An integrated approach for vehicle trajectory planning and tracking control is proposed in this paper, aiming at enhancing the ability to avoid obstacles by autonomous vehicles when they are in a complex dynamic environment. And the method combines the improved driving risk field model with model predictive control, and integrates speed planning, trajectory planning, and tracking control into a multi-objective optimization problem. First, considering the impact of dynamic traffic conditions on driving risk, an improved driving risk model is constructed by introducing the time to collision (TTC) indicator as well as the parameters of vehicle intrinsic properties and motion state attributes. Next, a dynamic vehicle model of three degrees of freedom (3DOF) is established. Based on the model predictive control (MPC) algorithm, the corresponding cost function and constraint conditions are designed by considering the multi-dimensional objectives of vehicle stability, comfortableness, driving safety, etc. Finally, different traffic scenarios were set up for simulation verification using the Simulink + Prescan + Carsim joint simulation platform. A comparative analysis was conducted between the integrated control method proposed in this study and the hierarchical control method and integrated control without considering TTC method, and the results indicate that the control method mentioned in this paper can achieve dynamic obstacle avoidance for autonomous vehicles in various traffic scenarios. And it can also achieve a more accurate and more stable driving trajectory and thus ensure the driving comfortableness and safety. And the improved driving risk field model mentioned in this paper draws a more accurate image of driving risks in a complex environment compared with the traditional driving risk field model which could improve the driving safety and applicability of autonomous vehicles in dynamic traffic environments.

In recent years, the autonomous driving technology has witnessed rapid global development, offering effective solutions to increasingly severe challenges related to traffic congestion and road safety. Among various paradigms, end-to-end autonomous driving has emerged as a promising alternative to traditional modular systems, due to its streamlined architecture, enhanced decision consistency, and superior generalization capabilities. This survey provides a comprehensive review of the evolution and core technologies of end-to-end autonomous driving. It emphasizes the applications and developments of imitation learning (IL), reinforcement learning (RL), and other paradigms. Furthermore, it highlights the emerging paradigm empowered by foundation models, such as large language models (LLMs) and vision-language models (VLMs), and systematically categorizes recent advances in planning, reasoning, data generation, and scene understanding. In light of persistent challenges, such as multimodal fusion complexity, low sample efficiency, and safety risks, this survey summarizes the representative research efforts and corresponding solutions. Finally, it outlines future directions, including the integration of world models (WMs) for unified data generation and inference optimization, the advancement of foundation model architectures toward modular design, sparse activation mechanisms, and knowledge distillation, and the realization of deployable, transferable, and unified multimodal frameworks. This survey aims to serve as a comprehensive theoretical and technical reference for researchers in end-to-end autonomous driving, facilitating its evolution toward higher performance, stronger generalization, and enhanced safety assurance.

Distributed cooperative control enhances the handling and dynamic response of autonomous vehicle, effectively addressing dynamic coupling effects among multiple actuators while reducing overall system complexity, which enables globally optimized cooperative control for X-by-wire chassis systems. So, this paper proposes a novel distributed coordinated control strategy based on reinforcement learning (RL) to dynamically balance the active front wheel steering system (AFS) and direct yaw moment control system (DYC). First, distributed state equations are formulated, and the information exchange process among agents is established. Second, the control objectives of each subsystem are defined and a cost function incorporating terminal and control constraints is designed. Then, the stability index is redefined by introducing yaw rate constraints into the phase plane, which together with subsystem tracking performance is used for the RL reward designed. This enables dynamic adjustment of the AFS and DYC control weights, achieving more efficient stability control while satisfying the trajectory constraints. Finally, co-simulation and hardware-in-loop (HIL) experiments are conducted to evaluate the control performance and experiment results validate the effectiveness of the proposed strategy for coordinated vehicle chassis control.

