Iterative learning control for uncertain nonlinear networked control systems with random packet dropout

Abstract

This paper proposes a novel networked iterative learning control (NILC) scheme with adjustment factor for a class of discrete-time uncertain nonlinear systems with stochastic input and output packet dropout modeled as 0-1 Bernoulli-type random variable. Firstly, the equivalence relation between the realizability of controlled system and the input-output coupling parameter (IOCP) is established. Secondly, in order to overcome the main obstacle arising from the unknown IOCP, an identification technique is developed for it. Thirdly, it is strictly proved that, under certain conditions, the tracking errors driven by the developed NILC scheme are convergent to zero along iteration direction in the sense of expectation. Finally, an example is given to demonstrate the effectiveness of the proposed NILC scheme and the merits of adjustment factor. KEYWORDS iterative learning control, networked control systems, stochastic packet dropouts, uncertain nonlinear systems, unknown input-output coupling parameter Int J Robust Nonlinear Control. 2019;29:3529-3546.wileyonlinelibrary.com/journal/rnc

Iterative learning control with discrete‐time nonlinear nonminimum phase models via stable inversion

“…Relevant applications include robotassisted stroke rehabilitation 1 , high speed train control 2 , laser additive manufacturing 3 , and vehicle-mounted manipulators 4 , all of which use nonlinear models. In fact, while the majority of ILC literature focuses on linear systems, the prevalence of nonlinear dynamics in real-world systems has motivated the development of numerous ILC theories for discrete-time nonlinear models [5][6][7][8][9][10] .…”

“…Additionally, Newton's method has been shown to deliver faster convergence in ILC than schemes such as P-type ILC 24, ch. 5 , upon which much of the relevant prior art on the ILC of discrete-time nonlinear systems is founded [5][6][7][8][9] . However, this work demonstrates that when synthesized from models with unstable inverses, i.e.…”

“…However, published discrete‐time‐nonlinear‐model‐based ILC theories exclude at least one of properties (P1)–(P4) from consideration. While the prior art makes important contributions such as foundational nonlinear ILC theory, 5‐7 relaxation of process repetitiveness assumptions, 8 robustness to packet dropout in measurement and controller signals, 9 and integration of ILC with adaptive control, 10 these studies' analyses are limited to specific system structures. As a consequence, (P1) is not addressed by References 6,7,9, (P2) is not addressed by References 5–7,9,10, (P3) is not addressed by References 5,8, and (P4) is not addressed by References 5–9.…”

Reference input and hysteresis quantizer based triggered control for networked control systems over limited channels

“…Networked control systems (NCSs), with the spatial distribution structure of the sensor, actuator, and controller, are about to replace the traditional point-to-point control systems for its superiority in design, installation and maintenance. [1][2][3][4] It has been widely applied to industrial processes such as unmanned aerial vehicles, 5 mobile sensor networks, 6 waste water treatment, 7 and remote surgery. 8 However, the utilization of communication network brings several challenging issues.…”

Open-Closed-Loop Iterative Learning Control with the System Correction Term for the Human Soft Tissue Welding Robot in Medicine

“…e history of iterative learning control [34,35] can be traced back to 1984, when Arimoto proposed the learning rules of P, PI, and PID and successfully applied them to the motion control of robots. After that, closed-loop and openloop combination learning control laws appeared.…”