This paper deals with some scientific approaches to curling. Curling is a winter sport called chess on ice. Players slide stones on a sheet of ice towards a target area. A stone is delivered with a gentle rotation, and then the stone curves in response to the rotation direction. The purpose of this game is to control the stone and obtain a higher score. There are still many aspects to analyze about curling, such as why a stone turns, how it curves, where it should be delivered at, how to develop several strategies, and so on. In order to find them, scientific approaches are necessary as well as intuitions and experiences. In these approaches, measurement, robotics, artificial intelligence, and information processing are included.
The ET Robot Contest is an attempt of education of software engineering for embedded system using robot contest. In the ET Robot Contest, the results are evaluated by both the competition result of autonomous mobile robot and the poster of software design for the robot. The robot used in the contest is a one make made by Lego Mindstorms. The result of competition derives from the software design. Evaluation on its poster is fed back to the contest participants by posting good points and points to be improved along with the poster. Also, at the workshop after the competition, the judges have a comment on the overall trend and particularly good points about its poster. ET Robot Contest provides technical education to contest participants. The technical education deals with element technologies to robot control, UML modeling and so on.
This paper deals with a transfer method for a thin plate by levitating with multiphase air flow. Non-contact transfer methods are required for glass plates and semiconductor wafers in order to keep them clean. In the proposed method, a thin plate was slightly levitated with a small inclination by blowing and sucking air from orifice restrictors, and airflow was blown against the lower side of the plate so that the driving force was generated. The plate was continuously carried by shifting the air flow pattern in accordance with the plate position. 48 air flow slots were arranged with intervals of 15 mm. The velocity of a plate with dimensions of 90 mm × 100 mm × 2 mm and a weight of 139 g was almost linearly increased at an acceleration of 15 mm/s2 and reached 50 mm/s after driving for 3 s. The levitated height was 0.13 and 0.28 mm at the lower and higher sides at air pressures of the positive and negative restrictors of +7.0 and less than -3.0 kPa, respectively. In the case of an initial velocity of 60 mm/s, the plate was turned in reverse after moving for 100 mm. The plate can move reciprocally by reversing the air flow pattern.
If an ultrasonic imaging inside an object is possible without scanning by using an ultrasonic probe having a small aperture, the ultrasonic imaging system becomes a simpler and easier system. We have developed a stacked encoding ultrasonic transducer which can transceive an M-sequence signal only by a pulse drive in order to realize an ultrasonic measurement with a high resolution and high sensitivity. When an ultrasonic probe with such transducer is used, a pulse compression measurement is enabled only with a general ultrasonic flaw detector and a PC without using special devices. In this study, we focused on the characteristic that the received waveform by this probe changes with transceiver directions of a wave. A stacked M-sequence encoding probe of 7 bits length was manufactured and the one-dimensional point spread function in transceiver angle was obtained. As a result, it was confirmed that this probe has not only the pulse compression characteristic but the angle selecting characteristic. Finally, by using the proposed probe, the cross-sectional imaging for point objects without scanning was demonstrated.
Micro flow rate measurements are required for manufacturing and medical fields. Straight type ultrasonic flowmeter can measure the liquid flow by using guided waves. However, guided waves propagate by vibrating intricately along boundaries, so it is difficult to reveal their characteristics. In addition, the time difference between two receiving waves across two transducers differs depending on the temperature of the liquid, which can cause errors in flow rate measurements. In this paper, the theoretical analysis method for guided wave flowmeter is introduced to investigate the influence of water temperature on flow rate measurement. The simulated time difference is calculated by using this method, while the experimental time difference is obtained in a circulation system. By comparing simulated and experimental values, this theoretical analysis can be confirmed to be valid for revealing the temperature dependence.
In this paper, we propose a new robust scheduling strategy. This strategy aimed at maintaining and increasing efficiency in productivity by using mathematical models. In this model, the deviation value between the schedule and the execution result was taken as an evaluation function. This deviation value is approximated by using the standard deviation. By optimizing the deviation and efficiency evaluation function, it is possible to suppress the influence of uncertain events and doing efficiency production. In addition, this model can suppress the influence of indeterminate event by putting forced idle time. Finally, Numerical experiments are based on the processing time of job following normal distribution and exponential distribution.
This paper proposes a knowledge transfer method based on state value for reinforcement learning (RL) agents. It has a fundamental problem that RL which is the one of machine learning techniques needs a lot of time or the number of trials because the agents acquire appropriate skills through trial and error in order to solve a task. Transfer learning (TL) allows the agent to transfer knowledge which is acquired by itself in other tasks, or previous knowledge to solve a target task. So, TL for RL is able to speed up the learning than simple RL. Our proposed method transfers both state value and a new policy which is given by state values of two selected knowledge to as initial knowledge for an unknown state. The effectiveness of the proposed method was verified with the simulation of the reaching problem for a multi-link robot arm. The proposed method has reduced the learning time 40% than the conventional method.
