Driver drowsiness detection is a critical service in road safety to significantly reduce vehicle accidents. However, existing solutions often prioritize detection accuracy while neglecting speed. This paper proposes a novel mechanism, named ALERT, based on Electroencephalography (EEG) signals for real-time and accurate drowsiness detection. ALERT consists of four phases: periodic driver monitoring, feature extraction, feature selection and drowsy driving detection. Specifically, a periodic collection model is introduced to monitor driver behavior, followed by extracting features from time, frequency, and time-frequency domains. To optimize performance, a Sigmoid-based Grasshopper Optimization Algorithm (SGOA) is proposed to select the optimal feature set with minimal redundancy. Finally, various classifiers are evaluated to identify the best performance. Simulation results on real EEG signals demonstrate that the system achieves an accuracy of up to 96
Accurate state of charge (SOC) estimation helps to maintain the battery running in a reliable and efficient way. The moving horizon estimation (MHE) algorithm is an efficient way to estimate SOC in the battery system with physical constraints and noises, but its huge computational burden hinders its online application. To reduce the burden, this paper proposes an explicit MHE method to offload the computationally complex optimization solution in the MHE. First, the complex and constrained optimization problem is solved offline by multi-parametric quadratic programming (mp-QP). Then, the explicit solution of optimal estimation is derived by piecewise affine (PWA) function, which is suitable for online estimation. The proposed explicit MHE method significantly reduces the online computational resources and enables real-time optimal SOC estimation. Finally, experiments and simulations validate the effectiveness of the explicit MHE method in ensuring estimation accuracy and saving computational cost. Results show the proposed method can achieve a max absolute SOC estimation error of 1.78
Rigorous safety evaluation of automated vehicles is crucial before widespread deployment. The rarity of safety-critical scenarios limits the efficiency of scenario-based safety evaluation. Accelerated evaluation methods using importance sampling offer promising solutions, though existing approaches struggle with local optima and inefficient search when constructing the importance sampling distribution in large parameter spaces. In this paper, a prior knowledge-guided genetic algorithm for importance sampling distribution (PKGA-ISD) is proposed to address these challenges. The PKGA-ISD extracts scenario risk-related prior knowledge through random forest-based variable importance quantification and statistical analysis, which leads to refined parameter search ranges. By constraining the population initialization within these refined ranges, high-quality initial populations focusing on promising parameter regions are created, enabling more effective exploration of large parameter spaces and faster convergence to the optimal distribution. Simulation validation in cut-in scenarios demonstrates that PKGA-ISD achieves improved evaluation efficiency with reduced test sample requirements while maintaining unbiased collision rate estimation. These results confirm the potential of integrating prior knowledge into importance sampling distribution construction for accelerated safety evaluation, facilitating the development and deployment of automated vehicles.
This paper presents an integrated decision-making and trajectory planning framework specifically designed for lane change maneuvers. Lane change trajectories for the host vehicle are generated using the Quadratic Programming by sampling lane change duration and longitudinal displacement. Subsequently, the problem is formulated as a non-cooperative Stackelberg game to model the competitive interaction between the host and obstacle vehicles. A utility function is proposed to effectively map the host vehicle’s various lane change trajectory options and the obstacle vehicle’s discrete acceleration possibilities. This facilitates the determination of optimal lane change strategies for the host and corresponding obstacle vehicle responses within a framework that maximizes multi-objective utility for each participant. Notably, the proposed method is designed to redirect the host vehicle to its original lane in the event of irrational behavior from the obstacle vehicle. Validation of the developed Stackelberg game against an open-source game solver confirms its reliability. Comparative experiments conducted using traffic scenarios from the HighD Dataset underscore the capacity of this integrated approach to effectively emulate human-like behaviors, thus highlighting its pragmatic utility. Lastly, the study delves into an analysis of cooperative and non-cooperative game solutions, providing valuable insights into real-world traffic dynamics.
