Reinforcement learning in neurofuzzy traffic signal control

A fuzzy traffic signal controller uses simple “if–then” rules which involve linguistic concepts such as medium or long, presented as membership functions. In neurofuzzy traffic signal control, a neural network adjusts the fuzzy controller by fine-tuning the form and location of the membership functions. The learning algorithm of the neural network is reinforcement learning, which gives credit for successful system behavior and punishes for poor behavior; those actions that led to success tend to be chosen more often in the future. The objective of the learning is to minimize the vehicular delay caused by the signal control policy. In simulation experiments, the learning algorithm is found successful at constant traffic volumes: the new membership functions produce smaller vehicular delay than the initial membership functions.     

A self-organizing adaptive fuzzy controller

There are two main parts in this paper. The first part presents a knowledge representation and reasoning scheme, called tree-searched neural networks (TNN). The TNN is based on a well-known intuitive knowledge representation (IKR) and can reduce the number of the processing nodes in the neural networks. The second part proposes a self-organizing adaptive fuzzy controller (SOAFC) based on the TNN model. It can help acquire control knowledge and thus can reduce the dependence on experts. Furthermore, designers do not need to predefine all membership functions to cover whole input space domain. For improving its performance further, we design a D-controller which is included within the SOAFC. Whether the fuzzy controller is incorporated with the D-controller or not, it is also guaranteed to be globally stable. Simulation results show that this approach has faster convergence speed, results in better transient response, and in addition requires less total control energy.     

Neural languages

In order to provide the theoretical framework necessary to study the neural mechanisms underlying languages, we present here a mathematical formalization of some neural behaviors. In such a context: (i) the neuron is defined as a coupling of automata, dealing with the transduction and codifying processes; (ii) the coupling between neurons is measured by a membership relation defined as the ratio between the transmitted and the system entropies; (iii) the attention given to the messages arriving at the system is considered as the difference between the couplings of the excitatory and inhibitory pools of neurons; (iv) the graphs in the neural systems are described by means of these couplings and, finally (v) the semantic productions are described by the fuzzy formal languages accepted by the automata formalized on these graphs. In such a way, both the verbal and neural semantics can now be correlated and experimentally investigated.     

Evolving fuzzy rule based controllers using genetic algorithms

The synthesis of genetics-based machine learning and fuzzy logic is beginning to show promise as a potent tool in solving complex control problems in multi-variate non-linear systems. In this paper an overview of current research applying the genetic algorithm to fuzzy rule based control is presented. A novel approach to genetics-based machine learning of fuzzy controllers, called a Pittsburgh Fuzzy Classifier System # 1 (P-FCS1) is proposed. P-FCS1 is based on the Pittsburgh model of learning classifier systems and employs variable length rule-sets and simultaneously evolves fuzzy set membership functions and relations. A new crossover operator which respects the functional linkage between fuzzy rules with overlapping input fuzzy set membership functions is introduced. Experimental results using P-FCS 1 are reported and compared with other published results. Application of P-FCS1 to a distributed control problem (dynamic routing in computer networks) is also described and experimental results are presented.     

Simulation and evaluation of dual fully fuzzy linear systems by fuzzy neural network

In this paper, a novel hybrid method based on fuzzy neural network for approximate solution of fuzzy linear systems of the form Ax = Bx + d, where A and B are two square matrices of fuzzy coefficients, x and d are two fuzzy number vectors, is presented. Here a neural network is considered as a part of a large field called neural computing or soft computing. Moreover, in order to find the approximate solution, a simple and fast algorithm from the cost function of the fuzzy neural network is proposed. Finally, we illustrate our approach by some numerical examples.     

Fuzzy disk for covering fuzzy points

In this paper, we consider an important fuzzy version of the well known smallest covering circle problem which is also called the Euclidean 1-center problem. Here we are given a set of points in the plane whose locations are crisp. However, the requirement for coverage of each point is fuzzy as represented by its grade of membership. As such we qualify the points as fuzzy. In the above framework, we wish to find a fuzzy disk with minimum fuzzy area for covering the given fuzzy points. After developing a general model, certain special cases are investigated in detail. Polynomial algorithms are presented for the special cases.     

