The precise positional controls of piezoelectric actuators (PEA) are problematic due to highly-nonlinear hysteresis behavior which is inherent in piezoelectric materials. In existing PEA positional control applications that are based only on neural networks, the obtained control response results are insufficient for practical usage. In this paper we apply a combined approach by using a feedforward neural network (FNN) jointly with a BAT search algorithm in order to improve the positional control of an X-PEA mechanism model by also taking into account the hysteresis behavior. The proposed positional controller was successfully implemented and it was capable of significantly improving the overall control response result of an X-PEA mechanism model by minimizing the overshoot value and steady-state error, and decreasing the settling time. In addition, the BAT search algorithm can also be used for training the FNN, optimizing the FNN topology and reducing the computational complexity. The presented simulation results confirmed that the proposed positional controller with combined approach provides better results compared to the classical FNN control approach.
COBISS.SI-ID: 18662422
This paper presents the synthesis of an optimal robust controller design using the polynomial pole placement technique and multi-criteria optimisation procedure via an evolutionary computation algorithm (ECG) – differential evolution (DE). The main idea of the design is to provide a reliable fixed-order robust controller structure and an efficient closed-loop performance with a preselected nominally characteristic polynomial. The multi-criteria objective functions have quasi-convex properties that significantly improve convergence and the regularity of the optimal/suboptimal solution. The fundamental aim of the proposed design is to optimise those quasi-convex functions with fixed closed-loop characteristic polynomials, the properties of which are unrelated and hard to present within formal algebraic frameworks. The objective functions are derived from different closed-loop criteria, such as robustness with metric H∞, time performance indexes, controller structures, stability properties etc. Finally, the design results from the example verify the efficiency of the controller design and also indicate broader possibilities for different optimisation criteria and control structures.
COBISS.SI-ID: 18589974
We propose a new compressed sensing MRI approach that uses the discrete nonseparable shearlet transform (DNST) as a sparsifying transform and the fast iterative soft thresholding algorithm (FISTA) for reconstruction. FISTA has a simple design and has shown good convergence behavior. The DNST transform has excellent localization properties within the space domain and excellent directional selectivity. We utilize the frequency representation of the DNST canonical dual filters to obtain a memory efficient modified FISTA based algorithm with a simple and efficient way of calculating the update, tuned to the non tight frame DNST transform. The proposed approach shows improved performance and similar execution time when compared with other state of the art reconstruction approaches.
COBISS.SI-ID: 18778390
This paper presents synthetic aperture radar (SAR) image categorization based on feature descriptors within the discrete wavelet transform (DWT) domain using nonparametric and parametric features. The first and second moments, Kolmogorov Sinai entropy and coding gain, are used for the nonparametric features within an oriented dual tree complex wavelet transform (2D ODTCWT). A Gauss%Markov random field (GMRF), triplet Markov random field (TMRF), and autobinomial model (ABM) are used for feature extraction using a parametric approach within an image domain. A single parameter of GMRF, TMRF, or ABM is used for characterizing an entire patch; therefore, higher model orders (MOs) are used. A database with 2000 images representing 20 different classes with 100 images per class is used for estimating classification efficiency. A supervised learning stage is implemented within a support vector machine (SVM) using 10% and 20% of the test images per class. The experimental results showed that the nonparametric features achieved better results when compared to the parametric features.
COBISS.SI-ID: 18252822
This paper presents a particle filter algorithm for distance estimation using multiple antennas on the receiver%s side and only one transmitter, where a received signal strength indicator (RSSI) of radio frequency was used. Two different placements of antennas were considered (parallel and circular). The physical layer of IEEE standard 802.15.4 was used for communication between transmitter and receiver. The distance was estimated as the hidden state of a stochastic system and therefore a particle filter was implemented. The RSSI acquisitions were used for the computation of important weights within the particle filter algorithm. The weighted particles were re-sampled in order to ensure proper distribution and density. Log-normal and ground reflection propagation models were used for the modeling of a prior distribution within a Bayesian inference.
COBISS.SI-ID: 18253078