
neuroWalknet, a decentralized, context sensitive controller to explain insect walking
Holk Cruse
Bielefeld University
https://neurotree.org/beta/publications.php?pid=8331
Date(s) : 09/06/2021 iCal
14h00 - 15h00
Legged locomotion comprises various tasks as are standing still, walking with different footfall patterns (or „gaits“), different velocities, forward or backward, negotiation of curves with different radii, dealing with disturbances, controlling stance, swing, and searching (and other behaviors as e.g. jumping or swimming). Here we show a controller for hexapod walking, neuroWalknet, that is considered part of a larger decentralized hierarchy, consisting of context sensitive local controllers being arranged at different levels of organization. neuroWalknet is able to describe and perhaps explain a large amount of behavioral data plus “behavior” of partly or fully deafferented animals, known e.g. from stick insect experiments. Although it has often been postulated that these behaviors are based on application of central pattern generators, our network shows that such elements are not necessary to explain the data. In contrast, not using CPGs may simplify its performance. Therefore, following Popper, we propose to endorse the simpler solution as best hypothesis.
https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1007804
Catégories Pas de Catégories