Evolutionary Learning of a Fuzzy Behavior Based Controller for a Nonholonomic Mobile Robot in a Class of Dynamic Environments

This paper presents an approach for evolving optimum behaviors for a nonholonomic mobile robot in a class of dynamic environments. A new evolutionary algorithm reflecting some powerful features in the natural evolutionary process to have flexibility to deal with changes in the environment is used to evolve optimum behaviors. Furthermore, a fuzzy set based multi-objective fitness evaluation function is adopted in the evolutionary algorithm. The multi-objective evaluation function is designed so that it allows incorporating complex linguistic features that a human observer would desire in the behaviors of the mobile robot movements. To illustrate the effectiveness of the proposed method, simulation results are compared using a conventional evolutionary algorithm.     

Variational Bayes solution of linear neural networks and its generalization performance

It is well known that in unidentifiable models, the Bayes estimation provides much better generalization performance than the maximum likelihood (ML) estimation. However, its accurate approximation by Markov chain Monte Carlo methods requires huge computational costs. As an alternative, a tractable approximation method, called the variational Bayes (VB) approach, has recently been proposed and has been attracting attention. Its advantage over the expectation maximization (EM) algorithm, often used for realizing the ML estimation, has been experimentally shown in many applications; nevertheless, it has not yet been theoretically shown. In this letter, through analysis of the simplest unidentifiable models, we theoretically show some properties of the VB approach. We first prove that in three-layer linear neural networks, the VB approach is asymptotically equivalent to a positive-part James-Stein type shrinkage estimation. Then we theoretically clarify its free energy, generalization error, and training error. Comparing them with those of the ML estimation and the Bayes estimation, we discuss the advantage of the VB approach. We also show that unlike in the Bayes estimation, the free energy and the generalization error are less simply related with each other and that in typical cases, the VB free energy well approximates the Bayes one, while the VB generalization error significantly differs from the Bayes one.     

Resource-Bounded Measure and Learnability

We consider the resource-bounded measure of polynomial-time learnable subclasses of polynomial-size circuits. We show that if EXP ≠ MA, then every PAC-learnable subclass of P/poly has EXP-measure zero. We introduce a nonuniformly computable variant of resource-bounded measure and show that, for every fixed polynomial q , any polynomial-time learnable subclass of circuits of size q has measure zero with respect to P/poly. We relate our results to the question of whether the class of Boolean circuits is polynomial-time learnable. Received July 15, 1998, and in final form September 9, 1999.     

Pattern recognition as conceptual morphogenesis

Pattern recognition is our mental activity in which we formulate, select, modify, and adjust our concepts or our frames of reference so that we can see a ``form' in objects. We can tentatively formulate this viewpoint as the principle of minimum entropy (or entropy-like function). Different algorithms in clustering and pattern recognition are given new interpretations in the light of this unifying principle.     

On intractability of the classUP

The classUP [V] is the class of sets accepted by polynomial-time nondeterministic Turing machines which have at most one accepting path for every input. The complexity of this class closely relates to that of computing inverses ofone-way functions, where a one-way function is a one-to-one, length-increasing, and polynomial-time computable function whose inverse cannot be computed within polynomial time. It is known [GS], [K] that there exists a one-way function if and only ifP UP. In this paper the intractability of sets inUP is investigated in terms of polynomial-time reducibility to a sparse set. It is shown thatUP has a set that is _m ^P -reducible to no sparse set ifP UP. We interpret this structural property in the relation with approximation algorithms: it is shown that ifP UP, thenUP has a set with no 1-APT approximation and, furthermore,UP has a set that is not _m ^P -reducible to any set with a 1-APT approximation. The implication of this result in the study of one-way functions is also discussed. In order to prove the main theorem, we introduce a variation of tree-pruning methods.This paper was written while the author visited the Department of Mathematics, University of California, Santa Barbara. This research was supported in part by the National Science Foundation under Grant CCR-8611980.     

Multiple pulse-heating experiments with different current to determine total emissivity, heat capacity, and electrical resistivity of electrically conductive materials at high temperatures

A modified pulse-heating method is proposed to improve the accuracy of measurement of the hemispherical total emissivity, specific heat capacity, and electrical resistivity of electrically conductive materials at high temperatures. The proposed method is based on the analysis of a series of rapid resistive self-heating experiments on a sample heated at different temperature rates. The method is used to measure the three properties of the IG-110 grade of isotropic graphite at temperatures from 850 to 1800 K. The problem of the extrinsic heating-rate effect, which reduces the accuracy of the measurements, is successfully mitigated by compensating for the generally neglected experimental error associated with the electrical measurands (current and voltage). The results obtained by the proposed method can be validated by the linearity of measured quantities used in the property determinations. The results are in reasonably good agreement with previously published data, which demonstrate the suitability of the proposed method, in particular, to the resistivity and total emissivity measurements. An interesting result is the existence of a minimum in the emissivity of the isotropic graphite at around 1120 K, consistent with the electrical resistivity results.     

Bubble wake visualization by using photochromic dye

Wake structure of a single “clean bubble”, rising in rectilinear, zigzagging or spiraling path, is experimentally investigated. A single nitrogen gas bubble was produced in a silicone oil pool and the wake structure development in the rear of the rising bubble was visualized by using photochromic dye. The flexibility of this visualization method enabled us to distinguish wake from drift easily. Both bubble motion and wake structure were recorded by using stereo high-speed video camera simultaneously. We present the first experimental support for the existence of the standing eddy at the rear of the clean bubble, as predicted by a previous numerical study by Ryskin and Leal [1984. Journal of Fluid Mechanics 148, 19-35], Dandy and Leal [1986. Physics of Fluids 29(5), 1360-1366] and Blanco and Magnaudet [1995. Physics of Fluids 7(6), 1265-1274]. We study motion of a pair of vortex filaments, which is called double-threaded type wake, in the case of bubble rising in an axi-asymmetric path. Visualization results of multiple formations of horse-shoe type vortices in one period of zigzag motion of rising bubble with shape oscillations, which has not been observed in previous studies are also presented.     

Aerosol Synthesis of Inhalation Particles via a Droplet-to-Particle Method

Inhalation powders with consistent particle properties, including particle size, size distribution, and shape were produced with an aerosol synthesis method. Compared to conventional spray drying, the aerosol method provides better control of the thermal history and residence time of each droplet and product particle due to the laminar flow in the heated zone of the reactor where the droplet drying and particle formation take place. A corticosteroid, beclomethasone dipropionate, generally used for asthma treatment was chosen as a representative material to demonstrate the process. Spherical particles were produced with a droplet-to-particle method from an ethanolic precursor solution. The droplets produced with an ultrasonic nebulizer were carried to a heated zone of the reactor at 50-150°C where the solvent was evaporated and dry particles formed. The mass mean diameter of the particles were well within the respirable size range (approximately 2 μm). The geometric standard deviation (GSD) of produced particles was approximately 2. The particle surface structure varied from smooth to rough depending on the degree of particle crystallinity and was affected by the thermal history of the particle. Amorphous particles with smooth surface were most likely obtained due to the rapid evaporation of the solvent from the droplet combined with the slow diffusion of the beclomethasone dipropionate molecule. The amorphous particles were transformed slowly to crystalline particles in the open atmosphere. In addition, the particle surface structure changed from smooth to rough during storage. The process was accelerated by thermal post-annealing. However, additional heating also increased particle sintering. By optimizing the reactor parameters, and thus increasing the molecular diffusion, stable, crystalline particles were produced at 150°C.