This paper presents an advanced post-processing framework to improve the limitations of deep learning-based segmentation models in handling local details and some unexplained segmentation errors. The proposed approach involves the detection of contours from segmentation results, followed by the extraction of local regions surrounding these contours through mapping to the original image. Critical known features are then extracted from these regions, and adaptive weights are calculated based on their reliability across different scenarios to facilitate feature fusion. The optimization process employs an energy function comprising internal energy gradients and fused feature energy gradients. And a new optimization approach, adam with adaptive points removal and movement constraint, is proposed to make the final contour align with the image edges and be smooth, while reducing noise segmentation regions by introducing a distance constraint, thresholds for contour length, and feature gradient variance on contour point updates in Adam. To evaluate the efficacy of the proposed framework, extensive experiments were conducted. When integrated with three types of models, it significantly enhanced segmentation quality. Furthermore, the method exhibited superior enhancement in comparison to other post-processing mechanisms. In real-world tests conducted on an autonomous driving platform, significant performance improvements were observed in campus and open park scenarios, highlighting the robustness and accuracy of the method.
Although without the capability to implement direct yaw moment technology, this popular dual-motor four-wheel drive (4WD) electric vehicle (EV) in the current market can provide an alternative solution for implementing indirect yaw-moment control (IDYC) by distributing the driving force between the front and rear axles to enhance vehicle cornering maneuverability. However, the effectiveness of IDYC to improve maneuverability is constrained by the driver’s sluggish torque request and restricted tire grip. To overcome this limitation and further enhance vehicle maneuverability, a novel longitudinal force control strategy is proposed by introducing G-vectoring control (GVC) on the basis of the IDYC. The strategy adopts the way of "GVC first, IDYC later" to control the longitudinal force distribution of the front and rear axles and the vertical load of the tire, so as to achieve the optimal balance of the tire grip utilization of the front and rear axles and the desired acceleration and longitudinal force distribution coefficient for the current driving conditions can be derived. Additionally, a feedforward plus feedback acceleration controller considering total disturbance estimation is developed to ensure precise tracking of the expected control variables. Typical U-turn and double-lane change maneuvers simulation tests demonstrate that the designed strategy can greatly improve the vehicle handling performance and make the vehicle dynamic steering characteristics more reaching to the steady-state steering characteristics. Compared with the case of no-control, the overall reaching degree is increased by 23
Managing mixed vehicle platoons, which integrate intelligent and connected vehicles and human-driven vehicles, presents significant challenges due to the uncertainties inherent in human driving behaviors. Although data-driven control techniques utilizing trajectory data have shown a potential to address these challenges, their performance is often undermined by inevitable noise and external disturbances. To address this limitation, we propose a tube-based robust data-driven predictive control (TRDDPC) framework to enhance the robustness of mixed vehicle platoons. The framework begins by constructing a matrix zonotope set from data, which provides an over-approximation of system dynamics under the influence of noise and disturbances. By decoupling the over-approximated system into nominal and error subsystems, a data-driven minimal robust positively invariant set is used to encapsulate the impact of disturbances and noise. The TRDDPC framework then formulates an optimization problem to compute robust control inputs, which are implemented through a tube-based control mechanism. Simulation results show that for a mixed vehicle platoon with three vehicles, TRDDPC achieves reductions of 30.6 % in velocity error and 26.7 % in spacing error compared to data-enabled predictive control in comprehensive scenarios, with further reductions of 30.6 % and 32.3 % in emergency scenarios. Moreover, TRDDPC decreases computation time by 90.6 % , significantly enhancing the robustness and efficiency of mixed vehicle platoon control.
Battery fault diagnosis is crucial to ensure the safe and reliable operation of electric vehicles (EVs). Cell inconsistency within a battery pack can trigger battery faults during long-term usage, and thus cell consistency evaluation is vital to identify battery faults in the early usage stages. This paper proposes a battery inconsistency evaluation method that combines the Shannon entropy and Cluster in QUEst (CLIQUE) algorithm to diagnose faulty cells. First, the charging fragments extracted from real-world EV operating data are used to calculate the normalized Shannon entropy (NSE) for each battery cell. Then, the CLIQUE clustering algorithm is utilized to identify the NSE outliers in a charging fragment. Finally, the NSE outlier proportion of each charging fragment is extracted and employed as an indicator for battery fault diagnosis. The real-world EV datasets with two different types of battery faults are applied to examine the effectiveness of the proposed scheme. The results show that the proposed method can effectually perform battery fault diagnosis and identify faulty battery cells before thermal runaway.
