Basically, my scientific interests concern how brains control behavior, i.e. how brains control complex movements. Pursuing this goal, we concentrate on control of walking in multi-legged animals, insects and crustaceans. Behavioural studies including walking on various substrates are accompanied by simulation studies on the basis of artificial neural networks. The resulting network, called Walknet, represents an extremely decentralized system able to control a body with many extra degrees of freedom and to simulate a large number of behavioural findings.
The properties of the simulation system are tested on simulated and physical robots, which is necessary because the body (plus the environment) is considered an essential element of the computational system.
This neural architecture has later been equipped with a further network, Navinet, that enables the robot to perform insect-like navigation (path integration, landmark navigation) without using a “cognitive map”.
More recently, we have expanded this network by a body model to allow internal simulation of behavior. This ability is used to test a newly invented behavior via “internal trial-and-error” before it will be performed in reality. This system, called reaCog, is considered to show cognitive properties based on reactive structures and is claimed to be endowed with emergent properties as are intention, volition, attention and some aspects of consciousness.