An active suspension system with preview control is important to enhance ride-quality and driving stability. The main problem with preview control is difficult to obtain road displacement in front of the ego-vehicle. In this paper, we propose a road displacement estimation method for preview suspension control by Using Stereo Camera. In particular, to obtain the road displacement, we reconstruct a 3D point cloud on road surface using stereo matching method which is SGBM (Semi-global block matching) and estimate 3D road displacement form the 3D point cloud on road by using road grid. Considering the 3D road displacement data acquired from different time, we enhance the estimation accuracy by using point cloud registration method which is ICP (iterative closest point). Finally, we present several simulation result of preview suspension control to evaluate an effectiveness the proposed estimation method.
In recent years, robots have become familiar with human society. Human beings are reading human attention from surrounding circumstances in daily life. However, a household robot that is currently popular is mostly what it is not possible. Humans have gained a lot of information by the visual information. Thus, by analyzing the human gaze movement, the robot is possible to estimate the human attention. If the robot obtain the attention, the robot would run the appropriate support action for human. Human attention is not only point of gaze. Therefore, method is needed for estimation of human attention map. In this paper, we focus on human gaze movement and estimate human's attention from the information. As a result, the proposed method brings about a result that changes according to human attention, and it can be used for estimation of human's attention.
Currently, the remote control such as the bilateral control which enables transmitting the multimodal information has been studied. However, remote control has not been reached while perceiving the wind, gravity, and the condition of the ground on the moving object at the remote place. When it is possible to perceive the external force applied to the moving object, it is possible to perform an appropriate operation sensuously according to a remote place such as a slope or a gravel road. Therefore, we propose a high immersive telexistence system with HMD and the six axis motion base which enables transmitting the inertia and force information of the moving vehicle, and 360-degree image of moving object's surroundings at real-time.
We proposed a method for automatic programming of inspection image processing system called “Image Processing Network Programming (IPNP)”. So far our studies make it possible to detect defect and make image processing algorithms automatically. However, IPNP need many types of node function to clear many kind of tasks, but it makes complicated for the node structure and image processing program. In this paper, we propose a new automatic generation method for visual inspection program inspired by human implementation approach.
In fruit farming, the appropriate time of harvesting high-quality and high-value fruits is determined by comparing the surface color of tree's fruit and a fruit color chart defined for each fruit cultivar. This comparative examination comprises visual inspection by the farmer and a sensory inspection based on sensibility information. Thus, the decision varies across individuals because the decision criterion is not defined quantitatively and is not unified. This study aims to develop a system to determine the appropriate time for harvesting fruit from the tree. In this paper, we used apples to develop a quantitative method for appropriate harvest time. We focused on the HSV and L*a*b* color spaces used by image processing. Consequently, it was found that using the value of H and a* in deciding on the proper harvest time was useful. We also conducted a comparative analysis of apple growth and color and color differences to investigate our quantitative method for determining appropriate harvest time. These results of the procedure show that quantitatively determining the proper harvesting time of fruit is possible.
This paper reports analytical evaluation for reverse recovery action and turn-off action from high level injection state of PiN diodes and IGBTs, respectively. In the analytical evaluation, we take account of structural parameters and current continuity at the both ends of i-layer. We evaluate optimum structure for minimum energy dissipation of these switching actions by using newly employed index RiE calculated by Ri(resistance of i-layer at high level injection state) ×E(energy dissipation on i-layer during these switching actions). RiE is minimized in the case that the carrier piled up at the portion from which depletion layer start to spreading. Besides, we analytically clarify the ideal carrier profile for minimum RiE.
Floating offshore wind turbines (FOWTs) have attracted interest in recent years. FOWTs operate under irregular wind and waves, and these disturbances cause generator output fluctuation and platform motions. In this paper, a PI controller in collective blade pitch control is used to mitigate these fluctuation and motions. The PI controller gain parameters are tuned by the FRIT (Fictitious Reference Iterative Tuning) method, and we design a gain-scheduled PI controller based on tuned gains at multiple operating points. We show that the gain-scheduled PI controller improves the disturbance suppression performance.
It takes much time to watch first person view videos captured with a head-mounted camera, because an important scene won't be found immediately. If we can leave a message for the video when recording, such a video can be utilized effectively. This paper presents a new method of real-time video annotation without using a sensor besides the camera. The proposed method records annotation to first person view videos captured with a head-mounted camera by writing to the palm of the hand. This method creates annotation videos which show hand writing scenes automatically after the work.
The experimental results show that it is a short time to acquire the proposed method, it is easy to annotate a video, the proposed method can create annotation video from video recorded by first person view and a character written by hand gesture shown in the annotation video can be recognized with high accuracy.
