Showing papers in "Iet Intelligent Transport Systems in 2019"

Hybrid dual Kalman filtering model for short-term traffic flow forecasting

Abstract: Short-term traffic flow forecasting is a fundamental and challenging task since it is required for the successful deployment of intelligent transportation systems and the traffic flow is dramatically changing through time. This study presents a novel hybrid dual Kalman filter (H-KF 2 ) for accurate and timely short-term traffic flow forecasting. To achieve this, the H-KF 2 first models the propagation of the discrepancy between the predictions of the traditional Kalman filter and the random walk model. By estimating the a posteriori state of the prediction errors of both models, the calibrated discrepancy is exploited to compensate the preliminary predictions. The H-KF 2 works with competitive time and space to traditional Kalman filter. Four real-world datasets and various experiments are employed to evaluate the authors’ model. The experimental results demonstrate the H-KF 2 outperforms the state-of-the-art parametric and non-parametric models.

Innovative method for traffic data imputation based on convolutional neural network

Abstract: The quality of traffic data is crucial for modern transportation planning and operations. However, data could be missing for various reasons. Hence, the data imputation approaches which aim at predicting/replacing the missing data or bad data have been considered very important. The traditional traffic data imputation approaches mainly focus on using different probability models or regression methods to impute data, and they only take limited temporal or spatial information as inputs. Thus, they are not very accurate especially for data with a high missing ratio. To overcome the weaknesses of previous approaches, this study proposes an innovative traffic data imputation method, which first transforms the raw data into spatial-temporal images and then implements a deep-learning method on the images. The key idea of this approach is developing a convolutional neural network (CNN)-based context encoder to reconstruct the complete image from the missing source. To the best of the authors' knowledge, this is the first time a CNN method has been incorporated for traffic data imputation. Experiments are conducted on three months of data from 256 loop detectors. Through comparison with two state-of-the-art approaches, the results indicate that this new approach increases the imputation accuracy greatly and has a stable error distribution.

Urban rail transit passenger flow forecast based on LSTM with enhanced long-term features

Abstract: Outbreak passenger flow is the main cause of rail transit congestion. In this regard, the accurate forecast of passenger flow in advance will facilitate the traffic control department to redeploy infrastructures. Traffic sequence is a typical time series with long temporal dependence. For instance, an emergency may cause traffic congestion for the next several hours. Only a few studies focused on the way to capture long temporal dependence of passenger flow in the rail transit system. Here, an improved model enhanced long-term features based on long-short-term memory (ELF-LSTM) neural network is proposed. It takes full advantages of LSTM Neural Network (LSTM NN) models in processing time series and overcomes its limitations in insufficient learning of long temporal dependency due to time lag. The proposed network strengthens the long temporal dependency features embedded in passenger flow data and incorporates the short-term features to predict the origin destination (OD) flow in the next hour. The experiment results show that ELF-LSTM outperforms other state-of-the-art methods in terms of forecasting.

V2V-based method for the detection of road traffic congestion

Abstract: The traffic congestion detection based on the internet of vehicles is gaining enormous research interest. A vehicle-to-vehicle (V2V)-based method for the detection of road traffic congestion is proposed. Firstly, a fuzzy controller was constructed based on the vehicle speed, traffic density, and traffic congestion rating system, and the level of local traffic congestion was evaluated. Then, the level of local traffic congestion of neighbouring vehicles was queried based on V2V communication, and the level of regional traffic congestion was obtained based on a large sub-sample hypothesis test. Finally, a simulation test platform was built based on vehicles in network simulation, and the back-off time slots and received packets of vehicle nodes were calculated. The accuracy of the proposed method for detecting road traffic congestion was compared to the cooperative traffic congestion detection ( CoTEC ) method and the geomagnetic coil method. The results show that the detection accuracy of the proposed method increased by 5.5 and 7.5%, respectively, compared to the geomagnetic coil method and CoTEC method. The V2V communication network overhead of the proposed traffic congestion detection method is reduced by 90.8% compared to the adopted CoTEC method. The communication overhead of the vehicle node using the proposed method is significantly decreased when there is no traffic congestion.