Fuel cell hybrid electric vehicles (FCHEVs) offer a promising solution for sustainable transportation. However, effective energy management remains a challenge due to the non-linear dynamics of the powertrain and the need to balance performance with the health of the vehicle’s energy storage system (ESS). This study proposes a novel adaptive differential evolution-based frequency separation energy management system (ADE-FS-EMS), featuring adaptive mutation control to dynamically calibrate frequency parameters to optimize load distribution among the proton exchange membrane fuel cell (PEMFC), lithium-ion battery, and supercapacitor (SC). The proposed EMS minimizes PEMFC fuel consumption while maintaining the state of charge (SOC) for both the battery and SC. Furthermore, by intelligently shifting high-frequency transient demands to the SC, the proposed EMS significantly reduces stress and degradation of both the PEMFC and battery. A comprehensive robustness analysis, including hardware-in-loop (HIL) testing, validates the real-time feasibility of the proposed EMS. Key findings demonstrate that the proposed EMS achieves a significant reduction in hydrogen fuel consumption, enables nearly 40
As a crucial component linking the perception and decision-making modules of automated vehicles, vehicle trajectory prediction plays a key role in improving driving safety and efficiency. In complex driving environments, simply extrapolating future trajectories from historical states is insufficient for accurate prediction, as vehicle interactions can induce significant trajectory variations. In this paper, a graph-guided vehicle trajectory prediction method inspired by bidirectional dynamic interaction is proposed. Given the cyclic coupling of interactions, undirected graphs are constructed based on distance and velocity. Multilayer graph convolutional networks (GCNs) are employed to extract bidirectional interaction features among vehicles at each time step. Building upon this, a temporal correlation extractor based on gated recurrent units (GRUs) is established, extending single-step interactions to continuous-time interactions and capturing dynamic interaction representations. Furthermore, to leverage the improved trajectory prediction accuracy enabled by bidirectional dynamic interactions, relevant features are mapped into a latent space via a conditional variational autoencoder (CVAE), enabling the modeling of continuous and complex driving behaviors. The proposed method achieves an 11.2
The conflict between human drivers and machines in shared control vehicles poses a daunting challenge and research focus. A hierarchical human-machine adaptive cooperative control strategy based on the dynamic game theory is proposed to tackle this issue. This strategy integrates an advanced intention recognition model, an intelligent trajectory planning method, and adaptive dynamic interaction mechanisms to enhance vehicle safety and mitigate human-machine conflicts. The strategy is composed of three layers: intention recognition layer (IRL), trajectory planning layer (TPL), and tracking control layer (TCL). IRL trains a CNN-GRU deep learning model utilizing the data obtained from the driver-in-the-loop platform for intention recognition. This model leverages the feature extraction capabilities of convolutional neural networks (CNN) and the temporal data processing advantages of gated recurrent units (GRU) to accurately identify the driver’s decision intention. Based on the recognized intention, TPL incorporates an improved risk potential field and employs a linear time-varying model predictive control (LTV-MPC) algorithm to plan a dynamic trajectory that aligns with the driver intention while ensuring safety. TCL designs an adaptive weight distribution mechanism under a dynamic game framework to achieve dynamic interaction of human-machine. The time-varying characteristics of the driving scenario and driver operations are considered in TCL. Based on the convex iterative method, the Nash equilibrium solution is obtained to realize the cooperative control. Hardware-in-the-loop experiments are utilized to validate the proposed strategy’s advantage. The results demonstrate that in complex scenarios, cross-layer dynamic synergy can effectively mitigate human-machine conflicts and enhance driving safety.
The sensorless control is a key technology of permanent magnet synchronous motor. The core contribution of this work is twofold: the design of an angular-velocity-based proportional-integral observer for rotor position estimation, and the inaugural derivation of the relationship between q-axis current error and angular velocity. This derived relationship allows the method to be applied universally across a wide speed range. Based on the rotation axis voltage model, the current error between the estimated model and the real model is utilized to estimate the angular velocity, which is used to compute the rotor angle through integration. Furthermore, the Lyapunov stability principle is employed to prove the stability of the algorithm. The algorithm is implemented in a motor control system and tested in a motor test bench, which achieves fast and precise speed control of a permanent magnet synchronous motor. Taken together, the experimental findings and statistical evaluations validate the angular-velocity-based proportional-integral observer as a compelling candidate for sensorless algorithmic control in permanent magnet synchronous motors.
