Reinforcement learning based adaptive PID controller design for control of linear/nonlinear unstable processes

Shuprajhaa, T; Sujit, Shivakanth; Srinivasan, K.

doi:10.1016/j.asoc.2022.109450

Cited by 45 publications

(12 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The genetic algorithm is an ordinary but mighty method, and it improves local optimisation capability as a part of the black tern algorithm. Since a good equilibrium between development and utilisation is important for all intelligent models, this research fuses black tern algorithm with the genetic algorithm to enhance local optimisation capability, improve optimisation productivity, and keep the diversity in subsequent running procession [19]. Genetic black tern algorithm combines the global optimisation performance of the black tern algorithm in terms of algorithm performance, so that it has obvious advantages in the ability to search in a wide scope; Simultaneously merits of genetic algorithm are used in local convergence, which makes it possible to avoid the feasibility of local optimisation and enhance ability of local optimisation.…”

Section: Training Algorithm Of Wnnmentioning

confidence: 99%

Retracted: Research on wavelet neural network PID control of maglev linear synchronous motor

Mao

2022

IET Circuits, Devices & Syst

View full text Add to dashboard Cite

Retraction: [Jun Mao, Jun Ma, Research on wavelet neural network PID control of maglev linear synchronous motor, IET Circuits, Devices & Systems 2022 (https://doi.org/10.1049/cds2.12136)]. The above article from IET Circuits, Devices & Systems, published online on 8 December 2022 in Wiley Online Library (wileyonlinelibrary.com), has been retracted by agreement between the Editor‐in‐Chief, Harry E. Ruda, the Institution of Engineering and Technology (the IET) and John Wiley and Sons Ltd. This article was published as part of a Guest Edited special issue. Following an investigation, the IET and the journal have determined that the article was not reviewed in line with the journal’s peer review standards and there is evidence that the peer review process of the special issue underwent systematic manipulation. Accordingly, we cannot vouch for the integrity or reliability of the content. As such we have taken the decision to retract the article. The authors have been informed of the decision to retract.

show abstract

Section: Training Algorithm Of Wnnmentioning

confidence: 99%

Retracted: Research on wavelet neural network PID control of maglev linear synchronous motor

Mao

2022

IET Circuits, Devices & Syst

View full text Add to dashboard Cite

show abstract

“…Using the derived integral control input in (22), the derivative of the Lyapunov candidate function with respect to time t in ( 21) can be rewritten as follows.…”

Section: Stability Analysismentioning

confidence: 99%

“…They showed that the proposed controller performed better than other controllers in terms of low tracking errors and low control effort. Shuprajhaa et al [22] developed an adaptive PID controller for controlling open-loop unstable processes based on generic data-driven modified proximal policy-optimization reinforcement learning. The main feature of the proposed control scheme is that it can eliminate the need for process modeling, as well as prerequisite knowledge of system dynamics and control parameter tuning.…”

Section: Introductionmentioning

confidence: 99%

Data-Driven Adaptive Steady-State-Integral-Derivative Controller Using Recursive Least Squares With Performance Conditions

Lee¹,

2023

IEEE Access

View full text Add to dashboard Cite

This paper presents a data-driven adaptive steady state-integral-derivative (SS-ID) control algorithm that uses gradient descent and recursive least squares (RLS) with a forgetting factor. A simplified first-order differential equation of the control system was designed and its parameters were estimated in real-time using the RLS algorithm. The steady-state control input for target-state tracking was derived based on the estimated parameters and steady-state performance conditions. The gradient of the integrated control error to the gain was estimated based on the least-squares method, using the saved past error and gain data in a finite sliding window to determine the control input. The integral gain was adapted based on the gradient descent method, using the estimated gradient, integrated error, and adaptation rate. Simplified control error dynamics were designed, and their parameter was estimated using the RLS algorithm. The derivative control gain can be adapted in real time using the estimated parameters from the simplified control error dynamics and time constant-based performance conditions. The proposed controller was designed in the MATLAB/Simulink environment. A performance evaluation was conducted under various scenarios using a DC motor simulation model and an actual test platform equipped with an optical encoder.

show abstract

“…After an adaptive PID controller is trained, its parameters change according to the changing state, achieving impressive performance. [29][30][31][32] In a previous study, 33) we combined DRL with a PID controller by integrating the A2C algorithm into the PID framework.The DRL-PID controller has been tested in the simulation environment and the actual machine with simple stepwise responses. The present study measures and analyzes the USLM's dynamic characteristics and extends the application of the DRL-PID controller for both position control and speed control, showing dynamic, versatile, and self-adjusting capabilities.…”

Section: Introductionmentioning

confidence: 99%

Dynamic characteristics and deep reinforcement learning of proportional-integral-differential controller for quadruped stator-based ultrasonic linear motor

Jiang,

Wang,

Sasamura

et al. 2024

Jpn. J. Appl. Phys.

View full text Add to dashboard Cite

Ultrasonic linear motors have many advantages, such as small size, light weight, and a lack of electro-magnetic interference. In our previous study, a quadruped stator-based ultrasonic linear motor that mimics the motion of an inchworm based on longitudinal and bending modes was proposed. However, our previous studies focused on voltage control, overlooking the influence of phase, which made accurate and sustainable operation difficult. This study measures the dynamic characteristics of this motor using polymer-based rods and combines deep reinforcement learning (DRL) with proportional-integral-differential (PID) algorithms for both position and speed control through phase manipulation. The proposed DRL-PID controller outperforms voltage control in both simulation and real-world experiments.

show abstract

Reinforcement learning based adaptive PID controller design for control of linear/nonlinear unstable processes

Cited by 45 publications

References 28 publications

Retracted: Research on wavelet neural network PID control of maglev linear synchronous motor

Retracted: Research on wavelet neural network PID control of maglev linear synchronous motor

Data-Driven Adaptive Steady-State-Integral-Derivative Controller Using Recursive Least Squares With Performance Conditions

Dynamic characteristics and deep reinforcement learning of proportional-integral-differential controller for quadruped stator-based ultrasonic linear motor

Contact Info

Product

Resources

About