Bisimulations for fuzzy automata

Bisimulations have been widely used in many areas of computer science to model equivalence between various systems, and to reduce the number of states of these systems, whereas uniform fuzzy relations have recently been introduced as a means to model the fuzzy equivalence between elements of two possible different sets. Here we use the conjunction of these two concepts as a powerful tool in the study of equivalence between fuzzy automata. We prove that a uniform fuzzy relation between fuzzy automata A and B is a forward bisimulation if and only if its kernel and co-kernel are forward bisimulation fuzzy equivalence relations on A and B and there is a special isomorphism between factor fuzzy automata with respect to these fuzzy equivalence relations. As a consequence we get that fuzzy automata A and B are UFB-equivalent, i.e., there is a uniform forward bisimulation between them, if and only if there is a special isomorphism between the factor fuzzy automata of A and B with respect to their greatest forward bisimulation fuzzy equivalence relations. This result reduces the problem of testing UFB-equivalence to the problem of testing isomorphism of fuzzy automata, which is closely related to the well-known graph isomorphism problem. We prove some similar results for backward-forward bisimulations, and we point to fundamental differences. Because of the duality with the studied concepts, backward and forward-backward bisimulations are not considered separately. Finally, we give a comprehensive overview of various concepts on deterministic, nondeterministic, fuzzy, and weighted automata, which are related to bisimulations.     

Bifurcations of fuzzy nonlinear dynamical systems

We present some recent developments of the fuzzy generalized cell mapping method (FGCM) in this paper. The topological property of the FGCM and its finite convergence of membership distribution vector are discussed. Powerful algorithms of digraphs are adopted for the analysis of topological properties of the FGCM systems. Bifurcations of fuzzy nonlinear dynamical systems are studied by using the FGCM method. A backward algorithm is introduced to study the unstable equilibrium solutions and their bifurcation. We have found that near the deterministic bifurcation point, the fuzzy system undergoes a complex transition as the control parameter varies. In this transition region, the steady state membership distribution is dependent on the initial condition. If we use the measure and topology of the α-cut (α = 1) of the steady state membership function of the persistent group representing the stable fuzzy equilibrium solution to characterize the fuzzy bifurcation, assuming the uniform initial condition within the persistent group, the bifurcation of the fuzzy dynamical system is then completed within an interval of the control parameter, rather than at a point as is the case of deterministic systems.     

基于神经网络的模糊决策方法

给出用神经网络去处理模糊决策问题的方法,此方法避免了模糊决策计算量大、计算复杂,隶属函数确定带有主观性等问题。     

一种基于神经元的自适应模糊推理网络

本文根据Ｔ－Ｓ模糊模型提出了一种新的基于神经元的自适应模糊推理网络，给出了连接结构和学习算法，它能自动学习和修正隶属函数及模糊规则，将其用于Ｂｏｘ的煤气炉，太阳黑子预报以及降雨量预报等不同类型的复杂系统建模，仿真结果表明，该模糊神经网络具有收敛速度快，辨识精度高，泛化能力强和适应范围广等特点，可当作复杂系统建模的一种有效工具。     

Solving fuzzy queueing decision problems via a parametric mixed integer nonlinear programming method

This paper proposes a mathematical programming method to construct the membership functions of the fuzzy objective value of the cost-based queueing decision problem with the cost coefficients and the arrival rate being fuzzy numbers. On the basis of Zadeh’s extension principle, three pairs of mixed integer nonlinear programs (MINLP) parameterized by the possibility level α are formulated to calculate the lower and upper bounds of the minimal expected total cost per unit time at α, through which the membership function of the minimal expected total cost per unit time of the fuzzy objective value is constructed. To provide a suitable optimal service rate for designing queueing systems, the Yager’s ranking index method is adopted. Two numerical examples are solved successfully to demonstrate the validity of the proposed method. Since the objective value is completely expressed by a membership function rather than by a crisp value, it conserves the fuzziness of the input information, thus more information is provided for designing queueing systems. The successful extension of queueing decision models to fuzzy environments permits queueing decision models to have wider applications in practice.     

