The social insects, such as ants, use pheromones as social information. Presence of other ants can be also useful information for ants' decision-making at the individual-level. In that sense, contacting with other ants may be a private information. In this study, we attempted to address the problem of individual decision-making by focusing on the complementary use of pheromones and the presence of other ants. Individual ants were allowed to enter a bridge at right angles where an ant trail was established. As a result, we found that target ants tended to move against the ant crowd. These results suggest that ants might determine the direction of movement on the ant trails based on contacting with other ants.
By combining image recognition technology and robotics, it is expected to save labor in daily surveys and inspections of social infrastructure. The purpose of this study is automation of wall inspection by a mobile robot and the automatic wall inspection method that meets the three requirements: comprehensive shooting of the wall surface, reliable detection of the recognition target and reducing inspection time. Especially, this paper focus on constraints for image recognition to reliable detection and nonlinear programming problems to reduce inspection time. First, to perform the requirement of comprehensive shooting of the wall surface, we propose the motion plan of the robot and camera module including the setting of the internal parameters of the camera. The method of considering the error that occurs in the robot and camera when shooting wall is also in the method. Next, to perform the requirement of reliable detection of the recognition target, we derived the constraint conditions that enable image recognition with reference to the research on the appearance inspection of the object by the camera module. Furthermore, by using the time required for wall inspection as an objective function and formulating a nonlinear programming problem within the constraints of image recognition, it is possible to achieve the requirement of reducing inspection time. We confirmed that proposed the automatic wall inspection method works correctly in a simulator. In addition, we conducted actual experiments by using a mobile robot to demonstrate the effectiveness of the proposed method. All the QR codes to be recognized which were attached to the wall surface were detected without omission in case of meeting constraints in image recognition. By comparing with other experimental conditions, we confirmed that the time required for the survey was shortened by using the parameters obtained by solving the nonlinear programming problems problem.
In this paper, we propose the method of the tracking with identification and separation using high-speed vision. The method is based on the template matching and update of the template images using instance segmentation. We used the image inside of the contour as template image for template matching and reduced the noice of tracking. Experiments for the images acquired by the camera mounted on actual vehicle are performed to validate the proposed method. Tracking based on the contour information realized more robust high-speed tracking than recent research when the vehicles overlapped.
Automation of toilet cleaning has been demanded at convenience stores. In this paper, we developed a toilet bowl that is easy to clean with the machine and has a cleaning mechanism to clean the side surface of the toilet bowl. The shape of the toilet bowl is cylindrical so that the cleaning mechanism can clean its side surface easily. The cleaning mechanism is installed behind the toilet bowl, and has a columnar microfiber cloth as a cleaning tool. The cleaning mechanism moves along the side surface of the toilet bowl and wipes urine with the rotating cleaning tool. First, we investigated the tendency of urine stains in the proposed toilet bowl using a simulated urine discharge machine. The result showed that the tendency was similar to that of a standard toilet bowl, but less urinary dripping on its side surface. The experiments were conducted to evaluate the performance of the cleaning mechanism. The results confirmed that the cleaning mechanism achieved an average removal rate of 91.5% by cleaning the simulated urine for 20 seconds. Moreover, it was able to clean the simulated urine at removal rate over 95% repeatedly by washing and dehydrating the cleaning tool after each cleaning. Finally, the prospect of a cleaning effect on assumed urine dripping was investigated using the simulated urine discharge machine. The results confirmed that the urine dripping can be cleaned with a high removal rate of 88.3% on average.
Although high-resolution and wide-area display holds promise for high immersion feelings, conventional projection methods have difficulty to manage both of them. We propose gaze oriented projection system using high-speed gaze point estimation and high-resolution tracking projection. Experimental results showed millisecond-order speed, sufficient accuracy, and wide-area of the gaze point estimation and low latency of the proposed system. We also showed two demonstrations: pseudo gaze-oriented view and user-demo without fixing the eye camera.
We propose a new method for object pose estimation using a force-torque sensor and deep neural network. To estimate the pose of object, our method uses contact positions and force directions obtained by a force-torque sensor attached on a robot end-effector. In addition, sets of the contact data have characteristics of both various length and random ordered. For the purpose to deal with such data sets, which conventional neural networks cannot, we employ a symmetric function. The proposed method can estimate 6 degrees of freedom pose of objects without considering optical properties. We confirmed in the simulation that our method can estimate the pose for different types of objects: nut, bolt, and wrench. The experiment of pose estimation for nut in a real environment shows that the performance of 3.9mm in position error, 12.7 degrees in angular error, and 20,000 times faster computation time than the conventional method.
Lacquer, which is called Urushi in Japanese, is natural coating material having various merits such as resistance to water and chemicals, and has been used in traditional Japanese crafts. We investigated electric insulation properties of lacquer films for using them as board and coating material of electronic circuits. The results of our experiments showed that the lacquer films have humidity dependence on the insulation resistivity in the 25-75% range of relative humidity, because they have hygroscopicity. The lacquer films have an insulation resistivity of approximately 1015Ω･at room temperature. However, the insulation resistance decreased by 4-5 orders of magnitude at higher than a certain voltage. It was clarified that the insulation performance at lower than 5 V in the electronic circuits could be maintained by using the lacquer film, which was more than 5µm of thickness.