In this paper, an optimal Takagi-Sugeno fuzzy control is designed with a novel method called combined discrete and continuous action reinforcement learning algorithm (CDCARLA). The proposed method is implemented for a nonlinear system that is Cart-Pole system. Simulation results show that the proposed method has significant performance. The advantage of CDCARLA is that it does not need system dynamics as well as any other information of power system. It can be said that, this method will consider nonlinear features of power system. It is shown that CDCARLA method can be considered as one of the automatic design technique for designing of controller parameters.

Since most of the systems in real world are nonlinear and include uncertainty in their nature, robust controller designing is one of the most important challenges for engineers. Controller designing for such systems is usually complicate with high computational cost. In contrast to this, state feedback controller designing, based on well-known Ackermann’ s formula, has simplicity in designing and application although global states controllability should be considered seriously. The aim of this paper is to design a state feedback controller for nonlinear inverted pendulum with uncertainty which close loop system has global asymptotically stability. For this reason, controllability property for nonlinear systems has been analyzed based on TS-Fuzzy model. In the existence of uncertainty, controllability property might be failed. In this case to handle the uncertainties in the systems, sufficient conditions have been investigated to guarantee the local and global controllability conditions and also global stability conditions. The advantage of this method is simplicity in implementation comparing to other complicated controllers.

This paper proposes robust H∞ output feedback control stabilization for uncertain Takagi– Sugeno (T-S) fuzzy systems via linear matrix inequalities (LMIs). In order to reduce the conservatism associated with T-S fuzzy system, a new form of non-monotonic Lyapunov functions is used. In the non-monotonic approach, the monotonic decrease of the Lyapunov function is relaxed which enables it to increase locally but vanish eventually. Based on the non-monotonic Lyapunov function approach, sufficient conditions for the existence of robust H∞ output feedback control stabilization are derived. The proposed design technique is shown to be less conservative than the existing non-monotonic approach, namely, 𝑘-samples variations of Lyapunov function. The effectiveness of the proposed approach is further illustrated via numerical example.

In this study a new type of Takagi-Sugeno-Kang (TSK) type fuzzy system with dimension reduction section at the input stage called Semi-polynomial data Mapping Fuzzy Inference System (SPMFIS) is proposed. In the proposed method a semi-polynomial feature map is used to transform the input variables to new extracted features with low dimensions. At the next step, these new features are used as the input vector of ANFIS structure. Also gradient descent algorithm is chosen for training parameters of ANFIS and SPM parts of the proposed method. In order to evaluate the capability of the proposed method, its applications in classification of some different benchmark data sets, system identification, and time series prediction have been studied. The results show that the proposed method performs better than the conventional models in classification, identification and time series prediction.

Prediction models and classification algorithms are widely used in many science and technology. Among their various methods, Well-known data-driven methods such as neural networks and neuro-fuzzy models because of their characteristics have been considered by many researchers. To develop and overcome the weak points of these models, the concepts of the human brain biological systems are used. Therefore, the brain's emotional limbic system is used to develop these models. Brain Emotional Learning (BEL) is an emotional artificial neural network based on the interaction of the thalamus, cortex, amygdala, and orbitofrontal components. This learning machine has different architectures and learning algorithms. In this paper, the online fuzzy extreme learning machine is used as the amygdala and orbitofrontal component in the brain emotional learning machine. To interact between the main components of the brain emotional learning machine, online recurrent memory sequential fuzzy extreme learning machine with different memory depth and transfer learning ability is used. The final design machine is called Brain Emotional Learning based on Online Recurrent Memory Sequential Fuzzy Extreme Learning Machine (BEL-ORMS-FELM). The proposed cognitive machine is designed based on learning the training data one-by-one but also chunk-by-chunk (with fi, xed or varying length) and it can discard training data that has already been trained. Performance comparison of the proposed method is done with other similar learning methods on the benchmark problems of chaotic time series. The results of analysis and simulations show that the performance and accuracy of the proposed method are higher than other methods.

1. Introduction Air pollution is a major threat to public health, especially in the metropolises (Kalo et al., 2020). Due to the disadvantages of air pollution, understanding the various aspects of this issue is of great importance. Producing accurate air pollution maps plays an important role in managing and quantifying existing and future health risks (Alimissis et al., 2018). Estimating the spatial distribution of air pollution continuously over a wide geographical area, especially in areas that have not been measured is a major concern in health studies (Masroor et al., 2020). Although spatial interpolation methods have been widely used in various applications to estimate unknown values in unsampled locations, many fundamental problems remain unresolved (Kalo et al., 2020). The superior methods extracted in previous research show that the results obtained in one phenomenon or one area are not extendable to all phenomena and places. Therefore, the evaluation and selection of interpolation techniques play an important role in the spatial zoning of CO pollution. Based on the results presented by Garcí, a-Santos et al. (2020) and given reviewing the methods used in previous research, Inverse Distance Weight (IDW), Kriging (simple, ordinary, and universal), and Radial Base Function (RBF) methods were selected as common and classical methods of evaluation. New interpolation methods including artificial neural networks (ANN) and fuzzy-based methods have been developed in various fields. Alimissis et al. (2018) expressed the ability of ANN in predicting the pollutants of nitrogen dioxide, nitrogen monoxide, carbon monoxide, sulfur, and ozone. ANN and linear interpolations have also been used for daily nitrous oxide measurements (Bigaignon et al., 2020). In performed research, only the temporal forecast of air pollution in each station is considered and no spatial zoning is done. Tutmez and Hatipoglu (2010) compared the Takagi-Sugeno fuzzy method with fuzzy clustering and Universal Kriging in nitrate modeling so that their study demonstrated the superiority of fuzzy methods. Since the spatial distribution of air pollutants is one of the major concerns of Tehran and authorities, the main objective of this research is to evaluate the capability of some proposed methods’,functionality (e. g., ANN and Fuzzy Sugeno by Fuzzy C-means Clustering) along with the common interpolation methods (e. g., IDW, RBF and Simple Kriging (SK), Ordinary Kriging (OK) and Universal Kriging (OK)) in estimating carbon monoxide gas pollution.

In this paper, the design steps of a multi-model adaptive receding horizon controller for a nonlinear dynamic system are investigated. To implement this control structure, the Takagi-Sugno-Kong (TSK) fuzzy inference system (TSK) has been used to predict the behavior of the dynamic system on a receding horizon. In the proposed controller, the linear part of the TSK fuzzy model is used as a linear model to implement a multi-stage receding horizon controller to calculate the optimal control input sequence. A standard least square algorithm is used to identify the rules consequent parameters of the TSK model. A clustering method is used for partitioning the input-output space in order to generate TSK fuzzy model. Each cluster represents a functional area of the complex dynamic system in the input-output space. In the proposed control strategy, it is assumed that the variables which are used in the premise of the rules are also those which are used in linear models that describe the consequents of the rules. For proper control of the nonlinear system, multiple models are used on the receding horizon. In order to evaluate the proposed control strategy, the proposed control structure has been used to control the power of a nuclear reactor in the charge pursuit problem. The simulation results show the good performance of the proposed control structure.