Autonomous driving: a bird's eye view

Abstract: The introduction of autonomous vehicles (AV) will represent a milestone in the evolution of transportation and personal mobility. AVs are expected to significantly reduce accidents and congestion, while being economically and environmentally beneficial. However, many challenges must be overcome before reaching this ideal scenario. This study, which results from on-site visits to top research centres and a comprehensive literature review, provides an overall state-of-the-practice on the subject and identifies critical issues to succeed. For example, although most of the required technology is already available, ensuring the robustness of AVs under all boundary conditions is still a challenge. Additionally, the implementation of AVs must contribute to the environmental sustainability by promoting the usage of alternative energies and sustainable mobility patterns. Electric vehicles and sharing systems are suitable options, although both require some refinement to incentivise a broader range of customers. Other aspects could be more difficult to resolve and might even postpone the generalisation of automated driving. For instance, there is a need for cooperation and management strategies geared towards traffic efficiency. Also, for transportation and land-use planning to avoid negative territorial and economic impacts. Above all, safe and ethical behaviour rules must be agreed upon before AVs hit the road.

SVRGSA: a hybrid learning based model for short-term traffic flow forecasting

Abstract: Accurate and timely short-term traffic flow forecasting is a critical component for intelligent transportation systems. However, it is quite challenging to develop an efficient and robust forecasting model due to complex non-linear data pattern of traffic flow. Support vector regression (SVR) has been widely employed in non-linear regression and time series prediction problems. However, the lack of knowledge of the choice of hyper-parameters in the SVR model leads to poor forecasting accuracy. In this study, the authors propose a hybrid traffic flow forecasting model combining gravitational search algorithm (GSA) and the SVR model. The GSA is employed to search optimal SVR parameters. Extensive experiments have been conducted to demonstrate the superior performance of the proposal.

Fatigue driving recognition network: fatigue driving recognition via convolutional neural network and long short-term memory units

Abstract: Fatigue driving has become one of the major causes of traffic accidents. The authors propose an effective method capable of detecting fatigue state via the spatial-temporal feature of driver's eyes. In this work, the authors consider fatigue detection as image-based sequence recognition and an end-to-end trainable convolutional neural network with long short-term memory (LSTM) units is designed. First, the authors apply a deep cascaded multi-task framework to extract eye region from infrared videos. Then the spatial features are learned by deep convolutional layers and the relationships between adjacent frames are analysed via LSTM units. Finally, through authors' model, a sequence-level prediction for driving state is produced. The proposed method achieves superior accuracy over the state-of-the-art techniques on authors' own dataset. Experimental results demonstrate the feasibility of authors' method.

Potential of wireless power transfer for dynamic charging of electric vehicles

Abstract: Wireless power transfer (WPT) offers a viable means of charging electric vehicles (EVs) whilst in a dynamic state (DWPT), mitigating issues concerning vehicle range, the size of on-board energy storage and the network distribution of static based charging systems. Such charge while driving technology has the capability to accelerate EV market penetration through increasing user convenience, reducing EV costs and increasing driving range indefinitely, dependent upon sufficient charging infrastructure. This study reviews current traction battery technologies, conductive and inductive charging processes, influential parameters specific to the dynamic charging state as well as highlighting notable work within the field of WPT charging systems. DWPT system requirements, specific to the driver, vehicle and infrastructure interaction environment are summarised and international standards highlighted to acknowledge the work that must be done within this area. It is important to recognise that the gap is not currently technological; instead, it is an implementation issue. Without necessary standardisation, system architectures cannot be developed and implemented without fear of interoperability issues between systems. For successful deployment, the technologies impact should be maximised with the minimum quantity of infrastructure and technology use, deployment scenarios and locations are discussed that have the potential to bring this to fruition.

