Topology Effects on Sparse Control of Complex Networks with Laplacian Dynamics

Abstract

Ease of control of complex networks has been assessed extensively in terms of structural controllability and observability, and minimum control energy criteria. Here we adopt a sparsity-promoting feedback control framework for undirected networks with Laplacian dynamics and distinct topological features. The control objective considered is to minimize the effect of disturbance signals, magnitude of control signals and cost of feedback channels. We show that depending on the cost of feedback channels, different complex network structures become the least expensive option to control. Specifically, increased cost of feedback channels favors organized topological complexity such as modularity and centralization. Thus, although sparse and heterogeneous undirected networks may require larger numbers of actuators and sensors for structural controllability, networks with Laplacian dynamics are shown to be easier to control when accounting for the cost of feedback channels.

Discovering adaptation-capable biological network structures using control-theoretic approaches

“…Beyond mathematical modeling, systems theory has been particularly useful to understand and characterize various biological systems [4]. Graph-theoretic tools have also found applications in analyzing and understanding biological networks as functional modules [5][6][7][8][9]. Notably, it has been seen through experiments [10] that the design principles, for any given biological response, are relatively conserved across organisms [11].…”

Network diffusion capacity unveiled by dynamical paths

“…It was also found that some diffusion processes present not trivial behaviors such as super-diffusion, by which the multiplex structure reaches a steady-state faster than any of the layers in isolation [35]. This approach has been successfully used in several applications [36][37][38][39][40] and later expanded to other phenomena [41] such as synchronization and reaction-diffusion processes [42][43][44][45]. Diffusion times computed by diffusion capacity are in excellent agreement with those determined by the Laplacian matrix's smallest positive eigenvalue (see section B of the SI).…”

Decomposition of control and optimization problems by network structure: Concepts, methods, and inspirations from biology

“…• In Constantino et al, 129 a large amount of random networks with diverse topological features were examined, and it was shown that modularity and centrality (or hierarchy), i.e., the existence of communities and the existence of a small core of nodes to which all nodes are inclined to connect, are the main topological features to lower the aforementioned total control cost when the weight of controller sparsity is moderate and high, respectively.…”