Railway operators and infrastructure companies strive to optimise the flow of passengers on and off vehicles whilst aiming to minimise accidents at the platform–train interface (PTI). An ideal solution (already available in some situations) would be step-free access to aid efficient boarding for everyday passengers and those with additional needs or reduced mobility. Out of many solutions existing today, a ‘kneeling vehicle’ seems a possible solution due to the opportunity to minimise the step and gap distances. In this study, the viability of an assumed kneeling mechanism retro-fitted to a contemporary suspension architecture is assessed by evaluating the possible improvement in the step/gap distances based on a detailed model of suspension movement. It is shown that for many different infrastructure scenarios that significant improvements in the PTI are shown for a modest and achievable kneeling action. This study also addresses fundamental operational concerns of a kneeling vehicle by assessing gauging (with respect to infrastructure and adjacent vehicles) and pantograph interaction.

To alleviate the driver's nervousness, a lane-keeping algorithm based on scene analysis is proposed. From the driver's point of view, two key parameters are studied, which are the distance between the traffics car and the lane centreline of the host car, and the time-gap difference between the host car and the traffic car. The designed scene analysis algorithm is mainly based on these two parameters to screen out dangerous traffic vehicles. To ensure driving comfort when driving states change, a five-power polynomial is used to plan a smooth path based on the location of the dangerous traffic vehicle. Considering the real-time and robustness of the control algorithm, this study proposes a feedforward and feedback control architecture combined with the preview-following theory. To verify the algorithm, a variety of typical tests are designed. The experimental results show that the algorithm can make the vehicle smoothly follow the planning path.

The vehicle sideslip angle is an important state for vehicle dynamic control, which needs to be estimated as it could not be obtained directly by the vehicle. To improve the estimation accuracy of the sideslip angle based on the intelligent vehicle platform, this study proposes a novel vehicle sideslip angle estimation algorithm with the fusion of dynamic model and vision information. Firstly, to further improve the model accuracy of the vehicle during lateral acceleration conditions, a vehicle dynamic model is established considering the acceleration error compensation with the assistance of attitude information. In addition, based on the lane line information obtained from the equipped camera in intelligent vehicles, a visual geometric model is established. Owing to the measurement delay and low sampling frequency of the camera, a multi-rate sideslip angle observer with delay compensation is designed to coordinate with the inter-frequency signal of the vehicle chassis. Finally, the effectiveness of the algorithm is verified by the slalom test.

Vehicular ad-hoc network (VANET) is a key enabling technology of intelligent transportation systems. VANETs are characterised by the rapidly changing topology and the unbounded network size. These characteristics present a range of challenges to different VANET applications such as routing and security. Clustering has strongly presented itself as an efficient solution to such challenges. In this study, the authors formulate the clustering algorithm as a many-objective optimisation problem. Then, they propose a unified framework to optimise the configuration parameters arbitrary clustering algorithms. Three many-objective metaheuristic optimisation techniques, ESPEA, MOEA/DD and NSGA-III, are compared in context of this framework, and various commonly used quality indicators are utilised to identify the metaheuristic with the best quality of solutions. The proposed framework is then used to optimise a recent clustering algorithm. Using the optimal configuration resulting from the proposed framework significantly improves the performance of the clustering algorithm under-test compared to the non-optimised algorithm as well as other clustering approaches. This is demonstrated by the simulation results which showed up to 182% improvement in the cluster head lifetime and a reduction of 36% in the clustering packets overhead in the highway environment.

This study presents an adaptive motion planning strategy for automated vehicle collision avoidance systems to be associated with the variation of collision speed region based on the position of the obstacle. This is done by designing the motion planner using an artificial potential field (APF) with the incorporation of an adaptive multi-speed scheduler using fuzzy system in the motion planning structure. The knowledge database information is developed based on the risk perception of the driver that consists of APF parameters and was optimised by using particle swarm optimisation algorithm. This study contributes to the improvement of a feasible reference motion generated by the motion planner that can be converted into desired control signals. The reference motion resulted to provide the control command that managed to avoid collision successfully by evasive manoeuvre without lane departure when adapting to variation in the vehicle speeds with different obstacle positions. The results indicated the reduction of the lateral error with respect to the reference avoidance trajectory data of up to 87% compared to base-type APF with maximum reference lateral motion is reduced of up to 26%. Then, a hardware-in-loop test is conducted to verify the proposed strategy using a steering wheel system.

