This paper proposes a new method of genetic algorithms (GAs) for discrete optimization problems. For continuous optimization problems, it has been reported that distributed genetic algorithms (DGAs) show the higher performance than conventional GAs. However, for discrete optimization problems, the performance of DGAs has not been clear so far. In this paper, we propose a new approach in DGAs to discrete optimization problems. The proposed method is based on the multiple crossovers applied to the population consists of offsprings from elite individuals in distributed subpopulations (Centralized Multiple Crossover : CMX). We examine the performence of a conventional GA, DGA and proposed method for a typical discrete optimization problem, the Traveling Salesman Problem (TSP). The experiments showed that the proposed method provides better performance than the conventional DGA.
Recently, World Wide Web plays an essential role in public and business applications such as information publication and electronic commerce, and its reliability is strongly required. Mark-based authentication has therefore started. It evaluates web sites and issues authentication marks to the sites. The marks are displayed on web pages of the sites, and users of the sites can judge their reliability by seeing the marks. This paper describes a web site authentication mark system which realizes generation, presentation and verification of marks. Especially the paper describes : (1) requirements for the system concerning security, usability, response speed, cost, and extensibility, (2) mark verification by digital signatures embedded in marks and by online queries, (3) implementation of verification programs as plug-ins of web browsers, (4) version management by embedded version IDs in marks, (5) life cycle management of marks using databases, (6) examples of the system behavior. The paper also explains reality of the proposed system through evaluations and comparison with previous systems.
In recent years, the integrated optimization of planning and scheduling from customer order management to delivery has been required from the viewpoint of Supply Chain Management. However, the simultaneous optimization model becomes increasingly complicated and often very difficult to be solved with the increase of the number of combinatorial alternatives. In this paper, we propose a decentralized supply chain optimization method for single stage production systems in which the total decision variables are optimized by solving several sub-problems consisting of Material Requirement Planning, Scheduling and Distribution Planning. A supply chain planning problem for single production system is solved by the proposed method and a hierarchical planning method. Numerical results show that the proposed method generates better solutions than the conventional method for the problems in which the planning decisions are relatively concerned with scheduling decisions.
Nonlinear optimal control using Monte-Carlo calculation of path integrals is proposed. Wave functions appearing in a quantum mechanical theory of nonlinear optimal control are represented as superposition of various paths connecting initial and final points. A transformation of a characteristic designer's constant HR to a pure imaginary value, HR=iHR is applied to these path integrals. Wave functions are then calculated as statistical mean values under the Boltzmann distributions with temperatures proportional to HR, the transformed values of the designer, s constant. Monte-Carlo methods enforced by Metropolis algorithm are then applied to evaluate wave functions and optimal control calculations. Validities of proposed scheme will be shown in simple systems with 1-input and 1-state variables.
This paper gives relaxation methods reducing a feasibility or optimization problem under a parametrized linear matrix inequality constraint to a finite number of LMIs constraint. The methods are based on convexification of difference of convex (d.c.) and multiconvexification. We propose a generalized relaxation method of d.c. convexification and a unified relaxation method between d.c convexification and multiconvexification techniques. These methods are applied to stability analysis and L2 gain analysis of parameter-dependent systems. Numerical examples are illustrated for each applications.
With a view to attaining the sharability and consistency of map information under distributed environment, we propose a Multi-level/Multi-theme map information model to maintain maps in consistency with original source datasets under distributed environment. However, the distributed management of spatial datasets results in a complex maintenance processing, especially when the modification refers to several datasets. To solve this problem effectually, in this paper we propose an index structure, MOR-tree (Multi-levels of Object-Relation tree), for organizing integrated maintenance procedure. MOR-tree is an extension of R-tree for indexing spatial objects of multi-levels in one hierarchy and records relations among objects at different levels. The performance of MOR-tree is also evaluated with a prototype system in this paper.
Optimization methods by using chaos dynamics are interesting as a class of global optimization methods by which the global minimum can be obtained without trapping in local minima. The chaos dynamics are classfied into discretized gradient models and continuous dissipative models with a nonlinear damping term. In this paper, two types of constrained optimization problems are considered in order to present nonlinear dissipative dynamics embedded in their constraints. One of types of the constraints is upper and lower bounds on each variable, and the other type is a simplex. For the each type of constraints, the inner state model with nonlinear dissipative dynamics w.r.t.inner states is introduced, which is composed of a nonlinear inertial model with the gradient and a nonlinear output function. As the nonlinear dissipative dynamics, Fujita-Yasuda's Model  and Tani's Model  are adopted. Especially, their revised models are proposed newly for the simplex type. The numerical simulations for a few constrained optimization problems demonstrate effectiveness of presented constrained global optimization methods.