In order to create a running and jumping quadruped robot composed of all articular joints, we have developed a mono-leg robot which simulates the landing and lift-off motions of kangaroos. From the photograph data of the gait motion of kangaroo, time responses of four fundamental quantities are approximated by solutions of first or second order differential equations. Then we propose a control strategy of the robot which realizes these differential equations as controlled constraints. Experimental results show that a running mono-leg robot is produced which has a smooth jumping gaits.
In this paper, the Lagrangian decomposition and coordination technique is applied to flowshop scheduling problems in which the sum of the changeover cost and the tardiness penalty is minimized. The proposed method possesses a feature that the scheduling problem is decomposed not into single job-level subproblems but into single machine-level subproblems. By decomposing the problem into one-machine multi-operation subproblems, the changeover cost and/or changeover time can easily be embedded in the objective function. In the proposed method, each subproblem for a single machine is solved by combining the simulated annealing method and the neighborhood search algorithm. In order to avoid oscillations in multiplier values, a new Lagrangian function is used to solve each subproblem. The effectiveness of the proposed method is verified by comparing the results of the example problems solved by the proposed method with those solved by the simulated annealing method in which a schedule of the entire machine is successively improved.
Even though Independent Component Analysis (ICA) has become an important technique for Blind Source Separation (BSS), it can provide only a crude approximation for general nonlinear data distributions. Karhunen et al. proposed Local ICA, in which K-means clustering method was used before the application of linear ICA. The clustering part was responsible for an overall coarse nonlinear representation of the underlying data, while linear independent components of each cluster were used for describing local features of the data. In this paper, we propose a method for extracting local independent components by using Fuzzy c-Varieties (FCV) clustering, which seems to be more natural than K-means or the like. Because FCV can be regarded as a simultaneous approach to clustering and Principal Component Analysis (PCA), the FCV takes part of the preprocessing of Fast ICA by Hyvärinen et al..
It is difficult to apply the existing adaptive control methods to an inverted pendulum and cart system if it is assumed that all physical parameters are unknown, because the number of control inputs is less than that of outputs. Regarding the inverted pendulum system, the parameter uncertainties of the cart are larger than those of the pendulum. In this paper, the partially adaptive control system is designed considering that the parameter uncertainties exist only in the cart. The inverted pendulum and cart system is divided into the known part which includes only the parameters of the pendulum and the unknown part which includes the parameters of the cart. Therefore, LQ control is applied to the known part and adaptive control treats the unknown part based on the backgtepping technique. Finally, the experimental results are provided and the usefulness of this technique is confirmed.
In our previous work, we have developed the backward selection method based on class regions approximated by ellipsoids. In this paper, we accelerate feature selection by the forward selection search, the symmetric Cholesky factorization, and deletion of duplicated calculations between consecutive factorizations. The feature selection for four data sets shows that our method is faster than and as robust as the previous method.
The simplified fuzzy reasoning model having multi-layer structure has a feature that generation of fuzzy rules is easy even under actual conditions. However, it is reported that redundant fuzzy rules are generated as division layers are iteratively increased. In this paper, we propose a fuzzy reasoning model of multiple division layers, which has a weight in the reasoning output. We reduce redundant fuzzy rules by tuning up the weight using an evaluation function based on Minkowski norm. Moreover, the validity of the proposed method is verified by identification experiments of two nonlinear functions having sparse domains in teaching data. Finally, we apply the proposed method to a modeling of ceramic kiln, and confirm that it is also effective for actual problems.