This paper presents a unified framework for airborne target perception, designed for unmanned aerial vehicles (UAVs) operating in non-cooperative airspace environments. The core contribution to artificial intelligence lies in the integration of electro-optical and infrared (EO/IR) sensors using a convolutional sparse representation-based image fusion algorithm, along with a novel spatiotemporal detection method that combines conditional random fields and motion history analysis. The engineering application focuses on real-time airborne Sense and Avoid (SAA) capabilities for small UAVs, where a local-angle-based collision avoidance path planning method is proposed to address the limitations of monocular vision-based perception. To validate the proposed framework, a distributed digital simulation and verification system is developed based on virtual camera feeds and local network communication. This system supports closed-loop testing of visual perception, target detection, and path planning in realistic airspace environments. Experiments conducted in three representative airport scenarios — Illinois State Hospital, Shanghai Pudong International Airport, and New York John F. Kennedy International Airport — demonstrate the framework’s effectiveness in enhancing visual quality under low illumination conditions, improving detection accuracy, and enabling robust and safe autonomous navigation. Specifically, the proposed system achieves a target detection accuracy of 94.6% and reduces false alarm rate to 2.1%, while successfully generating collision-free paths in 97.8% of dynamic encounters. Compared to existing state-of-the-art EO/IR fusion-based perception systems, our framework improves detection precision by 4.3% on average and increases planning robustness by 5.6% in complex airspace environments. These results validate both the effectiveness and the generalizability of the unified framework for real-world UAVs SAA tasks.

Real-time temperature prediction that achieves high accuracy while maintaining computational efficiency is essential for the effective thermal management of Axial Flux Permanent Magnet (AFPM) motors. This paper proposes an adaptive reduced-order modeling approach for the Proper Orthogonal Decomposition Reduced-Order Model (POD-ROM) of AFPM motors, which accounts for time-varying electromagnetic-thermal coupled characteristics. Initially, a high-fidelity Lumped Parameter Thermal Network (LPTN) model is developed, with convective heat transfer coefficients identified via the Particle Swarm Optimization (PSO) algorithm. The LPTN model demonstrates a maximum steady-state absolute error below 1.4 °C, corresponding to relative errors under 2%, when compared to electromagnetic-thermal coupled finite element method (FEM) simulations. Subsequently, the LPTN model is reduced through POD to establish a low-dimensional thermal model. To enhance the adaptability of the conventional POD-ROM under varying operating conditions, a temperature-dependent adaptive thermal conductivity matrix is introduced. By integrating this adaptive matrix with a real-time updated heat source vector, a dual-parameter adaptive correction strategy addressing both thermal conduction and heat generation is developed. An experimental platform for motor temperature rise is constructed to validate the model’s performance. Experimental results indicate that, under constant operating conditions, the enhanced POD-ROM achieves a steady-state absolute error within 1.21 °C and reduces computation time to 0.74 seconds, representing an 18.5-fold acceleration relative to the full-order LPTN model. Under variable operating conditions, the maximum absolute error remains below 2.1 °C, with relative errors under 4%. This study offers an effective methodology for developing high-accuracy, real-time temperature prediction models to support the thermal management of AFPM motors.

To address the engineering challenges of controlling and allocating torque in the transient over-actuated system during mode transitions of compound power split hybrid electric vehicles, the paper proposes an intelligent optimal torque distribution control scheme. First, the mode transition logic of the over-actuated system is established by analyzing the timing and sequence of dual-clutch operations. Based on this logic, a hierarchical coordinated control strategy is developed. At the upper layer, a controller and observer are designed using finitetime theory, with speed-tracking errors as state inputs, to guarantee transient performance during mode transitions. At the lower layer, simulated annealing is integrated with optimal control allocation to ensure optimal torque distribution in the over-actuated system. Finally, simulation and hardware-in-the-loop test results demonstrate that, compared with the terminal sliding mode control method based on control allocation, the proposed approach reduces jerk, slippage work, and transition time by 19.1 %, 52.7 %, and 16.7 %, respectively. The control framework provides a novel intelligent solution for mode transition control in hybrid electric vehicles.