Statistical‐based approach for driving style recognition using Bayesian probability with kernel density estimation

Abstract: Driving style recognition plays a crucial role in eco-driving, road safety, and intelligent vehicle control. This study proposes a statistical-based recognition method to deal with driver behaviour uncertainty in driving style recognition. First, the authors extract discriminative features using the conditional kernel density function to characterise path-following behaviour. Meanwhile, the posterior probability of each selected feature is computed based on the full Bayesian theory. Second, they develop an efficient Euclidean distance-based method to recognise the path-following style for new input datasets at a low computational cost. By comparing the Euclidean distance of each pair of elements in the feature vector, then they classify driving styles into seven levels from normal to aggressive. Finally, they employ a cross-validation method to evaluate the utility of their proposed approach by comparing with a fuzzy logic (FL) method. The experiment results show that the proposed statistical-based recognition method integrating with the kernel density is more efficient and robust than the FL method.

Integrated model predictive and torque vectoring control for path tracking of 4-wheel-driven autonomous vehicles

Abstract: In this study, an integrated path tracking control framework is proposed for the independent-driven autonomous electric vehicles The proposed control scheme includes three parts: the non-linear model predictive path tracking controller, the lateral stability controller, and the optimal torque vectoring controller Firstly, the upper bound speed limit is regulated based on the known curvature and adhesion coefficient of the road to prevent the tyre saturation The model predictive controller generates the steering angle and the desired longitudinal force for path tracking Simultaneously, the lateral stability controller calculates the desired yaw moment to balance the vehicle stability and motility under different situations Finally, the optimal torque vectoring controller distributes the wheel torques to generate the desired longitudinal force and yaw moment Three test cases are designed and verified based on a Carsim/Simulink platform to evaluate the control performance The test results illustrate that the proposed control framework has satisfactory path tracking performance, and the desired balance between vehicle mobility and stability is achieved under different road conditions

Investigating long-term vehicle speed prediction based on BP-LSTM algorithms

Abstract: Vehicle speed prediction is quite essential for many intelligent vehicular and transportation applications. Accurate on-road vehicle speed prediction is challenging because individual vehicle speed is affected by many factors related to driver-vehicle-road-traffic system, e.g. the traffic conditions, vehicle type, and driver's behavior, in either a deterministic or stochastic way. Also machine learning makes vehicle speed predictions more accessible by exploring the potential relationship between the vehicle speed and its main factors based on the historical driving data in the context of vehicular networks. This study proposes a novel data-driven vehicle speed prediction method based on back propagation-long short-term memory (BP-LSTM) algorithms for long-term individual vehicle speed prediction along the planned route. Also Pearson correlation coefficient is adopted to analyse the correlation of driver-vehicle-road-traffic historical characteristic parameters for the enhancement of the model's computing efficiency. Finally, a real natural driving data in Nanjing is used to evaluate the prediction performance with a result that the proposed vehicle speed prediction method outperforms other ones in terms of prediction accuracy. Moreover, based on the predicted vehicle speed, this work studies and analyses its effectiveness in two scenarios of energy consumption prediction and travel time prediction.

Trajectory tracking and prediction of pedestrian's crossing intention using roadside LiDAR

Abstract: Trajectory tracking and crossing intention prediction of pedestrians at intersections are important to intersection safety. Recently, on-board video sensors have been developed for detection of pedestrians. However, both the detection range and operating environment of video-based systems seem to be constrained by the advancement of image-processing technologies. Additionally, on-board systems cannot alarm pedestrians to take evasive actions when at risk, a feature which is critical to saving lives. This paper summarises the authors' practice on using roadside LiDAR sensors to monitor and predict pedestrians' crossing intention, as part of an ongoing effort to develop a pioneering LiDAR-based system to systematically reduce pedestrian and vehicle collisions at intersections. The LiDAR sensors were installed at intersections to collect pedestrian data such as presence, location, velocity, and direction. A new method based on deep autoencoder - artificial neural network (DA-ANN) was used to process data and predict pedestrian crossing intention. The case study shows the proposed model is about 95% prediction accuracy and computational efficiency for real-time systems. The roadside LiDAR system has great potential to significantly reduce vehicle-to-pedestrian crashes both at intersections and non-intersection areas, either used as a stand-alone system or in conjunction with the connected V2I and I2V technologies.

