Embedded Neural Networks

Using concepts and tools of embodied cognitive science, we investigate the implications of embedding neural networks in a physical structure, the body of a robot. Through this embedding the loop from a network’s outputs to its subsequent inputs is closed. This closure enables an embedded network to actively generate its own input data instead of only passively processing predesigned input patterns. We argue that this simplifies some of the hard problems faced by disembodied neural networks. It can also help to deepen our understanding of biological neural networks, which are likewise embedded in a physical structure, the animals’ body. The main arguments are illustrated with a series of case studies with simulated and physical mobile robots that are controlled by hand-designed as well as evolved neural networks.