Short-term passenger flow forecasting is a crucial task for urban rail transit operations. Emerging deep-learning technologies have become effective methods used to overcome this problem. In this study, the authors propose a deep-learning architecture called Conv-GCN that combines a graph convolutional network (GCN) and a three-dimensional (3D) convolutional neural network (3D CNN). First, they introduce a multi-graph GCN to deal with three inflow and outflow patterns (recent, daily, and weekly) separately. Multi-graph GCN networks can capture spatiotemporal correlations and topological information within the entire network. A 3D CNN is then applied to deeply integrate the inflow and outflow information. High-level spatiotemporal features between different inflow and outflow patterns and between stations that are nearby and far away can be extracted by 3D CNN. Finally, a fully connected layer is used to output results. The Conv-GCN model is evaluated on smart card data of the Beijing subway under the time interval of 10, 15, and 30 min. Results show that this model yields the best performance compared with seven other models. In terms of the root-mean-square errors, the performances under three time intervals have been improved by 9.402, 7.756, and 9.256%, respectively. This study can provide critical insights for subway operators to optimise urban rail transit operations.

Many studies have been conducted to estimate origin-destination (OD) demand based on vehicle trajectory data. However, the estimation accuracy heavily relies on the temporal-spatial distribution of trajectories, and its effect on OD estimation remains unrevealed and under-estimated. This study proposes a novel method for investigating the impact of the heterogeneity of trajectory data distribution on OD estimation at urban road networks. Synthetic scenarios are designed based on automatic license plate recognition data collected from a real-world traffic network in Kunshan, China. Four factors: test area, sampling rate, time period, and the sampling method are selected for scenario settings. Next, a particle filter-based method is implemented to reconstruct vehicle trajectories using the sampled trajectory data, and then the path flows and OD demands are extracted. Finally, a spatial statistics approach is introduced to reveal the spatial autocorrelation of trip generation/attraction variations, and the high-high clusters whose OD values are significantly affected are identified. Test results show that the heterogeneity effects of trajectory distribution on OD estimation can be effectively studied by the proposed method. Further investigation shows that the findings of spatial statistics can be applied for improving OD estimation accuracy.

Autonomous vehicles (AVs) are undergoing rapid worldwide development. They will only become a success if they are accepted by their users. Therefore, there is a need for user acceptance for these vehicles. Previous studies on acceptance of AV have identified several predictors. Inspired by these studies, the authors’ investigation is aimed at sociodemographic characteristics, including broader individual and societal acceptance, beyond technical issues to get a clear view of user acceptance. In this study, they surveyed 229 respondents, using a 46-item online questionnaire. Overall, the authors’ analysis revealed that the respondents are most concerned about crashing/malfunctioning, purchase price, liability for incidents, interaction with non-AV, performance in unexpected situations, hacking and safety. The results also revealed that the AV is perceived as ‘somewhat low risk’ to drive. Gender, age, education level and employment status had varied relationships with the perceived concerns and general attitude towards the AV. For instance, respondents with a university degree (Bachelor/Master/PhD) are less concerned about the liability of accidents and AV system failure than those without it. Similarly, respondents between 36 and 65 years of age are more concerned and even refused to drive AV in comparison to the age ranges of 18–35 years and 65 +  years.

Recently, low-mobility individuals have drawn growing attention in the traffic organisation field. Therefore, following the previous research efforts, this study tries to design and optimise a multimodal traffic strategy for low-mobility individuals. First, the preliminary application conditions and implementation effects of different traffic strategies are analysed to determine feasible multimodal traffic strategies for low-mobility individuals based on the optimisation goal and strategic orientation. Then, an optimisation model is proposed to select the optimal strategies from feasible strategies for different travel modes, aiming to minimise the weighted sum of public transport, walking, non-vehicle-based transportation, travelling by private car and strategy implementation costs. An appropriate heuristic algorithm is developed to solve the optimisation model. Finally, the case study of Wenling is presented to verify the feasibility and practicality of the proposed multimodal traffic strategy design and optimisation for low-mobility individuals. The results show that the proposed multimodal traffic strategy has stronger positive than negative effects from the travel cost aspect. The findings presented in this work can provide a reference for other, more in-depth, studies on the multimodal traffic travel of low-mobility individuals in the future.

Traffic flow prediction is regarded as an important concept used in traffic planning, traffic design, and traffic management. In this study, the authors propose a multi-component attention (MCA) method for traffic flow prediction, which may jointly and adaptively understand components of long-term trends, seasons, and traffic flow residuals that result from multi-dimensional decomposition. According to the highly non-linear nature of traffic flow, the proposed module consists of a one-dimensional convolutional neural network, a bidirectional long short-term memory, and a bidirectional mechanism with an attention mechanism. The former captures local trend characteristics of residuals, while the latter captures trends and seasonal time adjustments. Due to the randomness, irregularity, and periodicity of traffic flow at intersections, target flow prediction is related to various sequences. Through the introduction of the attention mechanism, highly related historical information may be connected for multi-component flow data in the final prediction. Compared to seasonal autoregressive integral moving average model, artificial neural network, and recurrent neural network, the experimental results demonstrated that the proposed MCA model can meet the accuracy and effectiveness of complex non-linear urban traffic flow prediction models.

