Japan is currently facing issues such as a shortage of truck drivers and an increase in logistics volume due to the spread of internet shopping services. The practical application of self-driving trucks is expected to solve these issues, and development is underway with the aim of realizing level 4 self-driving trucks on highways after 2026. When merging at a merging section of a highway, vehicles need to accelerate or decelerate depending on the speed of other vehicles traveling on the main line. In this study, we consider a smooth merging control algorithm for self-driving trucks using infrastructure cooperation.

LRT (Light Rail Transit) has been introduced in cities in Japan and overseas, and its effects have been confirmed. However, in Japan, although reevaluation is progressing, the introduction of LRT has not been seen to a large extent. Therefore, in this study, we examined the possibility of introducing LRT by focusing on core cities that are likely to enjoy the effects of LRT introduction. In this case, we modeled data such as population, geography, regional economy, industrial structure, and transportation as indicators, and attempted to identify indicators that are effective for classifying cities with introduction by principal component analysis. Next, we tried to extract indicators that have a large impact on the annual number of LRT and tram users by multiple regression analysis. We also examined the feasibility of introducing LRT in one of the extracted cities by checking the geographical information in detail.

This study focuses on the shape of the control interface and the direction of operation, specifically examining a bar- shaped control system (hereafter referred to as the "bar handle"). In the case of the bar handle, even when the gear ratio (steering angle/front wheel steering angle) is set to 1 (steering angle = front wheel steering angle), sensitive responses are less likely to occur. Additionally, it has been observed that even drivers with little experience can operate the vehicle without practice. It is hypothesized that intuitive operation is achieved because the bar handle can be steered in the direction of the target. Kitahara et al. have reported improvements in mental comfort and control performance du ring driving using an experimental vehicle equipped with a bar handle. Their findings show that ease of driving varies depending on the shape of the control interface, even for the same driver, though the underlying mechanism remains unclear. This research aims to explore the mechanism behind the improved ease of driving by investigating it through a driver model. As the first step, steering behavior during trajectory tracking is decomposed into feedforward and feedback components, and the ratio of these components is evaluated to examine how the driver model changes depending on the shape of the control interface and the direction of operation.

It is important for the safety of outdoor machines such as crawler cranes to avoid hazardous event such as tipping, structural damage, and excessive motion under the influence of wind and the ground supporting the machine, which is the natural environment. We have been conducting research to develop an operation support system using deep reinforcement learning, with a focus on lifted load sway. In this study, we describe the development status of a simulator that links a deep reinforcement learning agent created by MATLAB/Simulink with the Unreal Engine, a 3D game engine, to load sway control.

Self-position estimation is one of the technological elements of autonomous buses that have been developed in recent years. One of the technological elements of self-position estimation is the use of location information from magnetic markers installed on the road. However, it is necessary to estimate self-position using dead reckoning between magnetic markers. Therefore, it is necessary to design appropriate marker spacing considering the effects of disturbances. In this study, crosswind and sensor noise are selected as disturbances that affect self-position estimation of an autonomous bus, and their effects on self-position estimation and steering control performance are investigated through simulations. As a result, it was found that there is a large difference between the estimated self-position and the actual position because of integration errors caused by sensor noise. In addition, it was found that the error in the self-position estimate was larger when the entire side of the vehicle body was hit by crosswind than only a part of the side of the vehicle body.

This study aims to establish an index for quantitatively evaluating the physiological and psychological fatigue of the driver and passenger while the automobile is in motion. The driver drove a circulation course of about 400 m per lap for three hours, measuring the physiological and psychological responses of the driver and passenger during a simple driving operation or operation to stop randomly within the course. We examined the time-series variation of fatigue through physiological and psychological response measurement during automobile driving. We report findings on the fatigue of the driver and passenger during a three-hour drive derived from physiological and psychological responses measured in an actual vehicle experiment.

