Print this page

High-Assurance Control of Networked Control Systems (NCS)

 

Most control systems such as aircraft, industrial process control plants, biological systems, economical systems, networked control systems (NCS) are characterized by a dynamic phenomena due to time-delays. Further, their ever-growing complexity results in obtaining less accurate mathematical models (i.e., parametric and disturbance uncertainties) in order to explain their behaviors. The input time-delays are common in control systems due to the time it takes to acquire information, analyze, and compute control inputs, and further execute them. Also, the computational and communication delay manifest as input delays. This is unlike traditional feedback control systems which are point-to-point connected, where information acquisition, computation, and implementation of control actions are assumed to be instantaneous. Thus, the presence of time-delay in a system introduces additional dynamics that cannot be neglected, especially when the amplitude of the delay is comparable to the process response time. Additionally, most systems are subject to process variations such as parameter uncertainties (e.g., changing mass due to fuel consumption) and external disturbances (e.g., friction dynamics, unmodeled dynamics), which can be either stochastic or non-stochastic but bounded in nature. This results in most mathematical models used for controller design to be of low fidelity. Thus, the desired controller must be intelligent enough to self-correct and react appropriately to changing circumstances such as a combination of time delays, parametric and disturbance uncertainties using both online information and prior knowledge of the system and environment. The biggest challenge comes from how to develop theoretically provable learning algorithms under a generic architecture that can handle the above system complexities while accomplishing the desired tracking performance.

This research focuses on the development of a systematic and integrated control design methodology, for high-precision, fault tolerant, intelligent systems under the presence of parameter uncertainties, model disturbances, and the delayed actuator input to track a desired reference trajectory. In fact, we rarely meet a system in reality (e.g., fuel-to-air ratio (FAR) control in automotive systems, stabilization of the fingers of an underwater robot manipulator) without any of the above "annoying features". The above features of the NCS are summarized in the below Figure 1.

General Networked Control System

 

<Figure 1: General Networked Control System>