Meeting Abstract
High level tasks in biology, e.g. locomotion, are often achieved with distributed control of coupled subcomponents, e.g. muscles and limbs. The coupling of these subcomponents could range from weak and local, i.e. decentralized, to strong and global, i.e centralized. We developed a measure of centralization that compares information shared between control signals and both global and local outputs. We previously found that running cockroaches lie on the centralized side of this centralized/decentralized axis. To both validate the information measure of centralization and to contextualize the cockroach result with an intuitive system, we analyzed a computational model of coupled oscillators. Our centralization measure successfully reconstructs the shift from low to high coupling strengths. Intermediate values of coupling strengths which result in positive centralization correspond well to previous fits of this model to cockroach data indicating that this model can capture the centralization of the cockroach. Since mechanical coupling in a physical system might always result in a centralized system according to our measure, we analyzed a robotic model to contrast a decentralized architecture from the centralized cockroach. The robot control was designed to be reactive to local feedback, where coordination arises through mechanical coupling that can be altered by changing the robot’s inertia. The robot is decentralized compared to the cockroach according to our measure and is maximally decentralized when mechanical coupling is decreased even though leg coordination persists. These results affirm that the cockroach does use a comparatively centralized control architecture and shows that our measure successfully assesses centralization using empirical observations.
