Pattern recognition using feature feedback: Application to face recognition

In this paper, we propose a new pattern recognition method using feature feedback and present its application to face recognition. Conventional pattern recognition methods extract the features employed for classification using PCA, LDA and so on. On the other hand, in the proposed method, the extracted features are analyzed in the original space using feature feedback. Using reverse mapping from the extracted features to the original space, we can identify the important part of the original data that affects the classification. In this way, we can modify the data to obtain a higher classification rate, make it more compact or abbreviate the required sensors. To verify the applicability of the proposed method, we apply it to face recognition using the Yale Face Database. Each face image is divided into two parts, the important part and unimportant part, using feature feedback, and the classification performed using the feature mask obtained from feature feedback. Also, we combine face recognition with image compression. The experimental results show that the proposed method works well.     

Improving traversability of quadruped walking robots using body movement in 3D rough terrains

This paper presents a study on improving the traversability of a quadruped walking robot in 3D rough terrains. The key idea is to exploit body movement of the robot. The position and orientation of the robot are systematically adjusted and the possibility of finding a valid foothold for the next swing is maximized, which makes the robot have more chances to overcome the rough terrains. In addition, a foothold search algorithm that provides the valid foothold while maintaining a high traversability of the robot, is investigated and a gait selection algorithm is developed to help the robot avoid deadlock situations. To explain the algorithms, new concepts such as reachable area, stable area, potential search direction, and complementary kinematic margin are introduced, and the effectiveness of the algorithms is validated via simulations and experiments.     

Hybrid genetic algorithm with adaptive local search scheme for solving multistage-based supply chain problems

The optimal design of supply chain (SC) is a difficult task, if it is composed of the complicated multistage structures with component plants, assembly plants, distribution centers, retail stores and so on. It is mainly because that the multistage-based SC with complicated routes may not be solved using conventional optimization methods. In this study, we propose a genetic algorithm (GA) approach with adaptive local search scheme to effectively solve the multistage-based SC problems.The proposed algorithm has an adaptive local search scheme which automatically determines whether local search technique is used in GA loop or not. In numerical example, two multistage-based SC problems are suggested and tested using the proposed algorithm and other competing algorithms. The results obtained show that the proposed algorithm outperforms the other competing algorithms.     

Magnetic landmark-based position correction technique for mobile robots with hall sensors

We propose a precise position error compensation and low-cost relative localization method in structured environments using magnetic landmarks and hall sensors. The proposed methodology can solve the problem of fine localization as well as global localization by tacking landmarks or by utilizing various patterns of magnetic landmark arrangement. In this paper, we consider two patterns of implanted permanent magnets on the surface, namely, at each vertex of regular triangles or rectangles on a flat surface. We show that the rectangular configuration of the permanent magnetic bars is better for a robust localization under sensor noise. For the experiments, permanent magnet sets in rectangular configuration are placed on the floor as landmarks at regular intervals, and magnetic hall sensors are installed at the bottom of a mobile robot. In our implementation, the accuracy after the error compensation is less than 1 mm in the position and less than 1° in the orientation. Due to the low cost and accuracy of the proposed methodology, it would be one of the practical solutions to the pose error correction of a mobile robot in structured environments.     

Reconstruction of surface texture based on spatial information measured with a pen-type texture sensor

Surface texture is one of the important properties for the human to identify objects by touch. Effective reconstructions of textures are necessary for realistic interactions between the human and environment via human–computer interfaces. This paper presents a systematic approach for sensing and reconstructing periodic surface textures. Three significant issues are discussed: a pen-type texture sensor that measures the spatial information based on the measurements of contact forces; an algorithm for the reconstruction of periodic textures based on the obtained spatial information; and the method of incremental scanning to identify the polar spectrum of a surface by limited number of scans. The concept of polar spectrum is introduced to describe the spatial properties of the surface, that is, the relation between spatial frequencies and the direction of measurement. The pattern of polar spectrum is used to facilitate surface reconstructions. Experimental results based on the spatial information obtained with a laser displacement sensor and the pen-type texture sensor demonstrate the effectiveness of the proposed methods for the measurement and reconstruction of periodic textures.     

Characterization of a novel antibody immobilization combining protein G with parylene-H for surface plasmon resonance immunosensors

Here we present a method for selectively and efficiently immobilizing antibodies to enhance the detection performance of surface plasmon resonance immune-sensors (SPRIs) for diagnostic applications. To improve the performance of antibody arrays, protein G was used as antibody-selective linkage layer with aldehyde functionalized poly-(para-xylylene) film. To estimate the efficiency of antibody immobilization, immunoglobulin G (IgG) was measured using the anti-IgG immobilized SPRIs. To demonstrate the proof-of-concept validation, the signal detected from the IgG using parylene-H film was compared with that of a combination of parylene-H and protein G in SPRIs. The results showed that the detection of IgG on the immobilized anti-IgG layer using the combination of parylene-H and protein G has a larger change of signal than that of using parylene-H layer. These results also imply that the anti-IgG was densely and efficiently immobilized on the modified surface with the linkage layer in a combination with parylene-H and protein G. Therefore, we believe that this combinatorial approach could selectively immobilize the antibodies, and also be applied for detection and diagnosis of immune diseases in the field of many SPRIs applications.     

Risk-sensitive control of Markov jump linear systems: Caveats and difficulties

In this technical note, we revisit the risk-sensitive optimal control problem for Markov jump linear systems (MJLSs). We first demonstrate the inherent difficulty in solving the risk-sensitive optimal control problem even if the system is linear and the cost function is quadratic. This is due to the nonlinear nature of the coupled set of Hamilton-Jacobi-Bellman (HJB) equations, stemming from the presence of the jump process. It thus follows that the standard quadratic form of the value function with a set of coupled Riccati differential equations cannot be a candidate solution to the coupled HJB equations. We subsequently show that there is no equivalence relationship between the problems of risk-sensitive control and H ^∞ control of MJLSs, which are shown to be equivalent in the absence of any jumps. Finally, we show that there does not exist a large deviation limit as well as a risk-neutral limit of the risk-sensitive optimal control problem due to the presence of a nonlinear coupling term in the HJB equations.

     

Block-based neural networks

This paper presents a novel block-based neural network (BBNN) model and the optimization of its structure and weights based on a genetic algorithm. The architecture of the BBNN consists of a 2D array of fundamental blocks with four variable input/output nodes and connection weights. Each block can have one of four different internal configurations depending on the structure settings, The BBNN model includes some restrictions such as 2D array and integer weights in order to allow easier implementation with reconfigurable hardware such as field programmable logic arrays (FPGA). The structure and weights of the BBNN are encoded with bit strings which correspond to the configuration bits of FPGA. The configuration bits are optimized globally using a genetic algorithm with 2D encoding and modified genetic operators. Simulations show that the optimized BBNN can solve engineering problems such as pattern classification and mobile robot control.