It is needs self-localization without GPS, when an autonomous mobile robot is doing in indoor. One of the landmark detection is used self-localization method. Conventional landmark detection methods were used template matching or color data. However, it is difficult to the landmark detection using conventional methods in real environment, because the landmark detection requires a lot of template. In addition, the landmark detection methods proposed using Scale-Invariant Feature Transform (SIFT) and Speeded Up Robust Features (SURF). Among them, SURF is strong to scale-change, rotation, and varying illumination. But, the detection methods using features cannot detect the landmark from many features similar to the landmark in real environment. Therefore, in this study, we build several landmark detectors using SURF and environment information is strong to scale-change, rotation, and varying illumination. In addition, we build a rule to select the best detector from these detectors by the reinforcement learning. We examine usefulness of the proposed method by comparing a landmark detection rate of the conventional method that is the AdaBoost detection using Haar-like features. The experiment result shows that the proposed method surpasses the landmark detection rate than the conventional method.
Human tracking in crowded scenes is a challenging problem because of frequent occlusion and the presence of similar regions. In this paper, we propose an online human tracking method which can handle occlusion and targets with similar regions. Our method compares the target region with a surrounding region and targets with similar regions at current frame. In addition, we also compare the target region at current and previous frames. We reduce the probabilities of uncommon colors at current and previous frames, and the tracking accuracy is improved. The effectiveness of the proposed method has been demonstrated by comparison with state-of-the-art trackers on the PETS2009 dataset.
This paper proposes a total optimization method of a smart community (SC) by Differential evolutionary particle swarm optimization considering reduction of search space. Japanese experts have developed various sectors of SC models such as an electric utility model, an industry model, and a building model. This paper utilizes the models and tries to optimize whole of a SC in order to minimize total energy costs and shift electric power load peak of the SC. The simulation results by the proposed method are compared with those by Particle swarm optimization, Differential evolution, and Evolutionary particle swarm optimization based methods.
In recent years a study of evolvable hardware (EHW) which can adapt to new and unknown environments attracts much attention among hardware designers. EHW is reconfigurable hardware and can be implemented combining reconfigurable devices such as FPGA (Field Programmable Gate Array) and evolutionary computation such as Genetic Algorithms (GAs). As such research of EHW, Block-Based Neural Networks (BBNNs) have been proposed. BBNNs have simplified network structures such as two-dimensional array of basic blocks, and their weights and network structure can be optimized at the same time using GAs. SBbN (Smart Block-based Neuron) has been also proposed as a hardware implementation model of basic blocks which have four different internal configurations. SBbN preserves a sufficient number of weights so as to implement all four types of basic blocks. However, SBbN constantly needs to preserve weights unnecessary for some types of basic blocks, and thus consumes redundant hardware resources. In this paper, we propose a new model of BBNNs in which all weights in SBbN are used efficiently with modifying calculation procedures of outputs of basic blocks in order to eliminate the resource redundancy of SBbN. In the proposed model, the required number of basic blocks in BBNNs can be reduced because of using efficiently all weights in SBbN. In order to evaluate the proposed model, we apply it to XOR and Fisher's iris classification. Results of computational experiments indicate the validity of the proposed model.
This paper describes pattern databases for solving Hakoiri-Musume type puzzles using the A* algorithm and the IDA* algorithm. Although the original Hakoiri-Musume puzzle is not very hard to solve even on a small PC, it is not easy to solve extended versions of the puzzle, particularly, to obtain the optimal solutions of the puzzles. The puzzles have multiple goal states and are completely different from the sliding puzzles whose all pieces are distinguished such as the fifteen puzzle. In this paper, we defined relaxation problems of the puzzles and constructed pattern databases of the problems. Using the evaluation functions derived from the databases, we were able to obtain the optimal solutions of the puzzle of size 6 × 7 from randomly generated initial patterns about 100 to 10,000 times faster than the ordinary breadth-first search algorithm. For some of our pattern databases, the sum of the construction time of the database and the search time using the database was less than the running time of the ordinary breadth-first search algorithm.
In this paper, we propose high-accuracy occupancy detection using low-resolution electricity consumption data. In Japan, residential smart meters, which automatically read and transmit energy consumption data at each household to electric power companies, have started to be installed and will be set up in 80 percent of households by 2020. Occupancy detection is one of the major techniques leveraging electricity consumption data and is applicable various services such as ambient assisted living and peak load shifting. However, it is difficult to conduct high-accurate occupancy detection using the transmitted smart meter data because they are in 30-minute interval and truncated to 100Wh units. Especially, truncation makes difficult to analyze the change of demand by absence. Therefore, we propose a machine-learning based occupancy detection method combined with the estimation of the actual consumption from the truncated data using total variation regularization. In experiments, our method shows the performance is comparable to the result using the raw demand data in 1W unit.