Despite the increasing number of automated systems that have been introduced in vehicles over the past decade, highly automated vehicles are not yet capable of driving reliably and safely in all traffic scenarios and conditions. Until they are, humans will need to remain in the loop – to take back vehicle control when the driving capabilities of the automated system(s) are limited or systems fail. This automation-to-manual transition may be problematic if the driver is inattentive or distracted by competing activities. If so, this may compromise the driver's ability to take back control in a safe and timely manner. The aims of this study are to: (a) review what is known about driver inattention and distraction during periods of highly automated driving, (b) to outline countermeasures that have or may have potential to prevent and mitigate the effects of inattention and distraction during automated driving and (c) to highlight future research directions that may further inform the understanding and management of distraction and inattention as vehicles become increasingly automated. The paper concludes by contemplating whether a fully self-driving vehicle may itself be distracted or inattentive to activities critical for safe driving.

Smartphone usage while driving, and particularly texting, poses a major concern for road safety. The goal of this study is to suggest a novel and objective means to measure the smartphone usage among young drivers. A naturalistic study was conducted with 254 Israeli young drivers who installed a research-oriented smartphone app which continuously monitors smartphones usage while driving. The app captures the actual number of times drivers are ‘touching’ their smartphone screens, the speed at which these screen-touches occur, foreground apps and time stamps. The results, which are based on 3304 h of driving performed in 11,528 trips, indicate that young drivers touch their smartphone screen on average 1.6 times per minute of driving. Alarmingly, more than half of the screen-touches are performed while the vehicle is in motion, and some touches occur even at speeds higher than 100 km/h. The screen-touches occur throughout the trip regardless of its duration. Approximately half of them are performed while using WhatsApp, a popular free messaging app. These findings provide objective evidence to the actual and intensive usage of smartphones. While comparing these results to participants' self-reports, it was found that young drivers clearly underestimate their smartphone usage while driving.

Using a Wizard-of-Oz approach, the authors explored the effectiveness of engaging drivers in conversation with a digital assistant as an operational strategy to combat the symptoms of passive task-related fatigue. Twenty participants undertook two 30 min drives in a medium-fidelity driving simulator between 13:00 and 16:30 when circadian and homeostatic influences naturally reduce alertness. Participants were asked to follow a lead-car travelling at a constant speed of 68 mph, in a sparsely populated UK motorway scenario. During one of the counterbalanced drives, participants were engaged in conversation by a digital assistant (‘Vid’). Results show that interacting with Vid had a positive effect on driving performance and arousal, evidenced by better lane-keeping, earlier response to a potential hazard situation, larger pupil diameter, and an increased spread of attention to the road-scene (i.e. fewer fixations concentrated on the road-centre indicating a lower incidence of ‘cognitive tunnelling’). Drivers also reported higher levels of alertness and lower sleepiness following the Vid drive. Subjective workload ratings suggest that drivers exerted less effort to ‘stay awake’ when engaged with Vid. The findings support the development and application of in-vehicle natural language interfaces and can be used to inform the design of novel countermeasures for driver fatigue.

The recently proposed cognitive control hypothesis suggests that the performance of cognitively loading but non-visual tasks such as cell phone conversation selectively impairs driving tasks that rely on top-down cognitive control while leaving automatised driving tasks unaffected. This idea is strongly supported by the existing experimental literature and the authors have previously outlined a conceptual model to account for the key underlying mechanisms. The present paper presents a mechanistically explicit account of the cognitive control hypothesis in terms of a computational simulation model. More specifically, it is shown how this model offers a straightforward explanation for why the effect of cognitive load on brake response time reported in the experimental lead vehicle (LV) braking studies depends strongly on scenario kinematics, more specifically the initial time headway. It is demonstrated that this relatively simple model can be fitted to empirical data obtained from an existing meta-analysis on existing LV braking studies.

It is important to gain a better understanding of how drivers interact with in-vehicle touchscreens to help design interfaces to minimise ‘eyes-off-road’ time. The study investigated the relative effects of two interaction mechanisms (peripheral vision – PV, muscle memory – MM) shown to be relevant to visual behaviour when driving, on the time to press different sized buttons (small 6 × 6 cm, medium 10 × 10 cm, large 14 × 14 cm) on an in-vehicle touchscreen. Twenty-five participants took part in a driving simulator study and were presented with a single, white, square button on 24 successive trials. For MM conditions, participants wore a pair of glasses blocking their PV and for PV conditions, they were asked to keep their focus on the vehicle ahead. Results showed that task time gradually decreased when participants could only use MM. However, overall task time for MM conditions were significantly higher than PV conditions. Participants rated the use of MM to be more difficult than PV. In contrast, results suggest that for interfaces that utilise peripheral visual processing, the learning effect is not evident and operation times are constant over time. These findings indicate that in-vehicle touch screens should be designed to utilise PV for making simple button selections with reduced visual demand.