In recent years, the bus industry has faced significant challenges due to driver shortages and an aging workforce. Continuous monitoring of drivers is essential for ensuring safe and sustained driving. By understanding driving characteristics and providing safety education tailored to individual drivers, it is possible to enhance their driving skills and foster their professional development. This study aims to analyze and evaluate bus drivers' driving characteristics using logs collected during training sessions. In this paper, we analyze the training logs from two drivers—one novice and one veteran— across two sessions each, and assess their driving characteristics. The results confirm that it is possible to evaluate driving characteristics using the training logs. The analysis also revealed that the veteran driver demonstrated significantly higher driving skills compared to the novice, and that the novice driver showed improvement in driving skills over time.

There is a risk of slipping when driving on icy road surfaces or in case of heavy rains. Understanding driving in hazardous conditions is important for safety. However, it is difficult for people to understand it and the best driving reactions in depth. Videos are difficult to understand. Using an actual vehicle requires a lot of effort. Therefore, this study proposed driving experience mobility in hazardous situations using a prototype of omnidirectional vehicle, equipped with specific control logic that can emulate situations where the vehicle loses grip. This facilitates the experience of driving in hazardous situations and enables the learning of appropriate driving techniques in such conditions.

In this study, we assume a mixed environment of automatic level 4 vehicles and automatic level 2 driver assistance vehicles. Some HMIs, human-machine interface was proposed, and DS experiments were conducted to evaluate the effectiveness of communicating free space information that means there is no obstacle for Level 4 automatic driving obtained from the infrastructure to driver assistance vehicles. The results showed that the direct transmission of the information that there was no oncoming vehicle reduced the subjective load and the timing of right turn became faster, but it also induced accidents in some situations and the driver did not check the safety of the vehicle. When information about the presence of oncoming traffic or information that assists the driver's decision to turn right is communicated, accidents can be prevented, and the driver can check for safety in the same way as when driving manually without overconfidence in the system.

This paper employs deep learning techniques and an LSTM network to output driver attention weights in both temporal and spatial dimensions. A three-lane driving scenario was designed on a driving simulator, where vehicle trajectory data and eye movement data were collected from twelve drivers as they completed lane-changing tasks. The collected data were used to obtain the distribution of spatial and temporal weights under the DS dataset, and the results were compared with the conclusions derived from the NGSIM dataset. Finally, the processed gaze behavior data were fed into the LSTM model, improving the accuracy of trajectory prediction.

One of the most challenging collision avoidance scenarios is to avoid parked vehicles on Two-Lane Two-Way (TLTW) because of dynamic scenarios where divers must evade the parked vehicles while remaining alert of oncoming vehicles. Traditional behavioral decision-making systems for automated vehicles typically rely on processed image data. While these systems have made significant advancements, they often struggle to effectively prioritize and process the most relevant aspects of the scene, such as parked vehicles, oncoming vehicles, and pedestrians. This limitation can impact the overall safety and efficiency of collision avoidance. Saliency prediction offers a promising solution by identifying and highlighting the most important regions in a visual scene. By focusing on these salient features, saliency prediction can enhance the decision-making process by emphasizing what is most relevant in a given context. This study proposes a novel decision-making framework that utilizes saliency prediction to improve decision-making performance. Additionally, this study is aimed at developing a novel saliency prediction model for autonomous driving: Weighted Salient Object Detection (WSOD). WSOD gives varying importance levels to salient objects within a scene. This approach is expected to enhance decision-making performance by prioritizing salient objects.

As a train speeds increase, the elastic vibration of a car body tends to increase. To reduce the elastic vibrations of the car body, it is important to understand the natural vibration modes of the car body. This paper reports the experimental identification results of the natural vibration modes of the Shinkansen vehicle up to 40 Hz, which have not been reported so far.

The authors developed the Hybrid Simulator (HS) for Wheel Slide Protection (WSP). By using real pneumatic system hardware, the HS improves the reproducibility of the air brake response characteristics which are difficult to model. However, in order to correctly evaluate WSP performance, a test method considering the characteristics of the real pneumatic system hardware is required. In this study, we perform the WSP tests under conditions that suppress the fluctuation of the brake cylinder pressure setting value caused by the tolerance which is inevitably included in real pneumatic system hardware of the HS, and we report the effects of the test conditions on WSP performance evaluation.

