The rapid expansion of shared micro-mobility services has intensified safety concerns in dense urban environments in recent years. Traffic complexity and infrastructure limitations increase accident risks, yet safety aspects are often overlooked by operators and existing decision-support systems. Current safety-oriented approaches mainly rely on static analyses, historical accident data, or infrastructure-based interventions, which often lack adaptability to real-time urban dynamics and offer limited practical deployment at the user level. This paper addresses these limitations through DRL-SMSafe, a deep reinforcement learning–based framework for improving real-time micro-mobility safety via smart parking recommendations, focusing on the initial and final segments of trips. The proposed approach dynamically identifies alternative safer drop-off zones at the moment of a user’s parking request by leveraging both static and dynamic, easily accessible safety-related factors. To preserve user convenience, recommendations are restricted to neighboring zones, minimizing the distance between the user’s initially selected location and the suggested alternative. Comparative experimental results, supported by a detailed analysis and examination of the outcomes against baseline methods, demonstrate the effectiveness of the proposed framework and its suitability for real-time decision-making.

The cooperative adaptive cruise control (CACC) system is pivotal in improving traffic safety and efficiency by maintaining precise inter-vehicle gaps and regulating speed. However, communication and actuator delays often coexist and significantly affect the stability and operational safety of vehicle platoon systems. Although extensive research has been conducted on both types of delays, most studies have constructed and analyzed each delay independently, lacking a unified mitigation strategy for their combined effects. Therefore, this study introduces a novel distributed hierarchical control architecture that integrates upper-level decision-making with lower-level execution control. The upper-level controller employs a model predictive control (MPC) architecture enhanced by an adaptive Kalman filter (AKF) to mitigate state discrepancies caused by communication delays in car-following scenarios. This AKF-MPC controller further improves the real-time estimation of surrounding vehicle states, resulting in a more accurate desired acceleration. A Proportional-Integral-Derivative (PID) feedback controller is implemented at the lower level to track the desired acceleration from the upper-level controller, simultaneously compensating for actuator delays and enhancing closed-loop vehicle performance. The proposed architecture is validated through comprehensive simulations under diverse traffic conditions, demonstrating its effectiveness in significantly improving CACC system stability and safety under communication and actuator delays. This study can contribute to establishing a unified control architecture for cooperative vehicle systems, thereby providing a theoretical foundation for deploying delay-resilient CACC in intelligent transportation systems with heterogeneous communication constraints.

In order to ensure the safety and stability of vehicle platoons, the spacing error between vehicles needs to be maintained within a given range and gradually converge to 0. Due to actuator saturations and time-varying uncertainties in vehicle dynamics, the vehicle platoon control remains a major challenge. This paper presents a robust control method for vehicle platoons, ensuring the spacing error performance within a compact prescribed band region. Firstly, an improved spacing modeling method is presented, embedded with an auxiliary variable to modify the spacing error. It effectively circumvents actuator saturations caused by excessively large initial errors. Then, bilateral constraints are imposed on the modified error to characterize collision avoidance and compact formation (CA&CF) performance, and subsequently the constrained error is transformed into an unconstrained state via potential function transformation. The uniform boundedness (UB) performance of the unconstrained state ensures the compact prescribed band region performance of the spacing error. Then, based on the constraint-following control theory, a robust controller is designed to ensure the system’s UB and uniform ultimate boundedness (UUB) performance. Finally, simulation results demonstrate the effectiveness of the proposed control.

