In the field of autonomous vehicles, accurately predicting steering angle and speed is a pivotal task. This task affects the accuracy of the final decision of the autonomous vehicle and is the basis for ensuring the safe and efficient operation of the autonomous vehicle. Previous studies have often relied on data from only one or two modalities to make predictions for steering angle and vehicle speed, which were often inadequate. In this paper, the authors propose a Multi-Modal Fusion-Based End-to-End Steering Angle and Vehicle Speed Prediction Network (MFE-SSNet). The network innovatively extends the one-stream and two-stream structure to a three-stream structure and cleverly extracts features of images, steering angles, and vehicle speeds using HRNet and LSTM layers. In addition, in order to fully fuse the feature information of different modal data, this paper also proposes a local attention-based feature fusion module. This module improves the fusion of different modal feature vectors by capturing the interdependencies in the local channels. Experimental results demonstrate that MFE-SSNet outperforms the current state-of-the-art model on the publicly available Udacity dataset.

The design and testing of autonomous vehicles crucially depend on an understanding of human driving characteristics due to the coexistence of automated and human-driven vehicles. Previous studies had relied on personalized background vehicle models, which required either calibrating microscopic traffic model parameters or training supervised models. This paper proposes a method that considers human driving behavior as a collection of some empirical driving patterns. A pattern dictionary generation algorithm is presented that distinguishes real-world driving data into pattern representations in continuous space. Driving patterns extraction based on transformer (DPFormer), a self-supervised deep learning-based method, is proposed to embed high-dimensional driving data into a compact form without manually labeled driving maneuvers. The model is optimized using the pretext task which utilizes attention mechanisms to embed the current driving segment as the posterior embedding of driving context. The embedding representation facilitates the analysis of diverse driving behaviors. Furthermore, a driving characteristics analysis method is proposed based on the Wasserstein distance between driving patterns' transition probability matrices to quantify the differences between different drivers. Finally, the effectiveness of this proposed model is validated using a large amount of real-world driving data collected from 171 human drivers.

Optimizing the efficiency of traffic flow while minimizing fuel consumption is of significant importance in the context of resource scarcity and environmental preservation. Currently, the two-layer optimization strategy has been employed in autonomous intersection cooperation problems. The traffic efficiency is optimized on the first layer and energy consumption is optimized on the second layer based on an optimal timetable gained in the first layer. This operation prioritizes traffic efficiency over energy consumption, which may present a limitation in terms of equilibrating them. This paper develops an equilibrium control strategy for autonomous intersection. This control strategy includes vehicle schedule and equilibrium control. A schedule algorithm is initially proposed for platoons, in which the most passing sequence is gained when considering platoon formation. Then, a game deep reinforcement learning model is designed, and an equilibrating control algorithm is proposed, in which equilibrium state can be gained between traffic efficiency and energy consumption. Simulation results demonstrate that the proposed method can equilibrate traffic efficiency and energy consumption to an equilibrium state, as well as reducing trip cost compared with existing methods.

The highway on-ramp merging area is a common bottleneck prone to traffic congestion and accidents. With the current trajectory and advancements in automotive technology, intelligent vehicle infrastructure cooperative control based on connected and automated vehicles (CAVs) is a fundamental solution to this problem. While much existing research focuses solely on ramp merging control in single-lane highway scenarios, there is more than one main lane in the actual highway environment. Thus, this paper proposes a dual-clutch longitudinal-lateral cooperative planning model, inspired by the principle of dual-clutch transmission, to address this gap. Besides, considering the impact of communication delay on control effects within the internet of vehicles, the paper proposes a system delay prediction model, which integrates the adaptive Kalman filter algorithm, the elitist non-dominated sorting genetic algorithm based on imitation learning, and the radial basis function neural network. The delay predicted dual-clutch on-ramp merging control model (DPDM) applied to two-lane highways for CAVs makes up of these two models above. Then, the performance of the DPDM is analyzed under different traffic densities on two-lane highways through simulation. The findings underscore the DPDM's pronounced comprehensive advantages in enhancing group vehicle safety, expediting and stabilizing merging processes, optimizing traffic flow speed, and economizing fuel consumption.

