An Evolutionary Algorithm for Actuator-Sensor-Communication Co-Design in Distributed Control

Researchers Pengyang Wu and Jing Shuang Li have published a paper introducing a novel evolutionary algorithm designed to solve the complex co-design problem in distributed control systems. The core challenge is to jointly optimize the selection and placement of three critical elements: physical actuators, sensors, and the communication links between sub-controllers, all while minimizing both performance cost (measured by Linear-Quadratic/LQ cost) and the material cost of the hardware and network itself. Their approach starts with a baseline, densely connected LQR controller and uses an evolutionary algorithm to strategically 'prune' unnecessary components, effectively finding the most cost-effective architecture without sacrificing stability.

The algorithm's performance is demonstrated through simulations, with one standout result involving a 98-state model of a swing equation—a classic problem in power system stability. Remarkably, the co-design process for this complex model runs in seconds on a standard laptop. The research highlights that this intelligent, joint optimization significantly outperforms simpler 'naive' pruning methods, which remove components without considering their interdependencies, by over 50% in combined cost savings. For inherently unstable systems, where pruning risks causing failure, the authors provide a modified version of their algorithm that incorporates stability guarantees, making the tool practical for real-world engineering applications where reliability is non-negotiable.