This study proposes a concept for multithread active vision sensing that can measure dynamically changing displacement and vibration at multiple points on civil engineering structures. In multithread active vision sensing, a high-speed camera can function virtually as multiple tracking cameras by accelerating its measurement, computation, and actuation with ultrafast viewpoint switching at millisecond level. This enables simultaneous measurement of small vibrations distributed across a wide range, which cannot be observed using a single camera system, because its pixel accuracy is generally incompatible with its wide angle of view. We developed a galvanomirror-based high-speed multithread active vision system that can switch 500 different views in a second; it functioned as 15 virtual cameras each operating at 33.3 fps to observe multiple scenes in completely different views. The experimental results for a 4-m-long truss-structure bridge model to which 15 markers were attached show that a single active vision system can observe the deformation of the bridge structure and estimate modal parameters, such as resonant frequencies and mode shapes, at a frequency on the order of dozens of hertz.

Online food delivery services, provided under the multi-service transport platforms such as Grab and Gojek, could significantly change people’s eating-out behavior, which could also change the spatial distribution of restaurants in the long run. This study attempts to empirically identify factors affecting people’s preference on the use of online food delivery services using stated preference (SP) survey data collected with a multi-day smartphone-based travel diary survey in Jakarta, Indonesia. In the survey, we randomly chose observed eating-out trips (i.e., revealed preference (RP)) from a travel diary and asked whether the respondents would like to shift to an online food delivery service in a hypothetical situation in which the delivery cost, delivery time, food cost, and available food types vary across questions. This RP–SP combination allows us to elicit respondents’ preference under the real time–space constraints they had (e.g., he or she must start to work again from 13:00). Our empirical analysis confirms that delivery time and delivery cost are important factors affecting people’s preference. We also discuss the long-term impact of the behavioral changes on the spatial distribution of online food merchants and its policy implications.

Deep learning methods are being increasingly applied in transport studies, while the methods require modellers to go through a try-and-error model tuning process particularly on choosing neural network structure. Moreover, the accuracy level also depends on other factors such as the type of data, sample size, region of data collection, and time of prediction. To efficiently facilitate such a model tuning process, this study attempts to summarize the relationship between the prediction accuracy of deep learning models and the factors which influence it. We conducted a comprehensive review of the literature by adopting a detailed search strategy, followed by a meta-analysis on prediction accuracy. Four separate linear mixed effects models, taking into account unobserved heterogeneities in prediction accuracy across studies, were developed to statistically test the impacts of influential factors on prediction accuracy for (a) all observations (136 studies; 2314 cases), (b) studies with MAPE, MRE, and average accuracy indicators (86 studies; 1,878 cases), (c) classification-based studies with accuracy indicator (29 studies; 220 cases), and (d) traffic forecasting studies with MAPE, MRE, and average accuracy indicators (36 studies, 991 cases). The final model includes additional factors to test the influence of sample size and time horizon of prediction variables. The findings showed that, as expected, deep learning models, particularly ones that consider spatiotemporal dependencies of transport phenomena, show better prediction accuracies compared to conventional machine learning models. We also found that, on average, the prediction accuracy is improved by 5.90% with 100 million additional data, while the accuracy is reduced by 5.28% with 100 min increase in time horizon of prediction in traffic forecasting studies. We concluded this paper with a comprehensive summary of the existing findings on the applications of deep learning to transport studies.

Previously, we proposed a multithread active vision system with virtual multiple pan-tilt tracking cameras by rapidly switching the viewpoints for the vibration sensing of large-scale structures. We also developed a system using a galvanometer mirror that can switch 500 different viewpoints in 1 s. However, the measurement rate of each observation point is low, and the time density is not always sufficient. In addition, strong multiple illuminations are required for the system owing to the retro reflective markers attached to the object being observed. In this study, we propose a multiple vibration distribution synthesis method for vibration analysis that increases the sampling rate of each observation point in the multi-thread active vision system, which is subsequently modified to a system that requires only one illumination by using corner cubes as markers. Several dynamics-based inspection experiments are conducted for a 4 m long truss-structure bridge model. The proposed method and system are verified via a high-order modal analysis, which was impossible to perform in the previous method and system.

Autonomous vehicles (AVs) are rapidly transforming the automotive industry due to rising consumer interest in these vehicles worldwide. However, few studies have compared different countries in terms of public acceptance of AVs. This study compares public acceptance of AVs as a function of risk perceptions in two countries leading the AV industry—Japan and Israel. We set our study within the risk-as-feelings framework. In contrast to “risk as analysis,” which invokes factual reasoning to bear on risk assessment and decision making, “risk as feelings” takes affective cues such as the sense of dread and unfamiliarity into judgments of risk. To this end, we conducted two web-based surveys in Japan in 2017 and Israel in 2021. In a between-subjects design, we manipulated introductory video information to portray various combinations of risk factors commonly associated with AVs: system errors, external interferences with car controls (e.g., hacking), and the inability of the AV to cope with unexpected events. Next, participants were surveyed about how they perceive the risks of AVs and other well-known technologies and activities. Results showed that acceptable risk, perceived risk, and perceived benefit of AVs were all generally higher in Israel than in Japan. The opposite pattern was found for a “risk adjustment factor,” suggesting that the Japanese seek more safety before acceptance than Israelis. Furthermore, we conducted a factor analysis on seven risk dimensions, resulting in a two-factor model of dread and unfamiliarity. Cognitive mapping of AVs and other technologies and activities in the two-factor plane revealed that the AV technologies we studied (i.e., AV-car levels 3 and 4; AV-bus levels 3 and 4) have high unfamiliarity risk but moderate dread risk compared to technologies and activities such as smoking, flying, and handguns. After exposure to video-based educational content, unfamiliarity risk was less influential but dread risk—in particular, related to human-made risks—became more influential. The results indicated that manufacturers and policymakers should emphasize mitigating human-made risks instead of focusing on improving public familiarity with AVs to garner trust and improve public acceptance of the technology.