This study addresses the longitudinal control problem of a CAV convoy in a lead-follower communication topology, where both input delay and lead vehicle broadcast delay are present. Unlike existing research, which treats various types of delay uniformly, this study distinguishes the influence paths of broadcast delay and input delay at the system architecture level, thereby constructing a dual-delay convoy dynamics model with separate information and execution layers. Based on this, a hierarchical longitudinal control architecture based on a reference speed generation mechanism is proposed. By introducing a reference speed filter, this method decouples the lead vehicle’s broadcast information from the closed-loop error feedback channel, allowing the broadcast delay to participate in the construction of the control input as a smoothed reference signal. This achieves structural decoupling between the broadcast delay and the error feedback subsystem, effectively avoiding response jumps and high-frequency amplification issues caused by delayed information directly entering the feedback loop. Furthermore, based on Lyapunov stability theory, the stability of the error feedback subsystem under input delay is analyzed, and conditions for selecting control parameters are provided. Simulation results demonstrate that the proposed method effectively suppresses the propagation of errors in the convoy direction under perturbed conditions, with a headway error of less than 0.6 m, a 2.9% improvement in speed response, and earlier attainment of steady state.

This paper proposes a systematic approach to lateral stability analysis and coordinated control to address the issue of lateral instability in sport utility vehicles (SUVs) with air suspension systems under specific driving conditions. Firstly, a three degree of freedom dynamic model is established to accurately represent the lateral dynamic behavior of an air suspension SUV, taking into account the nonlinear mechanical characteristics of both air springs and tires. To analyze the stability of this high-dimensional, complex system, center manifold theory is employed for dimensionality reduction. Subsequently, a qualitative analysis of the reduced-order system is conducted, deriving the necessary and sufficient conditions for the system to undergo saddle-node bifurcation. Based on this, phase plane analyses are performed under four different operating conditions to reveal the mechanism of lateral instability in air suspension SUV under specific driving scenarios, providing a basis for weight distribution in the coordinated controller. The coordinated control strategy incorporates two sub-controllers: an active front steering sub-controller designed using a nonlinear model predictive control algorithm, and a direct yaw moment control sub-controller designed using the adaptive second-order integral terminal sliding mode algorithm. The two controllers cooperate to control according to the collaborative rules to maintain vehicle stability Hardware-in-the-loop simulation results demonstrate that the proposed coordinated control strategy, compared with traditional controllers, can effectively enhance the lateral stability of SUVs equipped with air suspension systems under various operating conditions.

Accurate motion prediction is critical for the safety of autonomous driving systems, as precise and diverse predictions enable downstream planning modules to better understand driving scenarios. However, existing methods typically struggle to capture the diverse behaviors of driving agents due to reliance on biased training data, which primarily records single-mode trajectories and inadequately reflects the complexity and variability of real-world scenarios. To address this issue, we propose HySA-Aug, a novel hybrid scenario-adaptive data augmentation framework aimed at enriching the behavioral distribution of scenarios and enhancing the multimodal representation learning capability of prediction models, thereby improving the diversity and accuracy of predicted trajectories. First, we construct a trajectory vocabulary from real-world driving scenes and implement a rule-constrained generation strategy to select augmented trajectory candidates through multi-level constraints, including road topology, interactions with adjacent agents, and self-motion consistency. Second, to further refine the quality of augmented trajectories, we introduce a data-driven trajectory scoring module that assesses the realism of candidate trajectories by modeling the feature distribution differences between real and augmented samples, effectively reducing the domain gap. A behavior partitioning strategy guided by scene topology is further proposed to improve the quality and diversity of training samples for the scoring module. Finally, we propose a training strategy that incorporates high-quality augmented trajectories into the prediction process through a multi-supervised regression loss, enabling efficient and lightweight integration of diverse behavioral patterns into existing prediction models. Experiments on various datasets, including Argoverse 1, Argoverse 2 and Waymo, demonstrate that the proposed augmentation framework improves behavioral diversity while maintaining high prediction accuracy.

