Kitakyushu National College of Technology

Electromagnetic wave (EMW)-absorbing materials have received significant attention in recent years owing to their important role in mitigating electromagnetic pollution caused by the increased use of wireless communication devices. SiC is expected to be one of the EMW absorbers because of its excellent properties such as chemical stability, mechanical stability, and high hardness; however, SiC alone does not have high EMW absorption performance. SiC-carbon composites were prepared as high-performance EMW-absorbing materials by thermal decomposition of 6H-SiC(0001) substrates. The surface morphology and chemical composition of the pure and annealed samples were evaluated using scanning electron microscopy and X-ray photoelectron spectroscopy, respectively. Carbon clusters with sizes of several micrometers were formed on the SiC surface after annealing, which formed in an interfacial structure between SiC and carbon and also increased the surface area of carbon. These microstructural changes improved the electromagnetic absorption of the SiC-carbon composite, which was confirmed by transmission attenuation measurements.

This research proposes a system that coordinates a home service robot with a drone to improve task efficiency.As an experiment, we conduct a search and pick-up task that integrates the home service robot and the drone.The drone's front camera and YOLOv8 are used to detect objects and send information to the robots.

This study investigated the efficacy of a Visual Analog Scale (VAS) and Likert Scale (LS) in assessing emotional intelligence in AI, specifically in the context of short-story comprehension. Although both scales are widely used in psychology and social sciences to measure human subjective experiences, their applicability to AI-generated emotional responses remains understudied. We conducted a comparative analysis using both scales to evaluate the emotional responses (interest, surprise, sadness, and anger) of human participants and an AI chatbot (ChatGPT) after reading short stories. The study revealed that the VAS demonstrated superior performance in capturing AI-generated responses, providing more precise values and greater sensitivity to nuanced emotions, especially with limited data points. In contrast, human responses showed a wider distribution across both scales, reflecting the subjective nature of emotional reactions in the literature. Notably, AI consistently overestimated positive emotions compared to human subjects, whereas its perception of negative emotions varied significantly across texts and measurement methods. These findings underscore the complex nature of evaluating AI-generated emotional responses and highlight the potential of the VAS as a more suitable tool for assessing AI emotional intelligence. This study contributes to the ongoing development of more accurate and nuanced methods for evaluating AI-generated content, thus paving the way for advancements in empathetic and human-like AI systems. Future research directions include expanding the range of emotions evaluated, investigating different AI models, and exploring cross-cultural factors in emotional responses to AI.

Currently, the number of farmers in Japan is decreasing and farmers are aging, creating a labor shortage in the agricultural sector. One solution to this problem is smart agriculture that incorporates AI and robot technology. We have been developing a tomato harvesting robot, and the grasping hand we have developed so far often comes in contact with tomatoes growing in clusters. This contact causes the hand to hit the surrounding area, resulting in the stem being entangled and other tomatoes being dropped, which can lead to poor harvesting. To solve this problem, we developed a tomato harvesting hand with a parallel link mechanism that maintains parallelism between the left and right hand parts when opening and closing the hand, thereby reducing the gap between the hand and the tomato to be harvested, which can cause stems and other tomatoes to wrap around the hand. A tomato harvesting hand with a parallel linkage mechanism was developed. The usefulness of this hand is demonstrated by preparing an experimental environment in which virtual tomato clusters are arranged and individual harvesting targets are harvested from the tomato clusters.

In recent years, smart agriculture has been gaining attention, and active research is being conducted on automation through robotics. Generally, farms involve vast areas for cultivation regardless of the type of crops, so it is more efficient for multiple harvest robots and container transport robots to collaborate on tasks rather than a single robot performing both harvesting and transportation tasks. While research on agricultural harvest robots is reported, there are few reports on the automatic transport of containers after harvesting or the collaborative operations between harvest robots and container transport robots. This research focuses on the development of harvest robots and container transport robots, and examines their collaborative operations and communication methods between the robots, which is reported here.

This paper proposes an RFID signpost system for the blind, giving precise positional information of landmarks. The proposed system consists of an RFID reader placed at location of navigational signs, a smartphone and a white cane with an ID-tag carried by the blind. As the blind traveler comes in a transmission area, the ID-tag is activated and transmits ID number. After the RFID reader receives the ID, an area information is transmitted to the blind’s smartphone. Then, the area information is provided via LINE application. Performance of the proposed system for orientation aids is also discussed in this paper.

Railroad structures must be installed at a sufficient distance from trains based on the railroad's structure gauge. Consequently, there is a demand for a system that can accurately and efficiently determine the positions of these structures. We have developed a system that uses a single 2D-LiDAR to measure structures such as tunnels inside a railroad and to generate accurate 2D drawings. However, it was not possible to clearly distinguish which structure the measured points on the 2D drawing indicated, so visual confirmation was required. In this study, we propose a material identification method that is based on the incidence angle of the laser light in 2D-LiDAR to identify the type of structure on the generated 2D drawing. The accuracy of the system, when using the proposed method, was evaluated by the identification rate and the error in the calculated incidence angle. Verification experiments demonstrated that the identification accuracy exceeds 90%, and particularly high accuracy was observed for certain materials. The results of this research are of great importance in maintaining and managing railroad infrastructure.

In an aging society, accidents involving the elderly constitute a significant social issue. Constant supervision by caregivers, regardless of day or night, is essential for accident prevention, but this leads to an increased burden on caregivers, posing a serious challenge. Against this backdrop, there is a demand for the development of bed-exit detection systems as a monitoring solution for the elderly. Previous research has proposed a system capable of detecting a person's bed-exit while preserving privacy using a sensor called Obrid-Sensor. However, it presented practical challenges. Thus, this study proposes a bed-exit detection method that can adapt to actual environmental changes. The proposed method installs the sensor with infrared LEDs and IR filters under the caregiving bed. Moreover, we extracted the features of the target by applying the inter-frame difference methods to the brightness waveforms obtained. Accordingly, it becomes feasible to robustly detect bed-exit events while minimizing the influence of natural and artificial lighting. To evaluate the effectiveness of the proposed method, verification experiments were conducted against diverse background changes in actual environments. The results revealed that the bed-exit detection rate reached 100%, confirming the practicality of the proposed method, regardless of changes in the real environment or nighttime conditions. Additionally, the false-positive rate was restrained to 10.7% or less against factors potentially leading to misidentification, ensuring a reduction in erroneous detections. These findings suggest that the proposed method can be effectively utilized as a monitoring system for the elderly in actual environments.