In this study, the authors examined an application of cutting force measurement using SAICAS as one of the methods for evaluating wear resistance that is simpler than full-scale brake tests. The relationship between the estimated shear strength obtained from cutting force measurements and the specific wear rate obtained from full-scale brake tests was investigated for several copper-based sintered alloy brake friction materials used in high-speed train, each of which has different composition. In the cutting force measurements, a cutting depth of 10 μm was applied to the specimen using a 0.3 mm wide blade with a rake angle of 5 degrees. The horizontal force measured was then used to calculate the estimated shear strength based on the cutting theory. Full-scale brake tests were repeated with an initial speed of 50 km/h and a contact force of 10.3 kN, and the specific wear rate was calculated from the measured wear mass. As a result, the comparison of the experimental results showed a strong correlation between the estimated shear strength and the specific wear rate of the friction materials. This suggests that the cutting force measurement can potentially be used to evaluate wear resistance.

Understanding the wear behavior of the brake block is necessary to reduce brake block wear and improve brake performance. However, measuring the amount of wear at each moment during braking is currently difficult. Therefore, focusing on the brake dust, a dust measurement method was developed to measure friction behavior with high time resolution and sensitivity. This method captured the differences in the dust emission characteristics of various brake blocks, and linearity was found between the amount of wear and dust except under high load conditions, suggesting the possibility of predicting the wear amount based on the dust measurements.

The application of ultrasonic acoustoelasticity for the safety assessment of tread-braked railway wheels subjected to malfunction drag braking was investigated. Ultrasonic residual stress measurements were found to be potentially more effective for detecting wheel anomalies compared to conventional techniques, such as strain gauges and X-ray diffraction. It was observed that the acoustic anisotropy of wheel cut pieces did not match the values reported in the literature, making direct, non-destructive stress measurement in specific wheels challenging. Therefore, a comparative assessment method was proposed, in which the measured stress values are compared to those from wheels of the same type and diameter, without abnormal loading histories.

During the pass-by of Shinkansen trains in open sections, low-frequency noise is emitted from the vehicles. We have been studying the mechanisms and countermeasures for this low-frequency noise. Due to its characteristics, measurement and evaluation are challenging. In this paper, we attempted to reproduce the low-frequency noise using a low-noise moving model test facility, which can launch a 1/20 scale train model. From the test results, trends in low-frequency noise generated at the bogies of the Shinkansen vehicles were observed.

In order to ensure good current collection quality for Shinkansen, it is important to maintain the lift force of a running pantograph within appropriate range. We usually investigate the lift force of pantograph using large- and small-scaled wind tunnels. The former enables us to test full-scale pantographs, but the costs tend to be high. On the other hand, although the latter costs less, only a limited part of pantograph can be installed due to the size limitation. It is necessary to take into account the differences in the flow field of the former and the latter ones. We therefore compare and investigate the measured pantograph lift force and the pressure distribution on the pantograph head surface in the large and the small wind tunnel.

To further improve the ride quality of Shinkansen vehicles, it is necessary to reduce the vibration of the vehicle carbody. A vibration analysis model is required in order to efficiently consider vibration reduction methods. In this paper, we proposed a vibration analysis model that can express the vibration modes of the carbody around 40 Hz. The validity of the model was verified by comparing the calculated results with the measured results.

In order to clarify the dynamic response of prestressed concrete (PC) sleepers under impact loads, impact tests are conducted on full-scale PC sleepers using a drop weight test machine. In addition, the effects of difference in the rail pad stiffness and the effects of using under sleeper pads (USPs) on the dynamic response of PC sleepers are investigated. The test results show that for the same input energy (calculated from the mass of the drop weight and the drop height), the maximum impact forces are approximately same regardless of the rail pad stiffness and use of USPs, and that it is not also affected by the presence or absence of the sleeper cracks. Furthermore, it is confirmed that the input energy at which cracking occurs increases when the low stiffness rail pad and the USPs are used.