Vehicle-following control systems represent the key autonomous driving technology, which can enhance the road utilization and relieve traffic congestion effectively. However, their stability is severely challenged by random disturbances in the external environment, parameter perturbations to their dynamic models, and the combinations of these effects. In the most adverse scenario, these effects can cause fatal traffic accidents. Therefore, focusing on the vehicle-following system design, the μ-synthesis robust control framework is proposed in this paper for autonomous vehicles to realize its speed- and spacing- tracking control that can adapt to uncertainty. The contributions of this paper are threefold: First of all, the external disturbance affecting the longitudinal motion of the vehicle is regarded as an uncertain parameter and incorporated as a variable in the dynamic control model. This effectively reduces the complexity of the control system and improves the real-time performance of the controller. Then, the state-space model of the vehicle-following system is reformulated using the descriptor form to achieve decoupled robust control of autonomous vehicles with multiple uncertain parameters. This can reduce the conservativeness of the controller. Last, with consideration of the nominal performance and robust stability, a μ-synthesis robust controller for the speed- and spacing-tracking by a vehicle-following system is developed. Experiments using the hardware-in-the-loop system are conducted to verify the effectiveness of the proposed contol framework. The results show that it has good tracking performance and is robust to parameter perturbations.

This paper presents a hybrid control approach, termed Model Predictive Backstepping Control, for steering angle manipulation. The design of stepping manifolds in the backstepping control incorporates the Lyapunov function, posing a challenge in determining optimal stepping parameters for each manifold. In contrast to the predominant focus on constant stepping coefficients in existing research, this study explores the less investigated variable approach. Stepping parameters are introduced as tunable variables and incorporated into a backstepping control law, computed using model predictive control with the backstepping control law using tunable variables as the system input. The hybrid control algorithm utilizes prior knowledge of the control system to identify optimal values for the stepping parameters. The paper comprehensively explores model predictive backstepping control applied to a steer-by-wire system, specifically addressing the resolution of variable stepping parameters through a cost function. The developed algorithm is implemented into a steering control unit and subsequently validated in a real-world steering test vehicle. The results demonstrate successful realization of angle following in the steer-by-wire system within an engineering practice context.

In this paper, trajectory planning for autonomous vehicle is investigated based on the model prediction control to plan a safe, comfortable and efficient driving path. Firstly, an expanded safe zone of the three circles is proposed to simplify the planning environment. Secondly, the process of trajectory planning is decoupled into lateral planning and longitudinal planning. The model prediction method is utilized in lateral planning to generate a series of candidate trajectories. In addition, the dynamic programming algorithm of unequal distance scattering points is used in longitudinal planning, which significantly improves the planning efficiency. Finally, the cost functions of path planning and speed planning are proposed to find the optimal driving path and speed in the current environment. The collision detection link is increased, which guarantees the absolute safety of trajectories. Simulation results show that the proposed approach is able to generate satisfactory lane change trajectories for automatic lane change maneuvering.

To improve the safety of the intended functionality (SOTIF) performance of autonomous driving systems, this paper proposed a design method for autonomous driving systems with uncertainties. The automatic emergency braking (AEB) system is taken as an example to demonstrate the methodology. Firstly, uncertainty parameters in the AEB system model of typical working scenarios are defined and quantified, and a stochastic model of the AEB system with uncertainty parameters is established. Subsequently, the Monte Carlo simulation is employed to ascertain the actual safety distance distribution characteristics of the AEB system with uncertainties. The variance and width of the actual safety distance distribution are taken as response values to measure the reliability and robustness of the AEB system. The Box–Behnken design method is employed to design the uncertainty combination simulation test schemes. The surrogate models of uncertainty parameters with response variance and distributed width are established respectively, and the significance analyses are conducted. Finally, based on the variance surrogate models, the impact of uncertainties on the AEB system reliability and robustness is analyzed. This analysis provides the basis for the design of AEB system sensors. Based on the distributed width surrogate model, a dynamic safety distance adjustment mechanism is established to adjust the theoretical safety distance according to different uncertainties, thereby improving the reliability and robustness of the AEB system with multiple uncertainties. The method proposed in this paper provides a new idea for solving the SOTIF problems for autonomous driving systems.

