Decision-making for automatic vehicles at unsignalized intersections with dense traffic is one of the most challenging tasks. Due to the complex structure and frequent traffic accidents, traditional rule-based methods struggle to flexibly address this issue and often produce suboptimal strategies. Recently, deep reinforcement learning (DRL) has garnered significant attention for its exceptional performance in decision-making problems. We propose a local attention safety deep reinforcement learning (LA-SRL) decision-making method for ego vehicle right-turns at unsignalized intersections. LA-SRL enables paying attention to different states of social vehicles within complex traffic environments and effectively deals with the impact of surrounding vehicles on ego vehicle. This contributes to enhancement of safe driving efficiency. To further balance the safety and efficiency of decision-making for ego vehicle at unsignalized intersections with dense traffic flow, we design a safety-reward function composed of risk reward and avail reward. The safety-reward function enables ego vehicle to promptly navigate out of high-risk areas, meanwhile avoiding collisions and reducing waiting periods. Finally, we evaluate our method in the CARLA simulator. The results demonstrate that LA-SRL outperforms state-of-the-art methods, achieving a remarkable success rate of 98.25% and reducing average time to 6.6 seconds

Electric freight vehicles have become an important means of transportation in connected and automated environments owing to their numerous advantages. However, the generally short driving range of connected and automated electric freight vehicles (CAEFVs) does not satisfy the growing transport demand. In this study, wireless charging technology is employed to construct a complex driving scenario including urban roads and dynamic wireless charging facilities. A combination of variable-scale elements consisting of vehicles, roads, and the environment is analysed hierarchically to develop a wireless charging scheme for urban transport systems. Using passage efficiency, energy consumption, and passenger comfort as the joint optimisation objectives, an optimal driving model for CAEFVs in wireless charging scenarios at signalised intersections combining scenario boundaries and vehicle dynamic constraints is proposed. Considering the differentiated charging needs of vehicles, this model is divided into a time priority strategy (TPS), balance priority strategy (BPS), and charging priority strategy (CPS). The obtained results reveal that the CPS is superior to the TPS in terms of the charging benefits but requires a longer travel time. Meanwhile, the BPS increases the charging benefits and passing efficiency. This study provides guidance for the deployment of wireless charging lanes with a high application value.

Projection micro-stereolithography (P-μSLA) processes have been widely utilized in three-dimensional (3D) digital fabrication. However, various uncertainties of a photopolymerization process often deteriorates the geometric accuracy of fabrication results. A predictive model that maps input shapes to actual outcomes in real-time would be immensely beneficial for designers and process engineers, permitting rapid design exploration through inexpensive trials-and-errors, such that optimal design parameters as well as optimal shape modification plan could be identified with only minimal waste of time, material, and labor. However, no computational model has ever succeeded in predicting such geometric inaccuracies to a reasonable precision. In this regard, we propose a novel idea of predicting output shapes from input projection patterns of a P-μSLA process via deep neural networks. To this end, a convolutional encoder-decoder network is proposed in this paper. The network takes a projection image as the input and returns a predicted shape after fabrication as the output. Cross-validation analyses showed the root-mean-square-error (RMSE) of 10.72 μm in average, indicating noticeable performance of the proposed convolutional encoder-decoder network.

Connected and automated vehicles (CAVs) represent a significant development in the transport industry owing to their intelligent and interconnected features. Potential field theory has been extensively used to model CAV driving behaviour owing to its objectivity, universality, and measurability. However, existing car-following models do not consider the impact of time delays and the influence of information from multiple vehicles ahead and behind. This paper focuses on the driving-safety risks associated with CAVs, aiming to enhance vehicle safety and reliability during travelling. We developed a multi-vehicle car-following model based on safety potential fields (MIDM-SPF), taking into account the characteristics of multi-vehicle connected information and time delays. To enhance the model’s precision, real-world data from urban roads were employed, alongside an improved optimisation algorithm to fine-tune the car-following model. The simulation experiment revealed that MIDM-SPF significantly reduces stop-and-go traffic, thereby improving traffic flow stability in urban areas. Additionally, we validated the stability of our model under varying market penetration rates in large-scale mixed traffic. Our findings indicate that increasing the CAV proportion improves the stability of mixed traffic flows, which has important implications for alleviating traffic congestion and guiding the large-scale implementation of autonomous driving in the future.

Cellular Vehicle-to-Everything (C-V2X) Mode 4, specified by 3GPP, enables distributed resource reservation in vehicular networks under out-of-coverage conditions. However, highly dynamic traffic and complex road environments can lead to increased channel contention, packet collisions, and transmission delays, significantly degrading system performance. To address these challenges, this paper develops a theoretical model to analyze the impact of vehicle density on key performance parameters, and proposes a Vehicle Density-driven Adaptive Resource Reservation (VD-ARR) method to dynamically adjust reservation parameters and mitigate persistent collision issues. A two-dimensional discrete-time Markov chain (DTMC) model is constructed to derive steady-state probability expressions, enabling a quantitative evaluation of VD-ARR. Building on this foundation, a VD-ARR Guided Double Deep Q-Network (VG-DDQN) framework is developed, integrating the analytical insights into a reinforcement learning architecture to enhance resource allocation adaptability. Simulation results show that the proposed VD-ARR achieves lower latency and reduced collision probability under varying vehicle densities, while VG-DDQN outperforms VD-ARR in highly dynamic environments, offering superior adaptability and robustness.