Measuring Spatiotemporal Coordination in a Modular Robotic System

In this paper we propose and verify spatiotemporal measures of coordination in a modular robotic system. These information-theoretic measures estimate the generalized “correlation entropy” K2 and the generalized excess entropy E2 computed over a multivariate time series of actuators’ states, while standard deviation of E2 is minimized over both space and time. Actuators are shown to be well-coordinated in individuals with fastest locomotion. These results support our conjecture that direct fitness functions can be approximated with generic selection pressures, leading towards information-driven evolutionary design.