Effects of ACC and CACC vehicles on traffic flow based on an improved variable time headway spacing strategy

Abstract: An improved variable time headway (VTH) spacing strategy for the adaptive cruise control (ACC) and cooperative ACC (CACC) system is proposed. On the basis of the novel strategy, the typical two-modes of ACC/CACC upper-level controller are redesigned. Numerical simulations for two traffic scenarios are performed to verify the efficiency of the improved strategy. The results demonstrate the suitability and advantages of the improved VTH strategy in comparison with the constant time headway strategy and the VTH strategy. Furthermore, the authors study the impact of ACC and CACC vehicles on traffic flow by multiple-types mixed scenarios: ACC/manual vehicles, CACC/manual vehicles and ACC/CACC/manual vehicles. The results illustrate that introducing the ACC/CACC vehicles into mixed traffic can improve traffic flow stability, enhance road capacity and alleviate the increasingly serious traffic congestion problem.

CB-MAC: a novel cluster-based MAC protocol for VANETs

Abstract: Due to relative mobility, topology changes rapidly with frequent link breakage in vehicular ad hoc networks (VANETs). Clustering VANETs into small groups limits channel contention and controls the network topology efficiently. In this study, a novel cluster-based medium access control (CB-MAC) protocol is proposed for VANETs. The cluster formation process is defined. Moreover, cluster head election and cluster merging processes are described for efficient communication in the cluster as well as out of the cluster. The mechanism defined in IEEE 802.11 standard is specially designed for only direct communications and is not suitable for cluster-based communications. Therefore, new control packets are introduced and the existing control packet format is modified to support cluster-based communications. For effective MAC protocol design, the request to send (RTS)/clear to send (CTS) mechanism is not used for safety messages which are of broadcast nature. On the other hand, the RTS/CTS mechanism is used for non-safety data delivery to eliminate hidden node problem. Markov chain model-based analytical model is presented to explore the performance of the proposed CB-MAC protocol. The proposed protocol is validated by numerical studies. The numerical results exhibit that the proposed CB-MAC protocol improves system performance and satisfies the delay constraint of 100 ms for safety messages.

Towards a fully automated car parking system

Abstract: Currently, payment at most car parking-lots is carried out in the following manner: a ticket machine at entrance prints a ticket with a bar-code for each car entering the parking-lot. This ticket is later scanned at a payment terminal to determine the amount to be paid. The procedure can be automated using face detection and recognition technology. This automation can help with the issue of ticket loss/car theft. This study describes an automated car parking system. The proposed system consists of a camera installed at the entrance/exit of the parking-lot. Frames are continuously acquired by the camera. If there is a detected face, it is registered in the database. When a driver is leaving, the face image is captured again at the exit of parking-lot and compared in the database to conclude the identity. The system at the parking entrance/exit is composed of the following processing modules: (i) image acquisition, (ii) vehicle and face detection, and (iii) feature extraction that also includes a feature comparison/classification module for face recognition. The authors propose suitable algorithms for each module and carry out ad-hoc experiments to check the feasibility of the proposed system.

Direct-yaw-moment control of four-wheel-drive electrical vehicle based on lateral tyre–road forces and sideslip angle observer

Abstract: Considering some technical and economic reasons, it is not easy to directly measure the vehicular moving parameters (such as tyre-road forces and vehicle sideslip angle) in electronic stability programme systems. This study proposes a method to estimate lateral tyre-road forces and vehicle sideslip angle by utilising real-time measurements, based on the unscented Kalman filter. Direct-yaw-moment control can effectively guarantee the stability of vehicle while steering at a high speed. This study proposed a hierarchical control strategy as the solution to the problem of the yaw-moment distribution. The overloop controller is designed to calculate the desired yaw moment based on the estimated lateral tyre-road forces and sideslip angle, using the sliding mode control. The servo-loop controller is designed to optimise the torque distribution using weighted-least-squares method based on the desired yaw moment obtained from the overloop controller. MATLAB/Simulink with Carsim is applied for the simulation experiment, the results demonstrate the effectiveness of the lateral tyre-road force and sideslip angle observer, and the optimal allocation controller could improve the handling stability and energy efficiency dramatically.