To improve the stability of the CAVP in dynamic communication delay and vehicle transmission delay environments, a control method based on PLF (Predecessor-Leader Following) communication topology is proposed, which integrates multi-state dynamic information such as speed difference, acceleration difference, vehicle distance error between the lead vehicle, neighboring front vehicles, and the self vehicle. A robust controller with adaptive regulation for the CAVP is constructed, and the parameter boundary conditions of the controller are derived based on Lyapunov stability theory to ensure system stability. To avoid potential system oscillations during the parameterization process, the controller sets different regulation constants for the pilot car and the neighboring front cars, respectively, to further maintain system stability. The experimental results show that the controller has significant advantages in reducing the platoon spacing error and improving response speed, especially under the influence of time-delay interference. It can advance the platoons to reach stabilization time by approximately 16 s, demonstrating better stability and efficiency. This study not only improves the adaptability of platoon control but also provides new theoretical support and experimental basis for the application of networked automated driving technology in dynamic communication environments, which is an important impetus for the development of intelligent transportation systems.

Key vehicle parameters (KVPs), such as vehicle mass and center of gravity (CoG) position, play a very important role in vehicle control and state estimation. This article proposes a novel KVPs estimation method based on an adaptive unscented Kalman filter (AUKF) and intelligent tire technology. Global sensitivity analysis (GSA) is used to determine the optimal in-tire installation locations of PVDF sensors for tire vertical load (VL) estimation. Raw voltage signals are collected and preprocessed to construct training datasets. An optimized Gaussian process regression (GPR) algorithm is then employed to establish an accurate tire VL estimation model. On this basis, two key modules are further developed: 1) a VL update-rate compensation algorithm for intelligent tires, which mitigates the adverse effects of vehicle speed variations and asynchronous VL estimation among the four wheels on KVPs estimation performance; and 2) an AUKF framework incorporating compensated tire VL information, enabling accurate estimation of KVPs. Both simulation and real-vehicle experiments are conducted to evaluate the performance of the estimator. The results demonstrate that the proposed method achieves maximum mean absolute percentage errors (MAPEs) of 1.08%, 1.61%, 0.11%, and 2.07% for the estimation of vehicle mass, longitudinal, lateral, and vertical CoG positions, respectively. In addition, the estimator exhibits strong adaptability to model parameter uncertainties, such as the tire-road friction coefficient.

Indoor parking lots represent a critical scenario for the commercialization of intelligent vehicles. The absence of GNSS signals, coupled with complex environmental structures and frequent dynamic interference, makes high-precision localization a core and challenging technology for achieving autonomous parking in such settings. Existing research methods based on multi-sensor fusion often lack strong constraints from high-level behavioral decisions when addressing these challenges, leading to insufficient localization accuracy and robustness in complex dynamic environments. To address this, this paper proposes an innovative approach that deeply integrates the planned path for autonomous parking with multi-sensor data to achieve high-precision localization in underground parking lots. Firstly, the A* and DWA path planning algorithms are improved to generate smoother trajectories that better adapt to dynamic environments. Secondly, in a novel contribution, the real-time optimized velocity and steering angle information fed back from path planning is utilized as key observations. This information is deeply fused with LiDAR and IMU data and embedded into a Bundle Adjustment (BA) optimization framework. This integration compensates for the accuracy degradation associated with velocity estimation relying solely on IMU pre-integration in traditional methods. Experimental results demonstrate the excellent performance of the proposed method across multiple real indoor parking lots. The average localization accuracy stably reaches around 0.2 meters, and notably, achieves 0.2315 meters even in complex commercial underground parking environments, confirming its stability and reliability in complex and dynamic scenarios.

Flying cars integrate the capabilities of ground and aerial vehicles, offering a transformative solution to urban congestion and significantly improving transportation efficiency and flexibility. Existing reviews have primarily focused on control, energy systems, vehicle configurations, or regulatory issues, whereas a path-planning-centric synthesis that explicitly links regulatory and operational boundary conditions, platform-dependent propulsion–dynamics–energy mechanisms, and planning problem formulations and solution strategies is still missing. The inherently cross-domain nature of flying cars, characterized by three-dimensional operation, multimodal dynamics, multiple transition phases, and propulsion-energy coupling, leads to strongly nonlinear, state-dependent, and nonconvex constraints, greatly increases planning complexity and renders path planning approaches developed for other vehicle platforms not directly applicable. This review provides a comprehensive examination of flying car path planning. First, it systematically reviews major countries’ airspace regulations and takeoff requirements, highlighting their roles as boundary conditions for path planning. Second, it compares flying cars with automobiles, unmanned aerial vehicles (UAVs), fixed-wing aircraft, and helicopters in terms of propulsion systems, dynamic models, and energy architectures, and explains how these subsystems jointly shape planning feasibility and optimization objectives. Third, it introduces a structured taxonomy of flying car path planning problems, distinguishing common issues shared with ground and aerial vehicles from the unique challenges introduced by multimodal configurations and air-ground transition phases. This review establishes a unified analytical framework that links physical mechanisms with planning strategies, laying a foundation for intelligent and certifiable urban air-ground mobility.

