SEMRINT - SEnsoMotorische Real-World-INTeraktion


The subject of the SEMRINT project is the real-world connection of the neuronal mechanisms developed in the DFG project SEMINT, with the aim of demonstrating that this "perception-through-participation" approach to the self-organisation of a system-specific understanding of space and form can also be used for the visual control of an autonomous mobile robot in a real natural environment.

It will be shown that planning mechanisms of action selection are suitable for local navigation based on information from the visual flow. By actions of the system in its environment, an associative knowledge base of general sensorimotor relations shall be built on-line. Based on this knowledge base and current flow field information, action sequences are to be planned by an internal simulation of future situations and actions. Action sequences with positive evaluation are then executed.

The specific aspects of visual processing, which are crucial for a real-world connection, are to be investigated and processed on the MILVA and Khepera mobile robot systems, with real-time implementation on the CNAPS neurocomputer forming a further essential focus.