The current development of Advanced Rider Assistance Systems (ARAS) would interestingly benefit from precise human rider modelling. Unfortunately, important questions related to motorbike rider modelling remain unanswered. The goal of this study is to propose an original cybernetic rider model suitable for ARAS oriented applications. The identification process is based on experimental data recorded in real driving conditions with an instrumented motorbike. Starting with a dynamic neural network, the proposed methodology firstly presents a non-linear rider model. The analysis of this model and some analogies with car driver modelling allow to deduce a quasi-linear parameter varying (quasi-LPV) rider model with explicit speed dependence and a clear distinction between linear and non-linear dynamics. This quasi-LPV model is further analysed and simplified and finally leads to a rider model with a reduced number of parameters and nice prediction capabilities. Such a model opens up interesting perspectives for the improvement of rider assistances.

For vehicle safety, the in-time monitoring of the driver and assessing his/her state is a demanding issue. Frustration can lead to aggressive driving behaviours, which play a decisive role in up to one-third of fatal road accidents. Consequently, the authors present the automatic analysis of the emotional driver states of frustration, anxiety, positive and neutral. Based on experiments with normal drivers within cars in real-world (low expressivity) situations, they use speech data, as speech can be recorded with zero invasiveness and comes naturally in driving situations. A careful selection of speech features, subject data identification, hyper-parameter optimisation, and machine learning algorithms was applied for this difficult 4-emotion-class detection problem, where the literature hardly reports results above chance level. In-car assistance demands real-time computing. A very detailed analysis yields best results with relatively small random forests, and with an optimal feature set containing only 65 features (6.51% of the standard emobase feature set) which outperformed all other feature sets, producing 35.38% unweighted average recall (53.26% precision) with low computational effort, and also reducing the inevitably high confusion of ‘neutral’ with low-expressed emotions. This result is comparable to and even outperforming other reported studies of emotion recognition in the wild. Their work, therefore, triggers adaptive automotive safety applications.

Traffic service is an important building block of smart cities, affecting citizens’ travel quality. Due to the characteristics of complicated and volatile traffic scenarios, fast and accurate modelling and processing requirements are vital for traffic service, such as traffic congestion solutions. Traditional deep reinforcement learning (DRL) approach can make decisions autonomously, but its complex network structure leads to time-consuming training and updating processes. In addition, it is not always feasible to provide a large amount of tagged data in real life. To solve these problems, this study proposes a semi-supervised double duelling broad reinforcement learning (semi-DDBRL) approach based on the broad reinforcement learning (BRL). It incorporates some algorithmic improvements into the BRL, such as duelling network and double Q-learning network, and adds semi-supervised learning to improve the accuracy of modelling and decision making. As a case study of smart city applications, the authors apply the proposed semi-DDBRL approach to the problem of traffic congestion. Based on the experiments, their approach can have a faster execution time than the DRL approach. Moreover, compared with the BRL approach, their approach can improve the performance by 11.7%.

Traffic state prediction plays an important role in intelligent transportation systems, but the complex spatial influence of traffic networks and the non-stationary temporal nature of traffic states make it a challenging task. In this study, a new traffic network state prediction model for freeways based on a generative adversarial framework is proposed. The generator based on the long short-term memory networks is adopted to generate future traffic states, and a discriminator with multiple fully connected layers is applied to simultaneously ensure the prediction accuracy. The results of experiments show that the proposed framework can effectively predict future traffic network states and is superior to the baselines.

Parking has been a common problem over several years in many cities around the globe. The search for parking space leads to congestion, frustration and increased air pollution. Information of vacant parking space would facilitate to reduce congestion and subsequent air pollution. Therefore, the aim of the study is to acquire vehicle occupancy in an open parking lot using deep learning. Thermal camera was used to collect videos during varying environmental conditions and frames from these videos were extracted to prepare the dataset. The frames in the dataset were manually labelled as there were no pre-labelled thermal images available. Vehicle detection with deep learning algorithms was implemented to perform multi-object detection. Multiple deep learning networks such as Yolo, Yolo-conv, GoogleNet, ReNet18 and ResNet50 with varying layers and architectures were evaluated on vehicle detection. ResNet18 performed better than other detectors which had an average precision of 96.16 and log-average miss rate of 19.40. The detected results were compared with a template of parking spaces to identify vehicle occupancy information. Yolo, Yolo-conv, GoogleNet and ResNet18 are computationally efficient detectors which took less processing time and are suitable for real-time detection while Resnet50 was computationally expensive.

