Robust LQR-Based Architecture for Faulty Networked Control Systems

This paper proposes a networked control architecture that provides robustness not only to parametric uncertainties in the model but also to packet dropouts and time-varying network-induced delays. Contrary to the common approach of nonlinear optimization solvers, the estimation and control approaches stand out by providing a direct solution derived from an Extended Kalman Filter and a Recursive Robust Linear-Quadratic Regulator. Moreover, we prove that the optimal robust solution for a single agent suffices for the control of a homogeneous multi-agent scenario. Numerous simulations analyze the impact of different network conditions on the proposed architecture for a multi-agent scenario. Finally, the architecture was implemented and made available as a Robot Operating System (ROS) package, and trajectory-tracking missions were experimentally carried out to show the effectiveness of the proposed system.