A neural fuzzy control system with structure and parameter learning

A general connectionist model, called neural fuzzy control network (NFCN), is proposed for the realization of a fuzzy logic control system. The proposed NFCN is a feedforward multilayered network which integrates the basic elements and functions of a traditional fuzzy logic controller into a connectionist structure which has distributed learning abilities. The NFCN can be constructed from supervised training examples by machine learning techniques, and the connectionist structure can be trained to develop fuzzy logic rules and find membership functions. Associated with the NFCN is a two-phase hybrid learning algorithm which utilizes unsupervised learning schemes for structure learning and the backpropagation learning scheme for parameter learning. By combining both unsupervised and supervised learning schemes, the learning speed converges much faster than the original backpropagation algorithm. The two-phase hybrid learning algorithm requires exact supervised training data for learning. In some real-time applications, exact training data may be expensive or even impossible to obtain. To solve this problem, a reinforcement neural fuzzy control network (RNFCN) is further proposed. The RNFCN is constructed by integrating two NFCNs, one functioning as a fuzzy predictor and the other as a fuzzy controller. By combining a proposed on-line supervised structure-parameter learning technique, the temporal difference prediction method, and the stochastic exploratory algorithm, a reinforcement learning algorithm is proposed, which can construct a RNFCN automatically and dynamically through a reward-penalty signal (i.e., “good” or “bad” signal). Two examples are presented to illustrate the performance and applicability of the proposed models and learning algorithms.     

H∞ estimation for fuzzy membership function optimization

Given a fuzzy logic system, how can we determine the membership functions that will result in the best performance? If we constrain the membership functions to a specific shape (e.g., triangles or trapezoids) then each membership function can be parameterized by a few variables and the membership optimization problem can be reduced to a parameter optimization problem. The parameter optimization problem can then be formulated as a nonlinear filtering problem. In this paper we solve the nonlinear filtering problem using H_∞ state estimation theory. However, the membership functions that result from this approach are not (in general) sum normal. That is, the membership function values do not add up to one at each point in the domain. We therefore modify the H_∞ filter with the addition of state constraints so that the resulting membership functions are sum normal. Sum normality may be desirable not only for its intuitive appeal but also for computational reasons in the real time implementation of fuzzy logic systems. The methods proposed in this paper are illustrated on a fuzzy automotive cruise controller and compared to Kalman filtering based optimization.     

A hybrid intelligent algorithm for portfolio selection problem with fuzzy returns

Portfolio selection theory with fuzzy returns has been well developed and widely applied. Within the framework of credibility theory, several fuzzy portfolio selection models have been proposed such as mean–variance model, entropy optimization model, chance constrained programming model and so on. In order to solve these nonlinear optimization models, a hybrid intelligent algorithm is designed by integrating simulated annealing algorithm, neural network and fuzzy simulation techniques, where the neural network is used to approximate the expected value and variance for fuzzy returns and the fuzzy simulation is used to generate the training data for neural network. Since these models are used to be solved by genetic algorithm, some comparisons between the hybrid intelligent algorithm and genetic algorithm are given in terms of numerical examples, which imply that the hybrid intelligent algorithm is robust and more effective. In particular, it reduces the running time significantly for large size problems.     

Precision of neural timing: The small ? limit

We explore the precision of neural timing in a model neural system with n identical input neurons whose firing time in response to stimulation is chosen from a density f. These input neurons stimulate a target cell which fires when it receives m hits within ? msec. We prove that the density of the firing time of the target cell converges as ?→0 to the input density f raised to the mth and normalized. We give conditions for convergence of the density in L¹, pointwise, and uniformly as well as conditions for the convergence of the standard deviations.     

Neuro-fuzzy technique for navigation of multiple mobile robots

In this paper, navigation techniques for several mobile robots are investigated in a totally unknown environment. In the beginning, Fuzzy logic controllers (FLC) using different membership functions are developed and used to navigate mobile robots. First a fuzzy controller has been used with four types of input members, two types of output members and three parameters each. Next two types of fuzzy controllers have been developed having same input members and output members with five parameters each. Each robot has an array of sensors for measuring the distances of obstacles around it and an image sensor for detecting the bearing of the target. It is found that the FLC having Gaussian membership function is best suitable for navigation of multiple mobile robots. Then a hybrid neuro-fuzzy technique has been designed for the same problem. The neuro-fuzzy technique being used here comprises a neural network, which is acting as a pre processor for a fuzzy controller. The neural network considered for neuro-fuzzy technique is a multi-layer perceptron, with two hidden layers. These techniques have been demonstrated in simulation mode, which depicts that the robots are able to avoid obstacles and reach the targets efficiently. Amongst the techniques developed neuro-fuzzy technique is found to be most efficient for mobile robots navigation. Experimental verifications have been done with the simulation results to prove the authenticity of the developed neuro-fuzzy technique.     

