof with respect to the surrounding space. Bodyspaces are multimodal, modular, and highly interactive. Bodyspaces develop by self-organization principles based on sensory-motor correlations. Thus, a bodyspace defines space not by absolute distances but by behavioral distances, which are inevitably body-grounded.

Learning focuses on distributing and connecting the bodyspaces "from scratch" - simply by experiencing body part activities during "motor babbling". Sensorimotor encodings will yield maximally behaviorally effective structures.

Behavior realized within bodyspace representations is goal-directed and highly flexible. Goal locations are activated dependent on homeostatic-based self-motivations and current bodily- and environmental constraints.