The conference at the conference "living machines" beginning with a pladoyer for central pattern generator (cpgs) as an unearthable component for the control of running behavior in living beings and robots. She ended with a lecture, which contradicted exactly this statement – and relieves the impression that the central questions of behavioral control and cognition are still open.
The contrast between the keynote of auke ijsperte, with which the conference open living machines, and that of holk cruse (uni bielefeld), which it was decided, but it will be mitigated that both relate to different organisms. While ijspeert priority reptiles as the salamander examined, cruse worked with insects. He presented a decentralized neural network, which controls the movement of the legs on the lowest level, regulates the navigation on a middle level and to make the highest level cognition.
The procedure was implemented on the six-legged robot hector. But not only here showed that cpgs are not required for the control of the running behavior, according to cruse. Also experiments with insects in which the signal transport was interrupted by the body to the brain by pilocarpin pointed out that the running apparently be controlled by a phase coupling of adjacent legs.
The video shows the first steps from hector 2014.
Watch your step
In the artificial neural network, the leg movements are coordinated by sensory memories (smm). As a rust that connects to the high levels of the network, the motivation unit serves (mu). Such a neuronal layer that is not needed to run to run is required to create cognition out of problematic situations. As an example, cruse called the case that a mid-leg of six-legged slipped into a hole and the robot now has a solution to continue running. In a corresponding experiment, the purely reactive system had been sufficient in 70.3 percent of the trap, at 28.9 percent hector has "ponder" (cognitive expansion) found a solution. Only with 0.7 percent of the experimental runs is the robot stuck.
In insects, the sensory feedback for the control of running is much more important than cpgs, cruse summarizes its executions. He realized, however, that this could look very different with vertebrates – and met again in this point with auke ijspeert. For his part, the importance of the cpgs opposite the sensory feedback was relativized in humans because of whose principal instability. So the conference is a deep impression of the enormous variety of nature – and of a community of researchers who are looking for with gross creativity and passion for answers to the related questions. That spectacular results have failed this year, let it be able to handle it without further ado.