Fig. 1
From: Information-based centralization of locomotion in animals and robots
Measuring Centralization of Locomotor Systems. a Locomoting systems can use numerous possible control signals which cascade information through successive levels of integration to effect a relatively small number of global task related outcomes. Coupling between control can range from weak and local (decentralized) to strong and global (centralized). Arrows indicated direction of information flow. Lines connecting the subsystems represent mechanical coupling, with thickness signifying coupling strength. Centralization could also be affected by the degree of coupling in the internal architecture of control, as well as the strength of global feedback versus local feedback. b Our centralization measure uses empirical observations of control signals, local states, and global states to infer the coupling of control. We estimate the mutual information between the control signal and both the global and local states. We measure centralization as IG minus IL, which removes any IR between the three variables. We expect decentralized systems to contain more IL and centralized systems to contain more IG. IUG, IUL, ISYN, and IR are the unmeasured constituent parts of the total information ITOT shared between the control signal and both local and global outputs
