Pingshu Ge

Intelligent vehicle path tracking is challenged by uncertain disturbances, such as modeling inaccuracies and external environmental influences, which will significantly compromise both the path tracking accuracy and stability. To address this, this paper proposes a fixed-time prescribed-performance (FTPP) path tracking control method based on adaptive neural network disturbance estimation. Firstly, a radial basis function neural network with an online-updated adaptive law is developed for real-time estimation of uncertain disturbances, effectively compensating for their impact within the control model. Subsequently, a backstepping controller with FTPP is designed by integrating a composite dynamic surface control method with finite-time control techniques. This approach not only enhances the system convergence rate but also mitigates the derivative explosion problem inherent in traditional backstepping, yielding a control law with adaptive disturbances compensation for precise steering control. Finally, based on Lyapunov stability analysis, the boundedness of the closed-loop signals is established under the given assumptions, and the lateral path tracking error is shown to remain within the prescribed-performance bounds under feasible initial conditions. CarSim-Simulink-based co-simulation results validate the effectiveness of the proposed control method in improving both path tracking accuracy and stability.

Camera-based 3D object detection has attracted widespread attention for autonomous driving applications. However, existing methods often lack effective feature screening mechanisms, resulting in an extremely low spatio-temporal signal-to-noise ratio in complex scenes. Specifically, distracting background projections, feature misalignment caused by dynamic objects, and frequent occlusions jointly lead to severe ambiguity and loss of object features. To alleviate these issues, we propose STRDet, a robust 3D object detection framework based on progressive spatio-temporal feature refinement. Firstly, we propose a semantic-guided context refinement module that explicitly suppresses background interference prior to the view transformation, thereby blocking noise propagation. Secondly, we design the differential-aware feature alignment and residual-based adaptive gated fusion modules, which leverage feature difference maps as motion saliency indicators to guide deformable alignment and employ gating mechanisms to selectively integrate historical motion cues, effectively resolving dynamic alignment failures and mitigating feature loss under occlusion. Extensive experiments on the nuScenes dataset demonstrate that STRDet effectively enhances feature purity and coherence, yielding significant improvements and achieving 46.95% mean average precision and 55.37% nuScenes detection score.

Level-2 (L2) driving automation still has many limitations, and drivers are responsible for L2 automation safety. This study investigates the impact of the criticality of takeover situations caused by longitudinal or lateral control system failures (critical, non-critical), the failure form (silent, alert), and non-driving related tasks (NDRTs) (no task, video tasks) on drivers’ physiological responses, visual behavior, and takeover performance. Sixty-eight participants completed experiments on a motion-based simulator. Results showed that participants were more challenged to recognize faded lane lines than stationary obstacles, resulting in higher accident rates. The failure form had limited effects on participants’ physiological responses and visual behavior, but the takeover request could guide participants to recognize takeover events earlier. Participants engaged in NDRTs were also responsible for monitoring the driving environment for safety. This dual task increased their visual workload and reduced their situation awareness during automation. Additionally, distracted drivers were more difficult to detect takeover events, which worsened takeover performance and increased accident rates. Specifically, 12.5% of distracted participants were in accidents, while 6.6% of those not engaged in NDRT. In addition, this study classified distracted participants into low, medium, and high-risk groups based on their takeover performance and examined the monitoring strategy and maximum eye-off-road time (max EORT). Results suggested that monitoring frequency and max EORT are more suitable for evaluating and classifying driver monitoring behaviors in L2 driving automation. The max EORT for the three groups were 6.2 s, 9.4 s, and 13.2 s, respectively. The findings of the study provide an important reference for engineers to design L2 automation systems and driver monitoring systems to assess distracted driver risk or predict their takeover responses.

Nowadays, most trajectory prediction algorithms have difficulty simulating actual traffic behavior, and there is still a problem of large prediction errors. Therefore, this paper proposes a multi-object trajectory prediction algorithm based on lane information and foresight information. A Hybrid Dilated Convolution module based on the Channel Attention mechanism (CA-HDC) is developed to extract features, which improves the lane feature extraction in complicated environments and solves the problem of poor robustness of the traditional PINet. A lane information fusion module and a trajectory adjustment module based on the foresight information are developed. A socially acceptable trajectory with Generative Adversarial Networks (S-GAN) is developed to reduce the error of the trajectory prediction algorithm. The lane detection accuracy in special scenarios such as crowded, shadow, arrow, crossroad, and night are improved on the CULane dataset. The average F1-measure of the proposed lane detection has been increased by 4.1% compared to the original PINet. The trajectory prediction test based on D²-City indicates that the average displacement error of the proposed trajectory prediction algorithm is reduced by 4.27%, and the final displacement error is reduced by 7.53%. The proposed algorithm can achieve good results in lane detection and multi-object trajectory prediction tasks.

