Vehicle object detection is a fundamental component of intelligent transportation systems, yet its deployment in resource-constrained environments is limited by the trade-off between accuracy and computational efficiency. This work systematically examines the relationship between network depth and detection performance through controlled experiments on both convolutional and attention-based architectures. The results reveal that accuracy does not scale indefinitely with depth, but instead exhibits diminishing returns and converges to a task-dependent performance plateau. A mutual information-based analysis attributes this saturation to fundamental limitations in information retention and representational efficiency. Motivated by this insight, we propose the TritonFocus Network (TF-Net), an architecture that integrates large-Kernel contextual perception with small-Kernel local aggregation for efficient feature representation. Extensive evaluations on the UA-DETRAC and BITVehicle benchmarks demonstrate that TF-Net achieves a state-of-the-art accuracy–efficiency trade-off among lightweight detectors. TF-Net achieves a relative detection accuracy improvement of approximately 5.8%, while maintaining high computational efficiency and approaching twice the inference speed of the baseline model. These findings validate the efficacy of saturation-aware architectural optimization and provide theoretical and practical guidance for deploying high-performance detection models in edge computing scenarios.

In cognitive radio ad-hoc networks (CRAHN), we can change the network topology through flexible spectrum allocation. In this regard, even though CRAHN is not prone to single points of failure, its network performance is entirely dependent on node connectivity. However, the existing solutions mainly focus on the connections between certain nodes, without fully considering the overall network connectivity. To this end, to avoid a large number of connection failures caused by local damage, it is necessary of improving system connectivity performance from a global perspective when allocating available spectrum. As such, we explore the relationship between spectrum allocation and global connectivity of CRAHN, and then propose a robust optimization scheme using optimal spectrum allocation (ROUOS). This scheme aims to create new connections in the connectivity vulnerable areas through spectrum allocation. Specifically, we establish a spectrum allocation model including available spectrum matrix, bandwidth benefit matrix, interference constraint matrix, communication connection matrix, alternate channel vector and optimal allocation matrix. After that, based on the connectivity quantitative index in [11], we design a network benefit function to measure the gain effect of different spectrum allocation solutions on network connectivity. Finally, we select the spectrum allocation solution (i.e., adding communication links) that most effectively improves network connectivity. Simulation results show that, compared to the benchmark scheme, our proposed ROUOS scheme increases the second smallest eigenvalue of the graph Laplacian matrix from 0.09 to 0.29, significantly enhancing the robustness of topological connectivity.

3D-stacked DRAM is a key technology enabling the development of large language models (LLMs). However, the intensive computational demands lead to substantial energy consumption resulting from large-scale data movement. Integrating processing-in-memory (PIM) within DRAM has been employed to reduce data movement, but it elevates manufacturing costs and incurs additional area and power consumption. On the other hand, implementing processing-near-memory (PNM) outside the DRAM controller fails to eliminate the high energy consumption associated with interconnects during data transfer. To address these challenges, we propose a processing-in-controller (PIC) architecture aimed at 3D-stacked DRAM for efficient data movement. A size-scalable General Matrix-Vector Multiplication (GEMV) structure is proposed, supporting configurations ranging from 16 × 16 to 128 × 128, thereby enhancing its adaptability for a variety of applications. To address the low throughput bottleneck caused by DDR read latency, a ping-pong buffer with weight prefetching is introduced in the PNM. Based on 28 nm CMOS technology, experimental results demonstrate that the energy consumption of this PIC is 0.22 pJ/bit, while data movement efficiency improves by nearly 92% compared to traditional DRAM operations.

Coordination at unsignalized intersections has attracted increasing attention in recent years, which aims at improving the efficiency of intersection operations, while eliminating conflicts and deadlocks for Connected and Automated Vehicles (CAVs). This paper addresses the challenging issue of systematically optimizing CAV trajectories, tackling the high computational cost for finding a good solution, especially as the number of lanes and CAVs increases. To do so, a novel systematic optimal trajectory planning model is designed to efficiently guide CAVs through intersections without conflicts and deadlocks. To tackle the computational hurdles and enable real-time applications, based on the model, we develop a Petri net-aided Iterative Trajectory Optimization (P-ITO) solution algorithm. Leveraging the unique characteristics of the problem, this algorithm first designs a Petri net-based controller for conflict and deadlock avoidance so that initial feasible solutions are generated. Then, refinement is made on the initial feasible solutions to obtain an optimal or near optimal solution by designing an iterative process. This algorithm effectively ensures solution feasibility and enhances computing efficiency by searching for an optimal solution in the feasible region. Numerical experiments for intersections with bidirectional six-lane configurations illustrate the efficacy of our model in facilitating the safe and efficient passage of all CAVs while mitigating the risk of deadlocks. The P-ITO algorithm significantly outperforms the commercial solvers in both solution quality and computational time. Furthermore, our method is versatile, applicable to diverse intersection scenarios, and capable of maintaining high computational efficiency.

The resilience of a transportation system is its ability to recover and provide timely services during traffic emergencies, which is crucial for highly urbanized societies. Traffic evacuation efficiency plays a significant role in improving resilience. However, previous studies have not considered maximizing resilience through a multidimensional approach to the siting of disaster response nodes, emergency evacuation, and resource allocation during emergencies. In this paper, we propose a two-stage optimization framework for evaluating the resilience of transportation systems, considering the uncertainty of the level of disaster. In the first stage. A hybrid algorithm combining discrete differential evolution and genetic algorithm was designed (H-DEGA). In the second stage, we calculate the optimal evacuation strategy of the system by continuously removing the important road sections in the traffic network using an optimization model. The system's resilience is derived by assessing its performance before and after disasters. Computational results for the Xi'an city network demonstrate that the model can determine the optimal recovery solution and improve system resilience. Finally, this method provides insights into traffic evacuation and intelligent city management.

