A Bio-inspired Distributed Control Architecture: Coupled Artificial Signalling Networks

This thesis studies the applicability of computational models inspired by the structure and dynamics of signalling networks to the control of complex control problems. In particular, this thesis presents two different abstractions that aim to capture the signal processing abilities of biological cells: a stand-alone signalling network and a coupled signalling network. While the former mimics the interacting relationships amongst the components in a signalling pathway, the latter replicates the connectionism amongst signalling pathways. After initially investigating the feasibility of these models for controlling two complex numerical dynamical systems, Chirikov's standard map and the Lorenz system, this thesis explores their applicability to a difficult real world control problem, the generation of adaptive rhythmic locomotion patterns within a legged robotic system. The results highlight that the locomotive movements of a six-legged robot could be controlled in order to adapt the robot's trajectory in a range of challenging environments. In this sense, signalling networks are responsible for the robot adaptability and inter limb coordination as they self-adjust their dynamics according to the terrain's irregularities. More generally, the results of this thesis highlight the capacity of coupled signalling networks to decompose non-linear problems into smaller sub-tasks, which can then be independently solved.