In work zone traffic management, warning systems provide road users with real-time traffic conditions through messages or signals. This process aims to adjust driver speeds and headway distribution to affect the smoothness of traffic flow, ultimately achieving safer and more efficient traffic operations. This study examined the distributions of time headways for different vehicle types when traveling upstream of freeway work zones equipped with advanced warning systems. Data were collected through video footage and field observations at two locations on Interstate 80 in Nebraska. Statistical tests were performed to compare headway distributions during periods with active and inactive warning systems. The results indicated that, overall, the activation of the advanced warning systems altered the traffic headway distributions. Specifically: (1) the overall mean headway was significantly smaller under warning conditions compared to no-warning conditions; (2) the mean headway for passenger cars following other vehicles was significantly greater under warning conditions compared to no-warning conditions; and (3) all lead-follow pairs involving a heavy vehicle showed no significant difference in the mean headways with or without activated warning systems. This study also identified the best-fitting distribution models for empirical headways influenced by advanced warning systems. The parametric lognormal distribution and non-parametric Gaussian kernel distribution models were found to provide the best fit, depending on the type of vehicles in lead-follow pairs.

Continuous Flow Intersections (CFIs), also known as Displaced Left-Turn (DLT), are designed to improve traffic efficiency and safety at intersections with high left-turn demand. While the CFI traffic signal control design and optimization have been already studied in the literature, there is little research on how CFI traffic signal control strategies should be adapted in mixed autonomy traffic with connected and autonomous vehicles (CAVs) and human-driven vehicles (HVs). This study proposes a real-time signal optimization model for a full symmetric CFI in mixed autonomy traffic. The proposed model consists of a real-time saturation flow estimator that accounts for the time-varying penetration of CAVs in traffic that feeds into a two-stage optimizer. The study proposes a queue-based as well as a platoon-based saturation flow estimator in mixed autonomy traffic. The model is evaluated with extensive simulation runs with different demand scenarios and is compared against a state-of-practice benchmark CFI signal control strategy. Simulation results indicate that the proposed model reduces average delay by 14% and average queue length by 16% for a full CFI across various demand patterns when compared to the benchmark.

With increasing travel demand, highway congestion and accidents have become more frequent. As an essential component of intelligent transportation systems (ITS), Hard Shoulder Running (HSR) provides a dynamic strategy to mitigate congestion by temporarily opening shoulder lanes, yet traditional methods often fail to adapt to real-time traffic changes and overlook the shoulder’s critical role in ensuring emergency vehicle access. This study proposes a novel HSR optimization framework based on the Long Short-Term Memory (LSTM) and Multi-Agent Deep Deterministic Policy Gradient (MADDPG) reinforcement learning algorithm, integrated with an improved A* algorithm for EV lane clearing. LSTM is used to extract temporal features from traffic data to support intelligent decision-making in MADDPG, enhanced with prioritized experience replay and importance sampling. The EV lane-clearing task is formulated using graph theory, and the improved A* algorithm determines the optimal path for clearing. A simulation case was developed using Simulation of Urban MObility (SUMO) for a section of the Jinan City Ring Highway, China, evaluating four levels of traffic service. Results show the proposed method reduces total travel time by 14.6%, Time Integrated Time-to-Collision (TIT) by 45.2%, and CO2 emissions by 11.9%. Additionally, with EV intervention, braking times are reduced by up to 376.3% and travel time by 18.1% using the improved A* strategy. These findings demonstrate that the integrated LSTM-MADDPG and A* approach effectively enhances highway traffic efficiency, safety, and sustainability under complex real-world conditions.

The growing demand for rapid, same-day deliveries has made the optimization of last-mile Hyper-local Networks (HLN) increasingly critical. These networks face challenges such as limited storage capacity and highly variable customer demand, which complicates delivery planning and resource allocation. This study aims to address the HLN design problem under uncertain and fluctuating demand conditions, to enhance delivery efficiency and system robustness. To achieve this, we propose a novel meta-heuristic optimization algorithm: the Composite Learning Strategy Criminal Search Optimization Algorithm ( CLS − CSOA ), enhanced with a Time-varying Inertia Weight mechanism. This approach is designed to balance exploration and exploitation phases during optimization, thereby producing more efficient and adaptive solutions. We assess the method using a combination of benchmark functions and realistic last-mile delivery scenarios. Comparative experiments show that CLS − CSOA consistently outperforms conventional algorithms in solution quality, robustness, and computational efficiency. Statistical validation via the Wilcoxon Signed-Rank test confirms its superiority, with a notably low Type II error rate. Three real-world case studies, reflecting varying levels of customer demand, demonstrate the practical applicability of the model. The results show significant improvements in delivery performance and a reduction in computational time. Importantly, the algorithm maintains high performance under increasing demand variability, underscoring its scalability and adaptability. This research provides practical implications for urban logistics by offering a lightweight, high-performance alternative to data-intensive methods, supporting agile and cost-effective last-mile delivery systems, especially for small and medium-scale logistics providers.