Public transport operators often struggle to provide a reliable and efficient transport service. A lack of comprehensive real-time operational data is often cited as a major cause for this state of things. In this study, the authors report on the design, implementation and testing of an Internet of Things-based system, named Bus Seating Information Technology system, which dynamically determines vehicle occupancy while the bus is in service. It uses an array of sensors for detecting events in the vehicle: infrared sensors ascertain whether passengers are entering or leaving the bus; force-sensitive resistors facilitate seat-occupancy detection; a Global Positioning System shield in conjunction with a Raspberry Pi microcomputer enables real-time tracking of the bus; and a USB camera connected to the same Raspberry Pi assist in cross-checking and validating the preceding information. The data collected is uploaded to an online IoT platform (thinger.io), through 3G or 4G if available, and can be visualised via an android app as well as through a desktop computer user interface. The planned functions of the system were tested in a 20-seater bus. Results showed that the system can track the vehicle location, as well as vehicle occupancy in real-time in most cases.

With the development of water transport, there is a rapid increase in maritime traffic density. Therefore, navigation security should be given adequate attention. In this study, a risk evaluation system combining wind field model, wave model and standards of risk grades was proposed and applied in the Qiongzhou Strait (QS), China. Based on meteorological data, a wind filed model was established by Fujita–Takahashi formula and was verified using observed data set of typical typhoon process. Then, a simulation wave nearshore wave model which was used to simulate the significant wave height (SWH) was established with bathymetry data. In addition, through the analysis of the types and quantities of wrecks and ships in the research area, the typical ship was selected and classified by tonnage. Combining the as low as reasonably practicable with wind speed and SWH, the risk evaluation system for navigation security was proposed. Applications on QS showed that the system could provide an hourly reference of risk grades to water traffic safety early warning. Also, the conclusions of this study could serve as recommendations for Maritime Administration to manage navigation safety.

With the growth of autonomous vehicles and collision-avoidance systems, several approaches using deep learning and convolutional neural networks (CNNs) continually address accuracy improvement in obstacle detection. The authors introduce a three-stage architecture that adds side channels as low-level features to serve as input to existing CNNs. In a case study, the architecture is used to extract depth from stereo cameras, and then compose RGBD inputs to state-of-the-art CNNs to improve their vehicle and pedestrian detection accuracy. This can be achieved by simple modifications on the first layers of any existing CNN with RGB inputs. To validate the architecture, the state-of-the-art matching cost-CNN, and cascade residual learning, both specialist algorithms to extract depth information combined to the state-of-the-art Faster-region-based CNN, MSCNCN, and Subcategory-aware Convolutional Neural Network (SubCNN) to yield the models to be tested using the KITTI dataset benchmark. In many cases, the accuracy (in terms of average precision) using their proposal outperforms the original scores in various scenarios of detection difficulty, reaching improvements up to +3.96% in the training and +1.50% in the testing KITTI datasets. This proposal also introduces efficient methods to initialise the weights of the depth convolutional filters during transfer learning using net surgery.

Realising the spatio-temporal evolutionary pattern of urban traffic can give advice about making personal trip route planning and improving road construction. A novel pattern-discovering model is presented to identify the traffic regularity and characteristics from spatial and temporal dimensions. To unveil this new method, there are two main parts as follows: first, by employing the constrained projected gradient of the non-negative matrix factorisation algorithm, the original traffic data matrix is decomposed into the feature matrix and the weight matrix. Necessary constraints are newly added so that the resulting matrices are ensured to make practical sense for reflecting the traffic spatio-temporal regular pattern. Then, the self-organising maps network is further used to cluster the factorisation error into several classes representing the disparate traffic pattern of each time. In addition, the experiment is conducted on real historical data to verify the performance of the algorithm. The global urban traffic flow for a week is summarised through a set of basic patterns with related weight distribution. The well-visualised result demonstrates that the authors method can achieve significant improvement in terms of computational efficiency and accuracy when compared with other widely-used methods.

Dynamic lateral lane change (DLLC) control of automated and connected vehicles (ACVs) is challenging because of the time-varying and complex properties of the traffic environment. This study proposes a DLLC control strategy combining dynamic trajectory planning and tracking. According to the real-time longitudinal accelerations and velocities of multiple surrounding vehicles, as well as the real-time states of the ACVs, the safe trajectory reference of DLLC is obtained by solving a case-dependent constrained optimisation problem. The lane changing efficiency, vehicle stability and passenger comfort are considered jointly in the trajectory planning. Then, the dynamic trajectory reference is tracked through a gain-scheduling control algorithm combining previewed trajectory feed-forward and ACVs states feedback. Gain-scheduling control algorithm based on a linear time-varying form is utilised to achieve the precise control of the different velocities and improve the real-time ability of the algorithm. The proposed strategy is tested through software and hardware-in-loop experiments, and in different test scenarios. The results of simulations and experiments show that the proposed control strategy can achieve a satisfactory performance of DLLC. The lane changing efficiency, safety, passenger comfort and vehicle stability are verified in complex traffic environments.