With the intelligent transformation of China's shipping network, large navigation hub infrastructures, represented by the Yangtze River trunk line, have become highly complex and heterogeneous under the empowerment of digitalization. These networks face severe cybersecurity threats, posing a challenge to shipping and even national strategic security. To address this challenge, a security protection framework for cross-domain collaboration paradigm was proposed in this paper. The framework aimed to integrate protection technologies and strategies from different security domains to achieve tight collaboration and efficient response. The research background, architectural framework, key technologies, and future research priorities of cross-domain collaborative security protection technology for large navigation hub critical infrastructure were delved, providing feasible references for enhancing its security capabilities.

OBJECTIVE:To address the limited consideration of driving style diversity in group-level risk propagation analysis, a Dynamic Interaction Potential (DIP)-based framework is proposed to capture the risk evolution patterns of vehicles with different driving styles. METHODS:Vehicle groups are identified by DBSCAN using the vehicle trajectories from HighD. To describe vehicle interactions within groups, DIP is constructed based on motion conflict and behavioral uncertainty. Fluctuation intensity and fluctuation rate are derived from DIP for risk measurement and driving style classification. The roles of different styles in risk propagation are quantified using Network analysis. A comprehensive vehicle group risk index is proposed that incorporates style-specific effects. RESULTS:Risk propagation network analysis reveals the distinct roles of driving styles, identifying aggressive drivers as key risk initiators within vehicle groups. The proposed style-aware risk index (CRAI_Style) demonstrates superior predictive performance compared to baseline indicators. In the one-second prediction window, the ROC-AUC for CRAI_Style is 0.845, followed by its style-agnostic counterpart CRAI (0.745) and traditional conflict-based indicator TTC (0.675). CRAI_Style consistently maintains the highest predictive accuracy when the prediction window extends to three seconds. CONCLUSION:These findings reveal the differentiated roles of driving styles in risk propagation, where aggressive drivers act as key risk initiators and conservative drivers as risk suppressors, and demonstrate that incorporating this heterogeneity significantly enhances the accuracy and robustness of vehicle group risk assessment.

With the rapid adoption of new energy vehicles, the safety and stable operation of charging piles have become key factors limiting the promotion of electric vehicles. To address the issues of insufficient fault diagnosis accuracy and robustness under complex operating conditions, this paper proposes a fault diagnosis framework for charging piles based on the fusion of vehicle-side and pile-side information. The framework is systematically designed from three aspects: feature optimization, model fusion, and decision robustness. At the feature level, R2 regression and Principal Component Analysis (PCA) are employed for feature optimization to reduce dimensional redundancy and emphasize key information. At the decision level, an improved KNN, AP clustering, and LSTM model are used to integrate the initial diagnostic results and construct a multi-source evidence body. D-S evidence theory is introduced to merge the results of multiple models, effectively alleviating uncertainty in high-conflict scenarios while fully exploiting the complementary advantages of each algorithm. Experimental results based on real operational data show that the proposed method achieves a diagnostic accuracy of 95

