Futoshi Kobayashi

Sensor Fusion System Using Recurrent Fuzzy Inference

In robotic and manufacturing systems, it is difficult to measure the state of systems accurately because of many uncertain factors and noise, and it is very important to estimate the state of systems. We must measure the phenomena of systems by multiple sensors and estimate the state of systems by acquiring information of sensors. However, we can not acquire all of sensor information synchronically, because each sensor has particular sensor information and measuring time. For estimating the state of systems by multiple sensors, a multi-sensor fusion system fusing various sensory information is needed. In this paper, we propose a Recurrent Fuzzy Inference (RFI) with recurrent inputs and apply it to a multi-sensor fusion system for estimating the state of systems. The membership functions of RFI are expressed by Radial Basis Function (RBF) with insensitive ranges. The shape of the membership functions can be adjusted by a learning algorithm. The learning algorithm is based on the steepest descent method and incremental learning which can add new fuzzy rules. The effectiveness of the multi-sensor fusion system using RFI will be shown through a numerical experiment of moving robot and estimation of surface roughness in grinding process.

Sensor selection by reliability based on possibility measure

Robotic or manufacturing systems become more and more complex for adapting to various environmental conditions. In these systems, sensor fusion methods for estimating states of a system from multiple sensor information have received much attention. Also, it is necessary to select the sensor information for adapting to various situations flexibly. In these days, various methods of fusing multiple information have been proposed so far, but these methods cannot select the sensor information. We propose a sensor selected fusion system using a recurrent neural network. The sensor selection method is based on the production system, considering the reliability calculated by the possibility measure. The effectiveness of the proposed method is shown through a simulation of a mobile robot.

Sensor selection based on possibility measure for grinding system

Grinding process is frequently used to produce a smooth surface in the manufacturing system. In this process, sensor fusion methods for inferring the surface roughness from online sensing information have received much attention, because it takes a long time to measure the surface roughness during process. In addition, it is necessary to select the sensing information for adapting to various situation flexibly. This paper proposes a sensor fusion system with sensor selection method for a grinding system. The sensor selection method is based on the production rules, considering with the reliability calculated by the possibility measure which is a type of the fuzzy measure. Then, the sensor fusion system fuses the sensor information selected by the sensor selection method using a recurrent neural network.

Shape reconstruction system integrating stereo vision and shape from shading with evolutionary programming

We propose a new method for measuring the shapes of objects. The proposed method is based on integrating stereo vision and shape from shading with evolutionary programming (EP). In some traditional methods, the reflection of objects is diffuse reflec- tion, but some methods cannot be applied to real objects because the reflection of real objects consists of diffuse and specular reflec- tion. Photometric stereo can be applied to real objects, but this method requires many original images to measure the shape accu- rately. The proposed method applies the Phong model to a reflec- tivity function and measures the shape of an object that has the diffused and specular reflection effect. Then this method uses EP to identify the reflectivity function accurately.

Sensor selected fusion using selection rules acquired by ES (application to inference of surface roughness in grinding system)

In grinding process, sensor fusion methods for inferring the surface roughness from online sensing information have received much attention, because it takes a long time to measure the surface roughness during process. We propose a sensor selected fusion system using reliability based on possibility. This method can select sensor information by selection rules of reliability determined by an operator in advance. However, this method cannot select sensor with consideration of the characteristic for each sensor because selection rules for all the sensors are same. In this paper, we propose an acquisition method of selection rules for each sensor using an evolutionary strategy (ES) for the sensor selected fusion system.

Shape measurement method integrating stereo vision and shape-from-shading with evolutionary programming

In this paper, we propose a new method for measuring the shape of objects. The proposed method is based on integrating stereo vision and shape-from-shading with evolutionary programming (EP). In some traditional methods the reflection of objects is treated as diffused reflection. However, the reflection of real objects consists of diffused and specular reflections, thus some methods cannot be applied to real objects. Photometric stereo can be applied to real objects, but this method needs many original images for measuring the shape accurately. The proposed method can apply Phong model to a reflectivity function and measure the shape of the object which have the effect of diffused and specular reflections. The method described using EP identifies the reflectivity function accurately.

Determining Internal Local Corrosion of Screwed Pipes Through Ultrasonic Testing

In diagnosing air-conditioning pipe deterioration, it is especially important to inspect screwed parts because their original thickness is thinner than that of straight parts and it is comparatively easy for leakage to occur if there is local internal corrosion. Usually, a radiographic test is used, but this method requires a licensed person to carry out the inspection. We examined a method of applying an ultrasonic test for the inspection, and proposed a method for detecting internal local corrosions of screwed parts and estimating the size of corrosion with surface SH waves. In this paper, we propose a method for detecting internal local corrosion of screwed parts of air-conditioning pipes with ultrasonic testing. The objective of the first step is to detect artificial corrosion and estimate its size and position. It is impossible to detect corrosion with a normal probe because a joint or valve overlapping the screwed pipe prevents ultrasonic waves from reaching the inside. Therefore, we used an angle probe, specifically a surface SH wave angle probe, because the thickness of the pipes requires a large angle of incidence, close to a right angle. We describe the following in this paper: ・ The results of experimental analysis on the echo from screwed parts of pipes. ・ The method used to detect local corrosion and estimate the size and position by using the pipe-end echo. ・ The results of tests on artificial corrosion pipes and the limitations of this method.

Sensor selected fusion system (application to interference of surface roughness in grinding system)

In grinding process, sensor fusion methods for inferring the surface roughness from online sensing information have received much attention, because it is quite time consuming to measure manually the surface roughness during process. In addition, recently, it is necessary to select the sensing information for adapting to various situation flexibly. This paper proposes a sensor fusion system with sensor selection method for a grinding system. The sensor selection method is based on the production rules, considering the reliability calculated by the possibility measure which is a type of fuzzy measure. Then, the sensor fusion system fuses the sensor information selected by the sensor selection method using a recurrent neural network.

Shape reconstruction system integrating stereo vision and shape-from-shading-shape reconstruction using pair improvement method

In this paper, we propose a new method for reconstructing the shape of objects by multiple vision modules. This proposed method uses the stereo vision module and the shape-from-shading module. In stereo vision module, the conjugate pair of edge pixels is determined by structural analysis and the height is calculated by the triangulation. In shape-from-shading module, we use the Phong model (1975) as the reflectivity function and identify the reflectivity function by evolutionary programming. Then, we estimate the normal vector of pixels. An important feature of our shape reconstruction is that these modules complement insufficient information each other. We show the simulated result for reconstructing the shape of simulated objects.