Train localization is an essential technology for ensuring safety in railways. We have previously proposed a low-cost, high-speed train localization method using a 1D LiDAR sensor and have achieved over 90% localization success rate. However, the problem of discriminating between similar structures remains. In this paper, we propose further miniaturization and increased train location information based on layering the cubic codes attached to structures for discrimination. The proposed method is verified through experiments using a scaled model and a linear motor. The experimental results show that the proposed method is successful in accurately decoding cubic codes for all patterns.

For many years, gas pressure welding has been used for the primary welding of rails for the high speed trains construction. In response to the recent labor shortage, flash butt welding was fully adopted from the construction of the Hokuriku Shinkansen (between Kanazawa and Tsuruga). We have established methods to address issues that arise due to the hot and humid environment in summer and the cold environment in winter, and to ensure continuous quality.

We have been studying and verifying the logic for detecting abnormal using physical quantities that act on the single link type traction device of the vehicle, with the aim of improving the performance and reducing costs of abnormal detection. In the previous report, we reported on determining the excess or deficiency of traction force during power running and braking using strain in the tensile and compressive direction, detecting the rotational velocity of the wheel set using frequency analysis, and detecting the bogie turning angular velocity using strain in the lateral bending direction. In this paper, we report on the detection of abnormal rotation of WN coupling and the detection of relative displacement between the car body and bogie using torsional strain.

As part of efforts to support track maintenance work on regional and local railways, a low-cost method of supporting train patrols using smartphones is being developed. In this study, the transition of on-board measured data on a regional railway line, such as train vibration and forward-view image, collected periodically during train patrols using a dedicated smartphone application, was analyzed. It was found that changes in dynamic track conditions can be confirmed from the changes in on-board measured data.

Diesel engines have various parts such as combustion systems, drive systems, intake and exhaust systems, cooling systems, etc. When defects such as deterioration of lubricating oil and deformation due to heat occur, metal-to-metal contact occurs and vibration and noise tend to increase. In the past, we measured the vibration generated by an engine and established a method to estimate the surface condition of engine parts without dismantling by machine learning. In this study, we will verify the improvement of the reliability of the prediction data and the optimization of the measurement time at the time of inspection．

The railway industry is experiencing manpower shortages due to a decline in the number of workers and changes in employment trends. In regional areas other than metropolitan areas, changes in population dynamics have become apparent, leading to unprofitable management. As it has become difficult to maintain the quality of maintenance, the company is considering reforming its maintenance business. This report presents a study of railway equipment maintenance considering risk assessment.

In recent years, new testing methods for evaluating vibration durability using random waves based on actual usage environments have been introduced in various technical fields. The current standard for vibration durability of railway signalling equipment (JIS E 3014) specifies only test methods using sine wave excitation, and until now, in the railway signalling field, there has been little progress in studying vibration durability tests using random waves that are suited to actual usage environments. In this paper, we present the physical characteristics and theoretical analysis results of random waves generated from vibration acceleration data obtained on-site, and report on the ideal form of vibration durability evaluation using random waves.

In the fault detection for track circuits, model-based methods by comparing measured waveforms and fault simulation waveforms may be effective. However, conventional track circuit simulations cannot adequately simulate the characteristics of measured waveforms with multiple frequency components. In this paper, we propose a transient analysis technique for track circuits, which can handle multiple frequency components simultaneously. In addition, we discuss the applicability of the proposed technique to model-based fault detection.

Since 2018, JR East has been utilizing track equipment monitoring devices. However, considering advancements in sensor and other elemental technologies, we have been developing the next-generation devices aimed at further utilizing track equipment monitoring data.

In this study, we have enhanced the track material monitoring devices by expanding the field of view through changes in optical system specifications, acquiring color images, implementing real-time detection of missing joint bolts, and incorporating a laser Doppler velocimeter to achieve higher functionality.

