In this paper we present a new meta-heuristic approach to the traveling Salesman problem (TSP) and evaluate its performance. Proposed method reproduces and selects a population of local optima searched by random start modified Lin-Kernighan (mLK) method. It enhances the search power and efficiency of mLK by fixing the part of the solution whose values coincide each other and thus reducing the search space of solutions. Results of numerical experiments on TSPLIB95 (500 or so cities instances) show that it is superiorly competitive to existing meta-heuristic methods for TSP. The reason of enhanced search power is also investigated through obserbation of its search processes.
A new jobshop scheduling problem with limited common buffers is dealt with in this paper. The standard genetic algorithm is not applicable to the present problem for deadlocks led by buffer constraints. A new algorithm is proposed by combining the genetic algorithm (GA) with a semi-active decoding which avoids deadlocks and satisfies the buffer constraints. Some benchmark problems appending with buffer constraints are tested by both the proposed GA and the multi-start local search method (MLS). Computation results show that the proposed GA outperforms MLS well. Computation is also made by varying the number of common buffers. These results are compared among the available results of the no-buffer jobshop problem and the normal jobshop problem. Consequently the results of the proposed GA are satisfactory.
Decision Directed Acyclic Graph (DDAG) and Adaptive Directed Acyclic Graph (ADAG) are the decision-tree-based support vector machines for multiclass problems. These methods show high generalization abilities but their abilities depend on the structures. In this paper, we determine the structures so that the unclassifiable regions caused by voting are resolved by the decision boundaries for class pairs with low generalization ability. Namely, at the higher level of the tree, we select a pair of classes with higher generalization ability that is estimated by the error bounds proposed for SVMs. We demonstrate the effectiveness of our method using benchmark data sets.
In this paper, we propose a design method for servo systems with tracking robustness. Here we consider the control configuration of robust servo system plus a reference model, in which the robust servo system is designed by the ILQ design method as proposed by the third author. First, we show the configuration of this servo system and then provide a theoretical basis for parameter tuning. Second, we clarify the relation between these parameters and the closed loop performance such as robust stability and tracking robustness. Finally we establish a new design method of ILQ servo system with tracking robustness, and show its effectiveness by a numerical example.