Dynamic integration and online evaluation of vision-based lane detection algorithms

Abstract: Lane detection techniques have been widely studied in the last two decades and applied in many advance driver assistance systems However, the development of a robust lane detection system, which can deal with various road conditions and efficiently evaluate its detection results in real time, is still of great challenge In this study, a vision-based lane detection system with dynamic integration and online evaluation is proposed To increase the robustness of the lane detection system, the integration system dynamically processes two lane detection modules First, a primary lane detection module is designed based on the steerable filter and Hough transform algorithm Then, a secondary algorithm, which combines the Gaussian mixture model for image segmentation and random sample consensus for lane model fitting, will be activated when the primary algorithm encounters a low detection confidence To detect the colour and line style of the ego lanes and evaluate the lane detection system in real time, a lane sampling and voting technique is proposed By combining the sampling and voting system system with prior lane geometry knowledge, the evaluation system can efficiently recognise the false detections The system works robustly in various complex situations (eg shadows, night, and lane missing scenarios) with a monocular camera

Automatic measurement of the traffic sign with digital segmentation and recognition

Abstract: Traffic sign detection assists in driving by acquiring the temporal and spatial information of the potential signs for road awareness and safety. The purpose of conducting research on this topic is introduced to a novel and less complex algorithm that works for traffic signs identification, accurately. Initially, the authors estimate the global threshold value using the correlational property of the given image. In order to get red and blue traffic signs, a segmentation algorithm is developed using estimated threshold and morphological operations followed by an enhancement procedure, the net outcome of which is provided the greater number of potential signs. Moreover, remaining regions are filtered in terms of statistical measures using the non-potential regions. Furthermore, detection is performed on the basis of histogram of oriented gradient features by employing the support vector machine (SVM)- K -nearest neighbour (KNN) classifier. The denoising approach with the weighted fusion of KNN and SVM is used in order to improve the performance of the proposed algorithm by reducing the false positive. A recognition phase is performed on the GTSRB data set in order to formulate the feature vector. The proposed method performed the significant recognition with an accuracy rate of 99.32%. It is quite comparable to the existing state-of-the-art techniques.

Deep learning-based helmet wear analysis of a motorcycle rider for intelligent surveillance system

Abstract: Automatic helmet wear analysis of a motorcycle rider is a promising video surveillance application, as the helmets are indispensable for saving the lives of humans from head injuries during road accidents. This article presents an intelligent video surveillance system for automatically detecting the motorcyclists with and without safety helmets. If the motorcyclists are found without the helmet, his/her license plate (LP) number is recognised to initiate further actions such as deduction of penalty amount from one's account linked with the vehicle license and Aadhar Number (Applicable to Indian Scenario) by the traffic police and the legal authority. First, the foreground objects are segmented, using Gaussian mixture model (GMM) and then labelled. Afterwards, the proposed system adapts faster region-based convolutional neural network (faster R-CNN) for the detection of motorcycles in the labelled foreground objects to ensure the presence of motorcyclists. Subsequently, faster R-CNN is also used for the detection of the motorcyclists with and without helmet. Finally, the LP number of the motorcyclists without the helmet is recognised, using character-sequence encoding CNN model and spatial transformer (ST). The proposed framework is evaluated, using the performance metric and mean average precision (mAP) on the surveillance dataset and it outperforms the state-of-the-art algorithms.