Topological relations between fuzzy regions: derivation of verbal terms

A method is presented to generate verbal terms about topological relations between fuzzy regions. The methodology relies on the fuzzy 4-intersection, which is a generalization of the crisp 4-intersection of Egenhofer and co-workers. The computation of the similarity between the fuzzy- and the crisp 4-intersection enables the verbal term, i.e., the linguistic variable, to be derived. The linguistic variable contains a semantic part which gives an immediate association to a crisp relation and a quantifier which indicates the strength of the relationship. Since the derivation of the linguistic variable depends on the definition of the boundary of the fuzzy regions, a method is presented to compute fuzzy boundaries. The approach here defines fuzzy boundaries so that each point in the fuzzy region is associated a partial membership in both the interior and the boundary of the region. This view is different from the boundary definition in crisp topology, but it agrees with the fuzzy set idea that elements can have partial membership in different sets. A simulation experiment demonstrates the properties of the proposed methodology, and it shows how the linguistic variable relates to an inclusion index. An example illustrates how some level of action can be associated to the linguistic variable, which is applicable in the course control of moving crafts, in military applications or in other kinds of operations where the level of warning or action depends on the topological relation between the fuzzy regions.The findings in this article are applicable to geographical information systems (GIS), the modelling of objects with indeterminate boundaries, in the reasoning about relations between geographical objects, or the evaluation of database queries. If the ideas in the present article are implemented in GIS, this will provide an enhanced user interface compared to most GIS today.     

Determination of the threshold value β of variable precision rough set by fuzzy algorithms

In the past, the choices of β values to be applied to find the β-reducts in VPRS for an information system are somewhat arbitrary. In this study, a systematic method which bridges the fuzzy set methodology and probabilistic approach of RS to solve the threshold value β determination problem in variable precision rough sets (VPRS) is proposed. Different from the existing probabilistic methods, the proposed method relies on the fuzzy membership degrees of each attribute of the objects to calculate β. The proposed method gives the membership degrees and fuzzy aggregation operators the probabilistic interpretations. Based on the probabilistic interpretations, the threshold value β of VPRS is directly derived from fuzzy membership degree by Implication Relations and Fuzzy Algorithms, in which the membership degrees are obtained by the standard Fuzzy C-means method. The argument is that errors of system classification would occur in the fuzzy-clustering phase prior to information classification, therefore the threshold value β should be constrained by the probability of belongingness of an object to the fuzzy clusters, i.e., through the values of membership functions. A few examples are given in the paper to demonstrate the differences with other β-determining methods.     

Clifford Fuzzy Support Vector Machines for Classification

A Clifford support vector machine (CSVM) learns the decision surface from multi distinct classes of the multiple input points using the Clifford geometric algebra. In many applications, each multiple input point may not be fully assigned to one of these multi-classes. In this paper, we apply a fuzzy membership to each multiple input point and reformulate the CSVM for multiclass classification to make different input points have their own different contributions to the learning of decision surface. We call the proposed method Clifford fuzzy SVM.     

Interval optimization based line sampling method for fuzzy and random reliability analysis

For structural system with fuzzy variables as well as random variables, a novel algorithm for obtaining membership function of fuzzy reliability is presented on interval optimization based Line Sampling (LS) method. In the presented algorithm, the value domain of the fuzzy variables under the given membership level is firstly obtained according to their membership functions. Then, in the value domain of the fuzzy variables, bounds of reliability of the structure are obtained by the nesting analysis of the interval optimization, which is performed by modern heuristic methods, and reliability analysis, which is achieved by the LS method in the reduced space of the random variables. In this way the uncertainties of the input variables are propagated to the safety measurement of the structure, and the membership function of the fuzzy reliability is obtained. The presented algorithm not only inherits the advantage of the direct Monte Carlo method in propagating and distinguishing the fuzzy and random uncertainties, but also can improve the computational efficiency tremendously in case of acceptable precision. Several examples are used to illustrate the advantages of the presented algorithm.