The Cloud-Based Predictive Cruise Control (CPCC) system obtain road and vehicle data from the Cloud Control Platform (CCP) and efficiently computes the optimal speed and trajectory for intelligent connected vehicles (ICVs). It holds significant potential for conserving vehicle energy and minimizing waiting times. However, existing research on CPCC often overlooks the practical feasibility from theory to application and lacks effective validation regarding the consistency and rationality between theoretical architecture and model application. In this paper, a co-simulation method for CPCC is proposed. Firstly, the overall architecture of the CPCC system is proposed, outlining the real-time acquisition of road information and the speed planning strategy for ICVs. Then, models and algorithms for road and vehicles within the CPCC system are introduced. Finally, a CPCC co-simulation platform is established to validate the collaborative feasibility of the CPCC architecture, models, and algorithms. Simulation results show that compared to scenarios without CPCC, the CPCC-controlled ICVs experience a 2% reduction in cruising time and a 14.7% decrease in fuel consumption. Additionally, ICV co-control simulation results indicate a 4.05% decrease in average queue length at intersections, a 4.58% reduction in average vehicle delay time, an 8.99% decrease in average number of stops, a 12.89% reduction in average vehicle energy consumption, a 7.81% decrease in CO emissions, an 8.08% reduction in NOx emissions, and a 7.46% decrease in VOC emissions.

Accurate and robust range estimation algorithms for battery electric vehicles have the potential to reduce range anxiety, increase the acceptance of lower-range vehicles, and improve the overall driving experience. However, developing such algorithms faces challenges due to the complexity of the driver-vehicle-environment system and the multitude of factors influencing a vehicle's energy demand. To address these challenges, this paper introduces a sensitivity analysis focused on driver- and environment-related factors, which are notably difficult to predict. Employing a global sensitivity analysis for factor prioritization, this study delineates and assesses the parameters and their value distributions using a validated vehicle simulation model. The co-simulation of a powertrain and an auxiliaries model enables the parameter-specific investigation of parameters related to the thermal system. The results are scenario-individual parameter rankings that show the importance of the considered factors in prediction algorithms and guide the strategy for the development of these algorithms. The acceleration behavior of the driver, often emphasized in literature, is shown to be of secondary importance to energy consumption. Moreover, factors such as air density and wind speed are identified as crucial in highway driving scenarios, whereas outside temperature and the probability of stopping at traffic lights are critical in urban settings. For validation purposes, the resulting rankings of the sensitivity study are validated by means of a convergence analysis.

Vehicle-as-a-Conveyor (VaaC) is a hypothetical partially-assembled battery electric vehicle (BEV) that self-guides through the automotive assembly processes using sensor skids temporarily attached underneath the BEV. The VaaC embeds a rechargeable energy storage system (RESS), so this innovation can use the vehicle’s on-board energy to move itself to multiple assembly stations without relying on expensive and energy-intensive transport systems such as conveyors. The VaaC can also supply the electrical power to manufacturing equipment such as assembly tools while vehicles are being assembled. Moreover, the VaaC mechanically powers devices to activate a passive motion based on the movement of one or more wheels to reposition the vehicle itself. However, these innovative features require significant consumption of the on-board energy, making efficient energy management a necessary condition to satisfy for the VaaC implementation. This paper introduces the concept of VaaC thoroughly and discusses its control methods such as early detection of RESS failure or damage at the manufacturing stage. In addition, this paper discusses the impact of vehicle route scheduling on the energy efficiency of VaaC.

The publication of this article unfortunately contained mistakes. Figure 13 was not correct. The corrected figure is given below.