The authors developed an imaging system and a diagnostic algorithm for automated visual inspection of underfloor equipment on passenger rail vehicles. The developed imaging system was installed for testing in a railway yard and recorded many images of rolling stock passing in front of the system for about a year. The system also recorded images of a test train simulating various visual anomalies on the underfloor equipment. The developed diagnostic algorithm was applied to the recorded images to evaluate its diagnostic performance. The results revealed the types of anomalies that can be detected by the proposed algorithm, and the influence of sunlight and rainfall on the diagnostic performance.

Various railway operators are advancing the use of on-board cameras and other methods to streamline track inspections. While methods involving cameras mounted under the floor to capture track materials have been implemented in other companies, there has not been one that can handle high-speed travel of over 200 km/h. Recently, we have developed a Track Material Monitoring System that can be installed on Shinkansen vehicles. A prototype will be manufactured in 2023 and installed on the Series N700S Shinkansen for running tests. In this article, we introduce the overview of the developed system and discuss the utilization of measurement data obtained from the tests.

Previous research has shown that when images acquired by smartphones are applied to the track component condition evaluation system under development, the system can accurately detect and identify the deterioration of wooden sleepers, PC sleepers, rail fastening systems, and other track components. However, the accuracy of rail joint detection and kilometer estimation has not been verified by using the developing system. In this study, images acquired by smartphones were applied to a track component condition evaluation system to verify the accuracy of rail joint detection and kilometerage estimation. As a result, it was found that the developing system was able to detect rail joints with high accuracy. On the other hand, the estimation error of the kilometerage was sometimes greater than that of the conventional method due to the failure to detect kilometerage post marker installed on concreate sleepers in some sections.

Currently, the condition of track equipment and track materials is checked by track patrol and onboard track patrol. Track maintenance engineers perform this task at high frequency through visual inspection and dimensional measurement. To reduce those works, we developed two algorithms using images captured by the track material monitoring system in this study: one for extracting rail corrugation, and another for extracting metal flow at glued-insulated joints. The results of our study are reported below.

Track measuring device with inertial mid-chord offset method is designed to measure rail position by irradiating a laser beam onto the rails, but measured data may not always be used as inspection data due to “whiteout” caused by rail discontinuity in turnouts or weather conditions such as raindrops. In this study, we made a basic analysis of weather-related whiteout in commercial train measurement data using track measuring device with inertial mid-chord offset method, with the aim of efficiently and appropriately selecting data that can be used for practical purposes from frequent track measurement data.

Regional railways, where equipment is aging, frequently experience derailment accidents due to gauge widening. This paper shows a detection method of gauge widening by measuring the positional relationship between the rails and the wheels. The proposed method uses a small, simple camera, which is easy to install and allows inexpensive measurements. The points where the gauge is widened, are identified by section images of rail and wheel extracted from moving image shot by the camera. An overview of the measurement device and experimental results of a commercial train are reported.

In the event of a disaster, we conducted a demonstration test of track inspection using a UAV. As a result, it was possible to fly over the track with the automatic navigation function. In addition, it was possible to grasp the presence or absence of abnormalities from the air. The acquired data can be checked immediately after flight from videos and still images, and by converting them into a point cloud, it is possible to grasp the approximate dimensions.

We have developed a method for measuring the three-dimensional vibration of overhead contact lines by analyzing the image data from videos taken by two cameras. This method makes it possible to measure the vibration of the target equipment from a position away from the equipment without making contact. In addition, because it is not necessary to attach special markers to the equipment, there is no need to enter the tracks and arrange for a power outage to carry out the measurement. This paper outlines the measurement procedure of the developed method, presents the results of the validation tests in the laboratory, and shows an example of the application to an actual overhead contact line equipment.

Influence of changes in the solar radiation condition caused by the shade of a building such as the occurrence of cloudy weather during a certain interval and the track direction on the rail neutral temperature is investigated. For this purpose, numerical simulation is carried out. Through numerical experiments, it is found that the weather change does not result in the reduction in the neutral temperature, while the track direction may rise the rail axial load.