How to enable autonomous vehicles to sustain high safety and rapid response capability in complex, dynamic environments remains a critical and urgent challenge. This study aims to elevate the learning efficiency of multi-sensor feature fusion in autonomous-driving tasks, thereby strengthening system safety and responsiveness. To this end, we present an innovative multi-modal, multi-sensor feature-fusion model that integrates two highly effective mechanisms (Sparse Channel Pooling and Residual Triplet-Attention) to refine the fusion process. First, the Sparse Channel Pooling mechanism allows the model to adaptively select salient feature channels while suppressing redundant information. Second, the Residual Triplet-Attention mechanism resolves inter-modal feature misalignment, enabling accurate cross-modal alignment of key features and markedly improving computational efficiency. Finally, a reinforcement-learning module is introduced to learn and optimize policies in continuous action spaces. Coupled with the feature-fusion framework, this module enables end-to-end, high-efficiency driving-policy learning within the CARLA autonomous-driving simulator. Experimental results demonstrate that the proposed method not only maintains safe operation in complex dynamic scenarios but also preserves real-time, high-level perception and decision-making accuracy, offering a novel paradigm for multi-sensor data fusion in autonomous driving.

With advancements in computational power and Vehicle-to-Infrastructure (V2I) technologies, modern traffic signal control (TSC) systems are designed in a more intelligent way by leveraging Connected Vehicle (CV) data. While prior research primarily focuses on CV positions and speeds, route data—which can provide foresight into future traffic demand—remains largely overlooked. To fill this gap, we propose a distributed TSC model based on the Soft Actor-Critic (SAC) reinforcement learning algorithm. Through V2I communication, the model incorporates vehicle positions, route information, downstream capacity, and neighboring traffic pressure into its state representation. This enables the agent to anticipate traffic demand and mitigate issues caused by partial observability. Furthermore, by acquiring the real-time data of priority vehicles and assigning weighted priority, the proposed model can adapt to real-time priority control. Simulation results show that the proposed model significantly outperforms baseline approaches, reducing average queue length and vehicle waiting time by approximately 19.2% and 57.9%, respectively. The model also demonstrates strong generalizability under traffic disturbances. In priority scenarios, it cuts the waiting time for high-priority vehicles by around 75%, with minimal impact on overall traffic efficiency. These findings demonstrate the model’s effectiveness, adaptability, and potential for deployment in intelligent traffic management systems.

With the emergence of studies on automated vehicles, the rapid development of new systems raises questions about road safety certification. Numerous methods have been developed to address this, such as Distance-Based and Scenario-Based approaches. The latter offers a time-saving advantage by avoiding redundant testing and focusing on traffic situations that pose risks to the system. However, scenarios can vary in levels of abstraction depending on the system’s design. Many studies attempt to identify safety criteria using indicators based on real-world scenarios, but perform at a low level of abstraction for the scenario description. Only a little draws analysis at the primary, i.e., abstract, level. Consequently, the qualification of abstract scenarios concerning safety indicators remains difficult and extremely dependent on field observations without harmonization. No link is clearly established yet between the abstract description of the experienced situation and indicators resulting from field observations. In a generic sense, abstract levels are managed at the expert level. This study establishes a rare connection between concrete (low abstraction level) and functional (high abstraction level) scenarios to compute the criticality of traffic scenarios using UAV data. Our approach develops the methodology to unify expert knowledge (top-down approach) with field observations (bottom-up approach).