Model-Based Reinforcement Learning Control of Electrohydraulic Position Servo Systems

Yao, Zhikai; Liang, Xianglong; Jiang, Guoping; Yao, Jianyong

doi:10.1109/tmech.2022.3219115

Cited by 39 publications

(22 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Additionally, M and M −1 d are diagonal matrices with J 2 = 0. If ℵ satisfies (17) and also (18) with the following parameters…”

Section: Resultsmentioning

confidence: 99%

“…Hence, according to the knowledge of the authors, adaptive IDA-PBC with respect to dynamical parameters and the parameters of the input mapping matrix have not been proposed in the literature due to the difficulty of the problem. Note that indirect adaptive control, which is based on the persistence of excitation condition or other similar conditions [17], or reinforcement learning method [18], are excluded, and the aim is ensuring stability without these conditions.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Adaptive total energy shaping of a class of manipulators

Harandi,

Taghirad

2024

IET Control Theory & Appl

View full text Add to dashboard Cite

Point‐to‐point control of underactuated mechanical systems is a challenging problem, especially if some of the system's parameters are uncertain. A popular method to stabilize underactuated manipulators is total energy shaping, in which a new Hamiltonian function is assigned to the closed‐loop system while a set of partial differential equations (PDEs) should be solved. In this study, an adaptive total energy shaping controller based on the approach called interconnection and damping assignment passivity‐based control (IDA‐PBC) for some underactuated robots is designed. The proposed method can overcome uncertainties in the dynamical parameters and also the parameters of the input mapping matrix under the satisfaction of the PDEs. The stability of the desired pose is ensured by the Lyapunov approach via suitable design of adaptation laws without the requirement for the persistence of excitation condition or other similar conditions. The results are applied to a 3‐DOF underactuated manipulator and verified by simulations.

show abstract

“…Additionally, M and M −1 d are diagonal matrices with J 2 = 0. If ℵ satisfies (17) and also (18) with the following parameters…”

Section: Resultsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Adaptive total energy shaping of a class of manipulators

Harandi,

Taghirad

2024

IET Control Theory & Appl

View full text Add to dashboard Cite

show abstract

“…For the future works, in addition to topics like observerbased control [38], [39], reinforcement learning control [40], [41], state/output formation control of multiple quadrotors subjected to parameter uncertainties, external disturbances, communication quantization errors, partially known external reference signals, prescribed performance, etc., is of particular interests (see [42] - [47] and references therein). Despite their achievements, however, most of these works used the thrust force and motor torques as control inputs and rely on TSS method incorporating the LBUA for parameter estimation.…”

Section: Discussionmentioning

confidence: 99%

Robust Adaptive Backstepping Control of Quadrotors With Unknown Input Gains

Huang

Jiang

2023

IEEE Access

View full text Add to dashboard Cite

Robust adaptive tracking control design for quadrotors with unknown parameters, including the thrust and the drag factors, is presented. Firstly, the adaptive linearizing control approach is employed for conquering the issues of unknown input gains in both the outer-and inner-loop design steps. Secondly, based on the barrier Lyapunov function (BLF) method, a robust virtual controller is constructed in the intermediate design step for counteracting the coupling nonlinearity and speeding up the convergence at the same time. Thirdly, the dynamic surface control (DSC) technique is applied for alleviating the measurement of acceleration signals and explosion of complexity problem simultaneously. Last, the serial-parallel identification model (SPIM) based composite update algorithms are incorporated for further improving the tracking performance. In particular, the prediction errors are ensured to converge to the vicinity of zero without relying on the parameter convergence. Simulation studies demonstrating its validity are carried out in the final.INDEX TERMS Composite update algorithm, quadrotors, robust adaptive, unknown input gains.

show abstract

“…To suppress these adverse effects, large control gains should be set to provide sufficient control effort, which may cause oscillations [29,30]. An effective approach to overcome this problem is to design composite control schemes utilizing fuzzy logic systems [31,32], neural networks [33], reinforcement learning [34,35], and disturbance observers for disturbance compensation. Among these techniques, the extended state observer (ESO), which can estimate the lumped disturbance with little model information, has been successfully implemented in various fields.…”

Section: Introductionmentioning

confidence: 99%

Enhanced command filtered control with extended state observer for dual‐motor servo mechanisms considering backlash and friction

Wang,

Liu,

Zhang

et al. 2024

IET Control Theory & Appl

View full text Add to dashboard Cite

A dual‐motor servo mechanism is a high‐order and strong‐coupling system with unknown nonlinearity, which brings challenges to controller design to realize high‐performance tracking and synchronization. This article proposes a finite‐time command filtered control strategy to address this problem. In tracking control design, finite‐time command filters are adopted to obtain the derivatives of virtual controllers, and the “explosion of complexity” problem in backstepping is thus solved. An improved compensation system is designed to reduce filtering errors. The tracking control is developed based on the combination of finite‐time control and command filtered approach. Moreover, to deal with unknown nonlinearities and uncertainties, fast finite‐time extended state observers are developed to observe lumped disturbances for the load and motor, respectively. In synchronization control design, two opposite control actions are developed using the speed difference to force two motors to rotate synchronously. The finite‐time convergence of the tracking and synchronization errors is proved. The efficiency of the proposed control strategy is verified via experiments conducted on a dual‐motor servo turntable.

show abstract

Model-Based Reinforcement Learning Control of Electrohydraulic Position Servo Systems

Cited by 39 publications

References 36 publications

Adaptive total energy shaping of a class of manipulators

Adaptive total energy shaping of a class of manipulators

Robust Adaptive Backstepping Control of Quadrotors With Unknown Input Gains

Enhanced command filtered control with extended state observer for dual‐motor servo mechanisms considering backlash and friction

Contact Info

Product

Resources

About