This study discusses the use of granular computing for representing multi-thinking cellular automata model. In addition, the learning in cellular automata model is examined from the viewpoint of granular computing. A granular cellular automata system for simulating the changes needed in cases based on assessments of individual and group decision from the viewpoint of soft computing as a new formulation of granular computing and cellular automata is presented here. The architecture of the proposed model and the results of simulation of novel approach are given. Results from the implementation enrich granular computing cellular automata hybrid system and shed a new light on the concept formulation of the model and the learning in it.

The number of pedestrian casualties in crashes with motorised vehicles is still alarming. Misunderstandings about the other road users’ intentions are certainly one contributory factor. Especially given recent developments in vehicle automation, informing about ‘vehicle behaviour’ and ‘vehicle intentions’ in the absence of any direct interaction between the driver and the outside world is becoming increasingly relevant. A frontal brake light which communicates that a vehicle is decelerating could be a simple approach to support pedestrians and other road users in the interaction with (potentially automated) motorised vehicles. To assess the effect of a frontal brake light on the identification of vehicle deceleration, the authors conducted a video based lab experiment. The brake light facilitated the identification of decelerations considerably. At the same time, the fact that only half of the decelerations were accompanied by the brake light resulted in increased identification times for decelerations in which the frontal brake light was absent compared to a control condition in which none of the decelerations was indicated by such a light. This finding points towards an increasingly conservative approach in the participants’ assessment of deceleration, which could be interpreted as an indicator of a potential safety effect of the frontal brake light.

Drivers always suffer varying degrees of performance decrements under insufficient visual feedback (VF) conditions. Nowadays, haptic guidance (HG) is a developing assistance technology to enhance steering performance; however, driver reactions to HG under degraded VF conditions are still unclear. Therefore, this study focuses on the influence of HG on driving behaviour when part of the road ahead is occluded. The experimental conditions combined three levels of HG, namely none, weak, and strong torques, with four scenarios of VF: whole, near, mid, and far segments. The driving experiment was conducted using a high-fidelity driving simulator with 12 participants. By analysing the standard deviation of lane position and time-to-lane crossing, it was shown that the lane keeping performance became worse without the HG for the degraded VF of near and far segments compared to that of whole and mid-segments. Furthermore, it indicates that the performance decrement in the worse cases was compensated by the implementation of HG, and the strong torque was significantly more effective than the weak torque. Additionally, the use of HG always resulted in an improved turning manoeuvre while approaching curves in the degraded VF of near and far segments.

Accurate traffic flow prediction can provide reliable and precise information for traffic departments to formulate effective management measures and assist drivers in performing more intelligent route planning and rerouting. The authors propose a short-term traffic flow forecasting framework for urban expressways based on data-driven mixed models including an approach to traffic flow threshold identification based on an improved semi-supervised K-means clustering algorithm, a hybrid multi-scale traffic speed forecasting method based on wavelet decomposition, and a traffic condition index corresponding to three-phase traffic flow theory for reflecting traffic status in real time. Model performance evaluation is performed using multi-source travel speed data. The results show that the traffic threshold recognition algorithm can correctly identify traffic speed thresholds confirming to the three-phase traffic flow transition and that the proposed short-term estimation technique outperforms traditional auto-regressive integrated moving average models, extended Kalman filtering methods, and artificial neural network models in terms of both accuracy and robustness. The proposed traffic condition index using adaptive thresholds and predicted speeds can provide real-time quantitative surveillance for urban expressway traffic.