The dual-winding steer-by-wire system represents advanced steering technology, where fault-tolerant control is essential to satisfy the redundancy requirements of autonomous driving. When one winding experiences an open-phase fault, the fault-tolerant control of the dual-winding steer-by-wire system encounters two primary challenges: model mismatch and increased system uncertainty. To address these issues, this paper proposes an event-triggered variable tube-based actively reconfigured model predictive fault-tolerant control strategy. The strategy incorporates event-triggered mechanism, actively reconfigured faulty model, adaptive unscented Kalman filter, and variable tube-based model predictive controller. The actively reconfigured faulty model is designed to minimize copper loss, fully utilize the faulty winding and reduce current harmonics and torque ripple. The adaptive unscented Kalman filter dynamically observes the fault coefficient matrix and the increased system uncertainty. The variable tube-based model predictive controller switches to the actively reconfigured faulty model as the prediction model via the fault event-triggered mechanism and adjusts the tube shape based on the adaptive unscented Kalman filter results to improve fault-tolerant tracking performance. The error event-triggered mechanism reduces the control frequency, conserving communication and computational resources. Hardware-in-the-loop test results demonstrate that the proposed control strategy enhances tracking performance, and the event-triggered mechanism effectively conserves communication and computational resources without compromising control performance.
Vehicles currently on the road are characterized by a growing level of automation. To improve safety standards and solve disputes on accident responsibilities, policy-makers have established sets of guidelines related to the monitoring of the automated vehicle (AV) behaviour, which requires logging and reporting of a set of data acquisitions, including environment perception and decision-making. Given the rising numbers of AVs, this paper provides a methodology to estimate the amount of data generated by logging and recording operations considering the following factors: AV adoption levels, sensing system architecture, decision-making pipelines, user familiarity with AVs, and point of data extraction along the vehicle sensing pipeline. The method is tested on a European scale, but can be replicated on broader scenarios. The results show the impossibility of uploading raw sensor output, due to the required bit-rate and storage memory, which are not sustainable by current infrastructure and technology. Upload of processed and annotated data in place of raw sensors output would downsize the volumes involved by several orders of magnitude, allowing for real time upload to the cloud. However, the lack of an annotation standard leads to greater uncertainty in the total volume of generated data, making cloud storage design more difficult for authorities. This demonstrates the stringent need for a data annotation standard to efficiently perform monitoring of AVs.
Distributed drive electric vehicle (DDEV) is experiencing rapid progress in development and industrialization, with different configurations to integrate numerous components in the chassis space. The mechanical and electrical connection network of all the components consists of the core element of system configuration, which should be considered systematically and comprehensively. Different from the traditional engineering drawing, this paper creatively introduces the graph theory concept as a systematic analysis methodology for DDEV configurations. By the position of the drive motor, the existing company products and scholarly research configurations can be classified into three types. They are: (i) DDEV with centralized motor and half shaft, (ii) DDEV with close-to-wheel motor and speed-reducer mechanism, and (iii) DDEV with in-wheel motor configuration. Furthermore, DDEV’s future trends are discussed, such as corner modules for chassis-by-wire technology, wireless power transfer technology, magnetic gear machine technology, and dynamic vibration absorber applications. Finally, the various DDEV configurations are evaluated and presented in the radar charts with conclusions. For the research of DDEV technology, this paper offers theoretical support for system analysis, configuration synthesis methodology, and technical direction prospection in industrial applications.