Beacon trust management system and fake data detection in vehicular ad-hoc networks

Abstract: A vehicular ad-hoc network is an emerged technology, where vehicles communicate with one another and Road Side Units. Vehicular ad-hoc networks are categorised as a sub-branch of mobile ad-hoc networks, and they help in enhancing traffic efficiency and safety. Vehicular ad-hoc networks facilitate communication by dissemination of messages relaying through vehicles. However, malicious nodes can propagate false safety alert in a network due to its malicious behaviour or selfishness. False messages in vehicular ad-hoc networks can change drivers behaviour and create a disastrous situation in the network. Therefore, it is more important to thwart these false messages. The article presents a beacon trust management system and fake data detection (BTMS-FDD) scheme. The trust management system uses speed and density information to establish a relationship with vicinity vehicles. The false safety event detection scheme uses relative safety and beacon messages to detect false safety event more accurately. For false safety events, a scheme uses position and speed information of beacon and safety-related messages. To analyse the performance of proposed detection scheme, it is compared with ELIDV and host IDS schemes. The simulation results demonstrate that proposed malicious data detection scheme is better in terms of true positive, false positive and overhead.

Comprehensive chassis control strategy of FWIC-EV based on sliding mode control

Abstract: Four-wheel independent control electric vehicle has possessed tremendous potentials because the enhancement of driving performance and energy savings can be simultaneously carried out by independent and precise driving/braking/steering control. The study has proposed a comprehensive control strategy aiming at all normal conditions, which employs hierarchical architecture to reach the above-mentioned control. In the high-level controller, sliding mode control scheme is developed to figure out total force and yaw moment. In the low-level controller, energy-efficiency optimisation allocation is presented to reduce motor power losses and obtain energy recovery based on motor efficiency map, and then steering angle allocation is conducted to decrease the lateral force so as to reduce power losses caused by the tyre sideslip. Considering insufficient motor braking torque during large deceleration or even larger, the blended brake control strategy with the motor brake and electric hydraulic brake and further anti-skid brake system control via adopting fuzzy logic method are carried out. The torque and pressure are gained to deliver the corresponding actuators model established according to their physical characteristics. Through CarSim-MATLAB/Simulink-AMEsim co-simulation, results suggest that the developed strategy can boost the vehicle manoeuvrability and reduce energy consumption generated by motors and tyre sideslip under all the conventional occasions.

Fatigue driving detection model based on multi-feature fusion and semi-supervised active learning

Abstract: Fatigue driving is one of the main factors of traffic accidents and there are many research efforts focusing on fatigue driving detection. With the extensive use of on-board sensors, a huge number of unlabelled driving data can be easily collected, however, it is a costly and laborious work to annotate semantic labels for these data manually, posing some difficulties to detect fatigue driving with these data. In this work, the authors propose a novel fatigue driving detection model based on multi-feature fusion and semi-supervised active learning. In the authors' model, the steering features of the vehicle and the facial features of the driver are fused to improve the accuracy and stability of the model. Semi-supervised active learning algorithm allows us to make semantic labels for only a small number of data that can be propagated to the rest data, and help us establish an efficient fatigue driving detection model with automatic label propagation. Some experiments are conducted to validate their model, the results show that the accuracy is 86.25%, which proves the effectiveness of the fatigue driving detection model.

Human–machine shared control for vehicle lane keeping systems: a Lyapunov-based approach

Abstract: In this work, a human-centred steering assist controller based on dynamic allocation of control authority between driver and automatic e-copilot has been proposed for lane keeping systems Cooperative control between driver and steering assist controller is addressed taking into consideration human driving behaviour The vehicle steering controller for lane keeping is designed using a driver model for representation of the conflict between the driver and the controller The steering controller is designed employing the integrated driver-vehicle model using Takagi-Sugeno control technique coupled with Lyapunov stability tools The proposed design is robust to longitudinal speed variations and involves a trade-off between the lane following performance and ratio of negative system interference The proposed approach was implemented on dynamic vehicle simulator SHERPA and the results presented in this study demonstrate the effectiveness of the proposed structure for cooperative control action between human driver and the steering assistance system Based on indices such as energies spent by driver, driver satisfaction level and contradiction level between driver and autonomous controller the proposed optimal approach shows 9348% and 8930% reductions in expended driver energy and contradiction levels Further, the satisfaction level of driver increased by 6780% while performing a lane change manoeuvre