This study proposes a feature selection-based approach to identify reasonable spatial–temporal traffic patterns related to the target link, in order to improve the online-prediction performance. The prediction task is composed of two steps: one hybrid intelligent algorithm-based feature selector (FS) is proposed to optimise original state vectors, which are designed empirically during the offline process and optimised state vectors are employed to carry out the online prediction. Numerical experiments by three non-parametric algorithms are conducted with taxis’ global positioning system data in an urban road network of Changsha, China. It is concluded that: (i) under optimised state vectors, the prediction accuracies improve or almost maintain the same; (ii) K-nearest neighbour (KNN) with the simplest state vectors obtains the greatest improvement of prediction performance; (iii) although the performance improvement of ɛ-support vector regression is limited with optimised state vectors, it always outperforms backward-propagation neural network and KNN; and (iv) three non-parametric approaches with optimised state vectors outperform auto-regressive integrated moving average in relatively longer prediction horizons. In conclusion, such FS-based approach is able to improve or guarantee the prediction performance under the remarkably reduced model complexity, and is a promising methodology for short-term traffic prediction.

The tracking accuracy of speed plays a significant role in the autonomous vehicle's control and safety management. In this study, we presented a novel method called self-adaptive proportional integral derivative (PID) of radial basis function neural network (RBFNN-PID) which is shown with improved longitudinal speed tracking accuracy for autonomous vehicles. A forward simulation model of longitudinal speed control for autonomous vehicles is established based on the driver model of self-adaptive RBFNN-PID and the vehicle dynamics model. Based on that, we used the traditional PID and fuzzy control methods as benchmarks to demonstrate the edge of the self-adaptive RBFNN-PID control under the new European driving cycle. Simulation results show the RBFNN-PID method is significantly more precise than the comparing groups, with a reduced error in the range of [−0.369, 0.203] m/s. The vehicle performance gives better ride comfort as well. In all, self-adaptive RBFNN-PID is proven to be effective in longitudinal speed control of autonomous vehicles and significantly outperforms the other two methods.

Road safety analysis is important for vehicles that travel at high speeds. This study proposes a novel viability kernel calculation method for road safety analysis. The viability kernel can be represented by a polyhedron, which is suitable for calculation of the algorithm. This facility solves the problem of analysing the permanent admissible areas of the road with constrains and all possible input of the vehicle. First, the concepts of the viability kernel are described and a simplified vehicle model is developed. Second, the viability kernel is applied to analyse road safety for vehicles on a straight road. Third, the analyses are extended to a straight-corner-straight model, which can be used to simulate general road conditions. The proposed method can predict the largest safety area of the road for vehicles travelling at a high constant speed. In the future, the proposed methods can be applied to calculate the radical safe control for high-speed vehicles in complex road circumstances.

High-speed rails have a significantly higher driving safety requirement than other public transport because of more passengers and faster speed. However, the particularity of train driving often leads to railway drivers more susceptible to drowsiness and fatigue, and this fatigue has a distinct personality. Here, the authors analyse the mechanism of individual fatigue generation of different drivers, and use eye movement data collected by non-contact means as an objective measurement, and combine subjective sleepiness assessment to reveal the existence of individual differences in fatigue in high-speed railway driving. Furthermore, a novel driver-specific feature weighted support vector machine (FWSVM) algorithm is proposed to handle the individual differences. In the FWSVM, features are assigned with different weights by the information gain to reflect classification importance and individual effects. The average accuracy of FWSVM is 90.98%, the average sensitivity is 92.01%, and the average specificity is 89.88%, which is better than the classical SVM. Such improvements are attributed to the quantitative evaluation of individual effects by the weighted features. These results can be used as a preliminary study to design a high-speed rail vehicle interface to prevent driver fatigue.

Advanced driving assistance systems (ADASs) play a vital role in the safety of transportation. The detection of lane markings is a very important part of ADASs. For the safety of autonomous vehicles and vehicles driven by human drivers, accurate detection results are necessary. In this study, the authors propose a novel algorithm based on lane structural analysis and convolutional neural networks (CNNs) for lane marking detection. First, a pre-processing stage is used to remove the pavement that constitutes the background of the lane markings. Next, a set of local waveforms from local images is used to generate a region of interest and a CNN classifier is employed to detect lane marking candidates. Finally, a lane geometry analysis stage determines whether or not the candidate is a part of a lane marking. The major contributions of this study can be summarised as follows. First, they propose a novel method to describe a road using waveforms. Second, they analyse the local and global characteristics of the road geometry to detect the lane markings. Third, they provide an effective method to obtain training data for the proposed machine learning method. Experimental results demonstrate that the proposed method outperforms conventional methods.

