Bi‐criteria optimal control of redundant robot manipulators using LVI‐based primal‐dual neural network

Cai, Binghuang; Zhang, Yunong

doi:10.1002/oca.897

Cited by 37 publications

(19 citation statements)

References 53 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We see that the computation of the LASSO solution is a Quadratic Programming (QP) problem. The QP problem could be solved readily using the MATLAB routine “QUADPROG” [24], or solved preferably by using neural networks [25,26]. Note that when s is large enough (i.e., λ = 0), LASSO is multiple linear least squares regression; when s ≥ 0 (i.e., λ > 0) is a smaller value, LASSO solutions are shrunken versions of the least squares estimates [19].…”

Section: Methodsmentioning

confidence: 99%

A novel artificial neural network method for biomedical prediction based on matrix pseudo-inversion

Cai

Jiang

2014

Journal of Biomedical Informatics

Self Cite

View full text Add to dashboard Cite

Biomedical prediction based on clinical and genome-wide data has become increasingly important in disease diagnosis and classification. To solve the prediction problem in an effective manner for the improvement of clinical care, we develop a novel Artificial Neural Network (ANN) method based on Matrix Pseudo-Inversion (MPI) for use in biomedical applications. The MPI-ANN is constructed as a three-layer (i.e., input, hidden, and output layers) feed-forward neural network, and the weights connecting the hidden and output layers are directly determined based on MPI without a lengthy learning iteration. The LASSO (Least Absolute Shrinkage and Selection Operator) method is also presented for comparative purposes. Single Nucleotide Polymorphism (SNP) simulated data and real breast cancer data are employed to validate the performance of the MPI-ANN method via 5-fold cross validation. Experimental results demonstrate the efficacy of the developed MPI-ANN for disease classification and prediction, in view of the significantly superior accuracy (i.e., the rate of correct predictions), as compared with LASSO. The results based on the real breast cancer data also show that the MPI-ANN has better performance than other machine learning methods (including support vector machine (SVM), logistic regression (LR), and an iterative ANN). In addition, experiments demonstrate that our MPI-ANN could be used for bio-marker selection as well.

show abstract

Section: Methodsmentioning

confidence: 99%

A novel artificial neural network method for biomedical prediction based on matrix pseudo-inversion

Cai

Jiang

2014

Journal of Biomedical Informatics

Self Cite

View full text Add to dashboard Cite

show abstract

“…x As for the above QP problem (30)-(31), according to the previous work [30,31], it can be converted into the following piecewise-linear projection equation (PLPE):…”

Section: Mke-type Zhang Equivalency (Mke-ze)mentioning

confidence: 99%

“…where  Ω (⋅) is a piecewise-linear projection operator [25,30,31], and the primal-dual decision variable vector y is defined as y = [x T , u T ] T ∈ R n+m , with u ∈ R m denoting the dual decision vector defined for equality constraint (31). In addition, the augmented coefficient matrix M ∈ R (n+m)×(n+m) and vector q ∈ R n+m are defined, respectively, as…”

Section: Mke-type Zhang Equivalency (Mke-ze)mentioning

confidence: 99%

“…In recent years, many researches have shown that the redundancy-resolution problem can be solved favorably in another manner via online optimization techniques [19,20,25,[29][30][31][32][33][34]. Such optimization techniques are generally expressed and unified as a quadratic programming (QP) formulation, which can elegantly incorporate inequality-based constraints and avoid Jacobian matrix pseudoinversion [30][31][32]. The resultant QP problem can be solved by many approaches, such as dual neural networks [32][33][34], LVI-based primal-dual neural networks [29,30], and numerical algorithms [31].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Analysis, Verification and Comparison on Feedback‐Aided MA Equivalence and Zhang Equivalency of Minimum‐Kinetic‐Energy Type for Kinematic Control of Redundant Robot Manipulators

Qiu

Zhang

Yang

2017

Asian Journal of Control

Self Cite

View full text Add to dashboard Cite

By following and generalizing Ma et al.'s inspiring work and Zhang et al.'s inspiring work, this paper proposes and investigates three kinds of equivalent relationships with error-feedback information incorporated for the minimum-kinetic-energy (MKE) redundancy resolution. Specifically, these three equivalent relationships are divided into two big classes in terms of scheme formulations: one is termed MKE-type Ma equivalence (MKE-ME); the other is MKE-type Zhang equivalency (MKE-ZE), which can be further subdivided into MKE-ZE-I and MKE-ZE-II. Also, the equivalent analyses as well as the primary distinctions about three such equivalent relationships are presented. The simulation results based on a PUMA560 robot manipulator and a three-link planar robot manipulator not only verify the efficacy of the corresponding velocity-level and acceleration-level schemes of MKE-ME, MKE-ZE-I and MKE-ZE-II, but also substantiate their reasonableness and effectiveness of these three equivalent relationships. More importantly, the comparative numerical results show their respective characteristics and advantages for kinematic control of redundant robot manipulators.

show abstract

“…Relatively, the online methods could be used to conquer the defects of the pseudo inverse based methods. In the online scheme, the equality and inequality constraints in the joint-velocity level [17,18] or in the acceleration-level [19,20] could be considered effectively. These schemes are generally configured as a quadratic program (QP) [20,21].…”

Section: Introductionmentioning

confidence: 99%

Online optimization scheme with dual-mode controller for redundancy-resolution with torque constraints

Fang

Zhao

Mei

et al. 2016

Robotics and Computer-Integrated Manufacturing

View full text Add to dashboard Cite

In this paper, an online optimization scheme is proposed for the real time path-tracking control of the redundant robot manipulator. In the proposed scheme, the inequality constraints are extended to the torque level to avoid the torque saturation. Besides, a dual-mode optimal controller is used to yield a feasible input during each control period, to resolve the solution space decreasing problem. Then the scheme is formulated as a Quadratic Program (QP), the computationally efficient Knitro-Based solver is applied to remedy the inescapable redundancy-resolution problem. Furthermore, the comparisons based on the different path tracking simulations between the proposed method and the velocity-level schemes demonstrate that the former is safer and more applicable. The experiment on a physical robot system further verifies the physical realizability of the proposed method.

show abstract

Bi‐criteria optimal control of redundant robot manipulators using LVI‐based primal‐dual neural network

Cited by 37 publications

References 53 publications

A novel artificial neural network method for biomedical prediction based on matrix pseudo-inversion

A novel artificial neural network method for biomedical prediction based on matrix pseudo-inversion

Analysis, Verification and Comparison on Feedback‐Aided MA Equivalence and Zhang Equivalency of Minimum‐Kinetic‐Energy Type for Kinematic Control of Redundant Robot Manipulators

Online optimization scheme with dual-mode controller for redundancy-resolution with torque constraints

Contact Info

Product

Resources

About