Hardware and software architecture of intelligent vehicles and road verification in typical traffic scenarios

Abstract: Intellectualisation is one of the three reforming technologies in automotive industry, which is now changing the mobility mode and human society. High safety and intelligence are the pre-requisites for putting self-driving vehicles into markets. This study presents the hardware and software architecture for intelligent vehicles, as well as their road verification in typical traffic scenarios. The hardware system includes environmental sensors, computing platforms, vehicle actuators, and vehicle platforms, which is able to provide redundant protection against the main controller failure. The software system includes environmental perception module, scene cognition module, decision and control module, human–computer interaction module and public service support module. To evaluate the performance of the developed architecture, the road tests of automated driving system were carried out in two typical traffic scenarios, including: (i) closed road test in Yuanboyuan region; (ii) open road test on Beijing-Tianjin highway. The real road test shows that the designed hardware and software systems for intelligent vehicles have desirable robustness, which can realise accurate and reliable environment perception, decision-making and motion control.

Design and optimisation of an In-wheel switched reluctance motor for electric vehicles

Abstract: To improve the mechanical performance of the In-wheel switched reluctance motor (SRM) used in electric vehicles (EVs), structure design and geometrical multi-objective optimisation strategy for the In-wheel SRM were developed in this study. The design method for major parameters of the In-wheel SRM was presented by means of design specifications of the EVs. According to requirements of the EVs, four indicators of the SRM were defined to evaluate the development of the SRM and perform the optimisation. To simultaneously improve the static performance of the SRM, a novel multi-objective simultaneous optimisation function was proposed by using four weighted factors and considering sensitivity analysis of the design variables on indicators. A four-phase 16/20 In-wheel SRM with an outer rotor was designed and optimised based on the proposed multi-objective optimisation method. The influence of design variables on average torque, torque ripple, efficiency, and torque density was analysed based on a combination of finite element analysis and orthogonal experiment design method. The static and dynamic torque performances of the optimised SRM were evaluated and compared with those of the initial motor. The comparison results showed that the proposed multi-objective simultaneous optimisation strategy can greatly improve the static and dynamic torque performances of the SRM.

Fast velocity trajectory planning and control algorithm of intelligent 4WD electric vehicle for energy saving using time-based MPC

Abstract: For intelligent four-wheel-drive (4WD) electric vehicle (EV), the vehicle speed can be planned and controlled for energy saving based on the slope information of road ahead. To reduce the calculation load of the optimisation algorithm, the model predictive control (MPC) method is formulated based on the time horizon in this study. Furthermore, a fast gradient method based control tool-GARMPC is used to solve the optimisation problem. First, the longitudinal dynamics model of 4WD EV based on time horizon and distance horizon is established based on the road slope information, respectively. Second, the MPC problem based on the time-discrete model is formulated and solved by GARMPC tool. For comparison, a dynamic program (DP) control method is introduced based on the distance-discrete model. Finally, the simulation is conducted under a designed road condition and a real measured road condition. The results show that the time-horizon based MPC method can significantly reduce the energy consumption compared with the proportion integration differentiation control method, which is similar to the driver's operation. Compared with the DP optimisation method, the time-based MPC method reduces the calculation time to smaller than 1 ms, which is essential for real-time application in a road vehicle.

Multi-model ensemble for short-term traffic flow prediction under normal and abnormal conditions

Abstract: Accurate traffic flow prediction under abnormal conditions, such as accidents, adverse weather, work zones, and holidays, is significant for proactive traffic control. Here, the authors focus on a special challenge of how to develop robust responsive algorithms and prediction models for short-term traffic forecasting in different traffic conditions. To this end, this study presents an ensemble learning algorithm for the short-term traffic flow prediction via the integration of gradient boosting regression trees (GBRT) and the least absolute shrinkage and selection operator (Lasso). Four different model structures whether considering the feature selection are proposed and tested for multi-step-ahead prediction under both normal and abnormal conditions. The results indicate that the proposed multi-model ensemble models are superior to the benchmark algorithms, i.e., support vector regression, and random forests, the GBRT model outperforms the Lasso model under normal traffic conditions, and the Lasso model has a better prediction accuracy under abnormal traffic conditions. In addition, the Lasso model with the feature selection is superior to the full feature model under either normal or abnormal conditions, while the GBRT model is not always better under normal conditions. The proposed integration framework is general and flexible to assemble various traffic prediction algorithms.