Despite the continuously tightening emissions legislation, urban concentrations of nitrogen oxides (NO _x ) remain at harmful levels. Road transport is responsible for a large fraction, wherein diesel engines are the principal culprits. Turbocharged diesel engines have long been preferred in heavy duty applications, due to their torque delivery and low fuel consumption. Fleet operators are under pressure to understand and control the emissions of their vehicles, yet the performance of emissions abatement technology in real-world driving is largely unquantified. The most popular NO _x abatement technology for heavy duty diesel vehicles is selective catalytic reduction. In this work, the authors empirically determine the efficiency of a factory-fitted selective catalytic reduction (SCR) system in real-world driving by instrumenting passenger buses with both a portable emissions measurement system and a custom-built telematics unit to record key parameters from the vehicle diagnostics systems. They find that even in relatively favourable conditions, while there is some improvement due to the use of SCR, the vehicles operate far from the design emissions targets. The archival value of this study is in quantification of real world emissions versus design levels and the factors responsible for the discrepancy, as well as in examination of technologies to reduce this difference.

Cyclist safety is increasingly becoming a societal problem in Europe, as shown by road safety statistics. Frequent stops for red traffic lights at intersections are experienced by cyclists as a major inconvenience. This study introduces a green wave concept for cyclists, with focus on the traffic management and control aspects under cooperative intelligent transport systems applications. It especially addresses increasing stability of the adaptive control system, to overcome the drawbacks of both actuated and traditional adaptive control (which are too unpredictable for a green wave speed advice). In addition, solutions for avoiding increased delays for other traffic are investigated, as generally result from a classic green wave approach (with only fixed-time control) and traditional adaptive control. This study introduces an adaptive control algorithm for a real-time model-predictive controller and implements a plan-deviation cost function to address stabilisation. Simulation results show that the developed method increases stability of the adaptive control system, limits extra delays for other traffic and yields a high success rate for the green wave concept.

Traffic flow prediction plays an indispensable role in the intelligent transportation system. The effectiveness of traffic control and management relies heavily on the prediction accuracy. The authors propose a model based on deep belief networks (DBNs) to predict the traffic flow. Moreover, they use Fletcher–Reeves conjugate gradient algorithm to optimise the fine-tuning of model's parameters. Since the traffic flow has various features at different times such as weekday, weekend, daytime and night-time, the hyper-parameters of the model should adapt to the time. Therefore, they employ the genetic algorithm to find the optimal hyper-parameters of DBN models for different times. The dataset from Caltrans Performance Measurement System was used to evaluate the performance of their models. The experimental results demonstrate that the proposed model achieved better performance in different times.

Automatic license plate recognition (ALPR) systems have become an important tool to track stolen cars, access control, and monitor traffic. ALPR system consists of locating the license plate in an image, followed by character detection and recognition. Since the license plate can exist anywhere within an image, localisation is the most important part of ALPR and requires greater processing time. Most ALPR systems are computationally intensive and require a high-performance computer. The present algorithm differs significantly from those utilised in previous ALPR technologies by offering a fast algorithm, composed of structural elements which more precisely conducts morphological operations within an image, and can be implemented in portable devices with low computation capabilities. The present algorithm is able to accurately detect and differentiate license plates in complex images. This method was first tested through MATLAB with an on-line public database of Greek number plates and was 100% accurate in all clear images, and achieved 98.45% accuracy when using the entire database which included complex backgrounds and license plates obscured by shadow and dirt. Second, the efficiency of the algorithm was tested in devices with low computational processing power, by translating the code to Python, and was 300% faster than previous work.

Visual inspection of target parts is a common approach to ensuring train safety. However, some key parts, such as fastening bolts, do not possess sufficient feature information, because they are usually small, polluted, or obscured. These factors affect inspection accuracy and can lead to serious accidents. Therefore, traditional visual inspection relying on feature extraction cannot always meet the requirements of high-accuracy inspection. Deep learning has considerable advantages in image recognition for autonomous information mining, but it requires a considerable amount of computation. To resolve the issues mentioned above, this study proposes a method that combines traditional visual inspection with deep learning. Traditional feature extraction is used to locate the targets approximately, which makes the deep learning purposeful and efficient. A composite neural network, stacked auto-encoder convolutional neural network (SAE-CNN), is provided to further improve the training efficiency. A SAE is added to a CNN so that the network can obtain optimum results faster and more accurately. Taking the inspection of centre plate bolts in a moving freight car as an example, the overall system and specific processes are described. The study results showed satisfactory accuracy. A related analysis and comparative experiment were also conducted.