For nonlinear systems, both the cubature Kalman filter (CKF) and square-root cubature Kalman filter (SCKF) can get good estimation performance under Gaussian noise. However, the actual driving environment noise mostly has non-Gaussian properties, leading to a significant reduction in robustness and accuracy for distributed vehicle state estimation. To address such problems, this paper uses the square-root cubature Kalman filter with the maximum correlation entropy criterion (MCSRCKF), establishing a seven degrees of freedom (7-DOF) nonlinear distributed vehicle dynamics model for accurately estimating longitudinal vehicle speed, lateral vehicle speed, yaw rate, and wheel rotation angular velocity using low-cost sensor signals. The co-simulation verification is verified by the CarSim/Simulink platform under double-lane change and serpentine conditions. Experimental results show that the MCSRCKF has high accuracy and enhanced robustness for distributed drive vehicle state estimation problems in real non-Gaussian noise environments.

High-speed and complex road conditions make it easy for vehicles to reach limit conditions, increasing the risk of instability. Consequently, there is an urgent need to solve the problem of vehicle stability and safety. In this paper, adaptive stability control is studied in BEVs driven by in-wheel motors. Based on the sliding model algorithm, a joint weighting control of the yaw rate and sideslip angle is carried out, and a weight coefficient is designed using a fuzzy algorithm to realize adaptive direct yaw moment control. Next, optimal torque distribution is designed with the minimum sum of four tire load rates as the optimization objective. Then, combined with the road adhesion coefficient and the maximum motor torque constraint, the torque distribution problem is transformed into a functionally optimal solution problem with constraints. The simulation results show that the direct yaw moment controller based on the adaptive sliding mode algorithm has a good control effect on the yaw rate and sideslip angle, and it can effectively improve vehicle adaptive stability control. In the optimal torque distributor based on road surface recognition, the estimated error of road adhesion is within 10%, and has a greater margin to deal with vehicle instability, which can effectively improve vehicle adaptive stability control.

Drivers are allowed to perform a nondriving-related task (NDRT) in the Level-3 (L3) automated driving, which inevitably promotes a variety of complex NDRTs. Hence, investigating the effect of NDRTs on drivers’ behavior is necessary to ensure a safe transition from automated to manual driving. The main aim of this study is to examine the effect of resource demands of NDRTs and its engagement time on drivers’ physiological responses, takeover performance, and subjective ratings. A total of 42 participants were recruited to conduct a driving simulator study. Results showed that visual and physical resource demands significantly increased participants’ response time and deteriorated their takeover quality. Engagement time significantly affected participants’ drowsiness and response time to the takeover request. In addition, participants rated the physical–visual–cognitive task as the most challenging task and the cognitive task as the least demanding task, which matches the takeover performance. Moreover, pupil diameter responds significantly to visual resource demands rather than physical resource demands. Physical resource demands had a significant tendency to elevate heart rate. This study suggested that tasks with a certain level of cognitive resource demand may be more appropriate for L3 automated driving. The findings of this study make it possible to use physiological responses to identify NDRT types or even predict drivers’ takeover performance in automated driving. Based on this, the time budget could be dynamically adjusted to ensure a safe transition.

The electronic differential control strategy for distributed electric vehicle (DEV) was proposed based on the optimum Ackermann steering model. To improve the stability and tracking accuracy of DEV when steering, as well as its adaptability to different working conditions, the ideal Ackermann steering model was optimized by introducing the tire slip angle correction coefficient. Electronic differential steering model was designed based on the optimum Ackerman steering. Speed controller based on PID was optimized by particle swarm optimization of BP network. The controller can achieve vehicle differential steering accurately and adaptive adjustment of PID control parameters online. Simulation results indicate that the proposed control strategy can achieve the stable differential steering under the condition of high speed and low adhesion conditions. The vehicle tracking accuracy can be improved and the influence of tire side angle on vehicle steering can be reduced under medium speed steering and high speed steering conditions.