SVM-based hybrid approach for corridor-level travel-time estimation

Abstract: The objective of this study is to develop an accurate model for corridor-level travel-time estimation. Different approaches, such as k-nearest neighbour (k-NN), gradient boosting decision tree (GBDT) and support vector machines (SVMs), were used in this study. Further, this study also developed a hybrid model combining a data-driven approach (SVM) and a model-based approach [particle filter (PF)] for corridor-level travel-time estimation. Both static and dynamic parameters, such as road geometry, intersection length, location information from Global Positioning System devices, dwell time etc. were used as influential factors for modelling. The proposed algorithm was tested on a study corridor of length 59.48 km, in the arterials of Mumbai, India. The data was collected using a probe-vehicle technique for five days during the morning peak period (from 8.00 am to 11.00 am) for two modes (car and bus). The mean absolute percentage error values obtained for the hybrid model for the two modes were: 9.96 (car) and 11.24 (bus). The performance of the proposed hybrid (SVM-PF) algorithm showed a clear improvement in accuracy in comparison to existing standard methods such as k-NN, GBDT and SVM.

Driver identification in intelligent vehicle systems using machine learning algorithms

Abstract: This study proposes an applicable driver identification method using machine learning algorithms with driving information. The driving data are collected by a 3-axis accelerometer, which records the lateral, longitudinal and vertical accelerations. In this research, a data transformation way is developed to extract interpretable statistics features from raw 3-axis sensor data and utilise machine learning algorithms to identify drivers. To eliminate the bias caused by the sensor installation and ensure the applicability of their approach, they present a data calibration method which proves to be necessary for a comparative test. Four basic supervised classification algorithms are used to perform on the data set for comparison. To improve classification performance, they propose a multiple classifier system, which combines the outputs of several classifiers. Experimental results based on real-world data show that the proposed algorithm is effective on solving driver identification problem. Among the four basic algorithms, random forests (RFs) algorithm has the greatest performance on accuracy, recall and precision. With the proposed multiple classifier system, a greater performance can be achieved in small number of drivers' groups. RFs algorithm takes the lead in running speed. In their experiment, ten drivers are involved and over 5,500,000 driving records per driver are collected.

Gated branch neural network for mandatory lane changing suggestion at the on-ramps of highway

Abstract: A gated branch neural network (GBNN) is proposed for modelling mandatory lane changing (MLC) behaviour at the on-ramps of highways. It provides a core algorithm for an MLC suggestion system for advanced driver assistance systems (ADAS), where the main challenge is the trade-off between computational speed and prediction accuracy for both non-merge and merge events. The GBNN algorithm employs a gated branch based on correlation analysis, scaled exponential linear units activation function, and adaptive moment estimation optimiser. The algorithm has been evaluated using the real-world dataset of U.S. Highway 101 and Interstate 80 from Federal Highway Administration's Next Generation Simulation (NGSIM). Input features are extracted from NGSIM and pre-processed by standardisation and principal component analysis. TensorFlow framework and Python are used as the development platform. Results show that the proposed GBNN algorithm with the Pearson correlation method has values of 97.7%, 96.3%, and 0.990 for non-merge accuracy, merge accuracy, and receiver operating characteristic score, respectively. It outperforms other traditional binary classifiers for MLC applications, and is more light-weight than a convolutional neural network (AlexNet) of deep learning algorithm. Owing to its compact architecture, the GBNN provides high accuracy and efficiency, demonstrating promising usage as an MLC suggestion system in ADAS.