To mitigate range anxiety, the energy density of lithium-ion batteries has been continuously improved. In ternary lithium batteries, adding nickel in cathodes or increasing the voltage efficiently boosts the energy density but also escalates the thermal runaway (TR) risks. However, how these factors affect battery material reactions under abuse and overall battery behaviour remains unclear. In this study, the TR characteristics and thermochemical reaction behaviours of battery materials in high-voltage systems (LiNi0.6Mn0.2Co0.2O2 | Gr) and high-nickel systems (LiNi0.9Mn0.05Co0.05O2 | Gr) with similar energy densities are compared and analyzed. The results indicate that the decomposition temperature of LiNi0.9Mn0.05Co0.05O2 is lower than that of LiNi0.6Mn0.2Co0.2O2, resulting in a lower TR trigger temperature (T2) for the high-nickel system battery. Furthermore, the release of more oxygen from LiNi0.9Mn0.05Co0.05O2 promotes a more thorough and intense reaction with the anode, leading to a higher peak temperature (T3) during TR. The gas venting behaviour of the high-nickel system battery is also more severe than that of the high-voltage battery. These results indicate that batteries with high-voltage systems show better thermal safety. Therefore, for ternary lithium batteries, increasing the nickel content in the cathode significantly reduces the battery’s safety performance, and increasing the battery’s operating voltage appears to be a promising design direction for high-energy–density batteries. This study provides guidance for the material design of high-energy–density batteries.
Accurately estimating electrical machine temperatures, particularly the magnet temperature, remains a challenging task. Data-driven approaches can be a new method if they offer highly accurate temperature predictions with compact models suitable for real-time embedded systems. This paper presents an investigation on thermal neural networks, a data-driven modeling approach. The main goal of this study is to investigate the benefits and drawbacks of this approach. Firstly, the model architecture and the input dataset from simulations in the ANSYS MotorCAD design software are presented, followed by the training routine. The initial analysis evaluates the best number of nodes for the model, already indicating its strong performance, with a maximum absolute error of the permanent magnet temperature of 1.37 K on an unknown validation dataset. The result of this first analysis is the base model with four internal nodes. This study then explores methods to enhance the physical interpretability of this thermal neural network. Firstly, a sparsing technique reduces the model’s parameters by removing connections that lack physical meaning. Secondly, in an experiment the ambient node is eliminated and a virtual node with constant temperature is introduced to isolate the influence of the ambient node. This model with virtual node has a maximum absolute error of the permanent magnet temperature of 2.75 K and offers faster training with fewer parameters than the base model. The paper finally outlines key steps toward creating a more diverse training dataset, ensuring real-world applicability and usability for temperature-based control systems in permanent magnet synchronous machines.
This paper presents a novel two-level Q-learning algorithm designed to improve path planning and collision avoidance in autonomous electric vehicle (EV)-charging systems. The proposed method leverages point cloud data from a single camera to construct a 3D grid-based representation of the environment. Fuzzy logic is used to dynamically determine the number of grid cells, ensuring sufficient resolution to handle minor misalignments during the charging process. The typical Q-learning approach is enhanced with a guided error mechanism and adaptive step size, which adjusts exploration strategies and optimizes step size based on real-time feedback. This adaptation allows the robot to avoid high-risk areas near obstacles, significantly improving obstacle avoidance and boosting the success rate of the charging process. Through extensive experimentation on multiple EV models, the proposed method demonstrated reduced computational time and enhanced robustness, achieving a success rate of 95
The accurate prediction of the remaining useful life (RUL) for lithium-ion batteries (LIBs) is crucial for the effective management of electric vehicle energy systems. However, the prediction accuracy and adaptability are frequently compromised by the phenomenon of capacity regeneration. To surmount this challenge, a novel RUL prediction methodology is introduced for LIBs that integrates variational mode decomposition (VMD) with deep learning techniques. Initially, the LIBs capacity data are decomposed at multiple scales using VMD to extract the intrinsic mode functions (IMFs) that represent the stochastic fluctuations in battery capacity and residual (RES) components that characterize the global degradation trend. The RES trend and each IMFs are modelled using a deep belief network (DBN) and a Bayesian optimized gated recurrent unit (GRU) network, respectively. The resultant predictions are subsequently combined to estimate the final RUL. Finally, the observed data is used to test the predictive performance of the proposed model. The results demonstrate that the proposed model achieves high accuracy and strong robustness for RUL prediction compared with other alternative prediction models, with the average relative error strictly controlled within 0.5