Airborne small human detection over dynamic water surfaces is challenging in maritime search-and-rescue scenarios, because human targets are often extremely small and easily confused with waves, foam, and specular reflections. From an instrumentation and measurement perspective, this task can be regarded as a UAV-borne optical measurement problem, where clutter-induced false alarms and weak target responses directly affect sensing reliability. Although shallow attention enhancement improves sensitivity to weak targets, it may also amplify clutter-dominated background responses. To address this issue, this paper proposes Attention-Conditioned Spectral Modulation (ACSM) for airborne small human detection over dynamic water surfaces. The key idea is to introduce frequency-domain discriminability into attention-enhanced shallow features, so that structurally coherent target components and texture-dominated clutter responses can be more effectively disentangled. Specifically, shallow channel attention is first used to enhance weak target cues, and adaptive band-wise spectral modulation is then performed to suppress clutter-related frequency components while preserving target-related structural information. A residual correction mechanism is further introduced to integrate the refined representation into the baseline detector in a plug-and-play manner. Experiments are conducted on three public maritime human-detection datasets, namely SeaDronesSee, ManOverBoard, and AFO, together with a self-collected real-flight UAV dataset for deployment-oriented validation. On the challenging SeaDronesSee benchmark, ACSM improves AP50 from 0.504 to 0.520 and reduces FP/image from 0.632 to 0.504 over the ESOD baseline, while achieving the best precision and F1-score. On ManOverBoard and AFO, ACSM remains competitive and preserves a favorable balance between detection accuracy and false-alarm control. Feature-level analyses further verify that the proposed method alleviates shallow-layer representation entanglement and selectively suppresses clutter-dominated spectral responses. Overall, ACSM provides an effective and practical solution for improving the reliability of UAV-borne optical measurement for airborne small human detection in dynamic maritime environments.

Integrated active-passive safety plays an important role in improving vehicle crash safety, especially in hazardous scenarios where collision avoidance may become dynamically infeasible. Conventional active-safety strategies mainly focus on collision avoidance and may not fully account for occupant-injury outcomes once an impact becomes unavoidable. To address this issue, this study proposes a stability-constrained injury-aware Model Predictive Path Integral (SCI-MPPI) framework for integrated active-passive safety decision-making. A driving safety domain is constructed based on nonlinear vehicle dynamics, tire-force saturation, and braking-steering stability boundaries, and is embedded into trajectory planning as a dynamic feasibility constraint. High-fidelity crash simulations and THOR anthropomorphic test device (ATD)-based occupant-injury responses are used offline to train a physically partitioned radial basis function (RBF) surrogate model for rapid online Weighted Injury Criterion (WIC)-based injury prediction in frontal asymmetric collisions. SCI-MPPI then optimizes sampled trajectories while considering road boundaries, obstacle avoidance, vehicle-footprint constraints, stability limits, and injury-related costs. Simulation results show that SCI-MPPI generates collision-free trajectories and smooth dynamic responses in avoidable scenarios. Comparative evaluations against baseline methods indicate that SCI-MPPI shows a favorable balance between nonlinear trajectory-optimization performance and computational efficiency. In unavoidable-collision cases, SCI-MPPI achieves a mean per-case first-impact WIC reduction of 18.56% while keeping injury-oriented maneuvers within the driving safety domain. These results indicate that SCI-MPPI supports the coordination of active obstacle avoidance and first-impact injury mitigation within stability-constrained maneuvering limits, providing guidance for improving collision safety in future intelligent-vehicle systems.

Recent advances in cooperative driving have garnered significant attention, as it is believed to have greater potential than single-vehicle systems in complex environments. However, gaps in current research frameworks impede its further development. Dataset-based frameworks provide a limited perspective, failing to capture the dynamic, asynchronous, and interactive complexities of the real world. Simulation-based frameworks often employ synchronous implementation and oversimplified communication models, impeding meaningful real-time performance evaluation in asynchronous environments. To this end, we propose RAMACoDrive, a novel real-time asynchronous framework for cooperative perception research. RAMACoDrive implements complete process separation, allowing agents and their internal components to operate asynchronously and independently. It uniquely integrates hardware-in-the-loop V2X communication using an external device for realistic data forwarding. Furthermore, RAMACoDrive incorporates an asynchronous online evaluation method with real-time performance metrics, enabling comprehensive evaluation beyond mere accuracy. Through experiments centered on cooperative 3D object detection, we demonstrate its effectiveness in revealing the impact of real-world asynchronous effects and communication constraints. RAMACoDrive provides a valuable and realistic testbed for cooperative perception algorithms and helps to advance the deployment of practical applications. The code is available on our GitHub repository.