Monitoring driver's behavior is the foundation for ensuring road safety. However, identifying unsafe behavior of the driver during the driving process still faces many challenges. In this paper, we propose a recognition network based on graph convolution, aimed at effectively identifying various driver's unsafe behaviors during the driving process. The SDRM-CSTJM graph convolutional network (SC-GCN) comprises two modules: the stacked deconvolution residual module (SDRM) and the channel-spatial-temporal joint module (CSTJM). The SDRM extracts information about the driver's unsafe actions across various scales by stacking multiple unit structures and introducing multi-level residual blocks. The CSTJM extracts features from different dimensions by integrating channel, spatial, and temporal information. Integrating SDRM and CSTJM into the spatial temporal graph convolutional network (ST-GCN), and a two-stream network is introduced to handle joint data and skeletal data separately. Subsequently, integrating the outputs of the two streams. Achieving recognition of various behaviors exhibited by the driver based on their body posture features. Finally, we evaluated the proposed method on the dataset. Experiments demonstrate that the proposed method achieves the highest recognition accuracy on the driver action dataset.

The advent of Connected and Autonomous Vehicles (CAVs) offers significant potential to mitigate traffic accidents and reduce congestion through advanced unsignalized intersection technologies. However, this potential also introduces substantial modeling and computational challenges. This paper establishes a Mixed Integer Linear Programming (MILP) model and a Simulated Annealing Algorithm (SAA) to solve the unsignalized intersection problem. It begins with an indepth analysis of the conflict relationships among vehicles at intersections, laying the foundation for the model. Building on this, a novel vehicle trajectory planning model is introduced, leveraging the MILP model to ensure faster and more practical trajectory calculations, meeting real-time computation requirements. Finally, this paper designs a SAA to determine the optimal vehicle passing order, enhancing both safety and efficiency at intersections. Through numerical simulations, the research verifies the efficiency of the proposed MILP model and the effectiveness of the SAA in optimizing vehicle trajectories.

Nowadays, on-ramp area is one of the bottlenecks of traffic congestion. With the advancement of intelligent connected technology, it is important to study how to improve traffic efficiency when ensuring the safety of connected and autonomous vehicles (CAVs) and human-driven vehicles (HDVs) in ramp merge zones. Especially under high traffic flow conditions, HDVs often fail to yield to each other, leading to frequent traffic congestion and gridlock. To address these issues, this paper proposes a new multi-vehicle platooning strategy and designs a bilevel trajectory planning framework to ensure the safe merging of vehicles and improve traffic efficiency in mixed traffic environments. Additionally, the framework further enhances the computational efficiency of vehicle trajectory planning by using a multi-step increasingly approximating algorithm to transform the non-linear trajectory planning model into a linear model for solving. The multi-vehicle platooning strategy was tested through a series of experiments under different traffic states and market penetration rates (MPRs) of CAVs. The results demonstrated that the proposed strategy significantly reduced model-solving time and enhanced traffic efficiency compared to scenarios where vehicles do not form platoons.

Intersections are high-congestion areas in urban roads. Bus priority signal control can achieve optimal intersection performance while providing priority service to buses. This study first analyzed two operational scenarios of buses entering the detection area and calculated the time for buses to reach the stop line and delay variation under different conditions. Next, a bus priority signal control model was established to minimize average passenger delay, using active priority strategies including green light extension and red light early termination to prioritize bus passage. Then, fuzzy control theory was combined with a genetic algorithm to solve for the optimal signal timing adjustment scheme at the intersection, and simulations were conducted using SUMO. The final results show that this method can effectively reduce intersection delays compared to traditional fixed-time methods. When bus proportions are 2%, 4%, and 6%, this method can reduce average delays and queue lengths by more than 20% and 30%.

Existing studies on post-disaster road network repair strategies have ignored the impact of different levels of road damage and recovery on the efficiency of network repair. To solve this issue, this study integrates the construction material distribution (CMD) with the repair crew scheduling and routing problem (RCSRP), and determine the level of road recovery through the CMD. Then, a bi-level optimization model is proposed with network performance resilience and recovery speed resilience as the optimization objectives. A two-stage optimization algorithm (TSOA) composed of a genetic algorithm with an improved coding method (ICM-GA) and the Frank-Wolfe algorithm (FW) is then employed to solve this model. Finally, the effectiveness of the model and algorithm is validated through simulation experiments. The results indicate that, under given material and time constraints, the optimal repair strategy proposed in this study outperforms the repair strategy without considering road recovery level by 17.51% and 5.42% in terms of network performance resilience and recovery speed resilience, respectively. This demonstrates the positive significance of considering road recovery level in formulating road network repair strategies. Besides, this strategy can be applied to optimize the configuration of workstation count for different-scale networks.

Self-attention-based transformers have outperformed recurrent and convolutional neural networks (RNN/ CNNs) in many applications. Despite the effectiveness, calculating self-attention is prohibitively costly due to quadratic computation and memory requirements. To solve this challenge, this article proposes a hybrid analog-ReRAM and digital-SRAM in-memory computing accelerator (HARDSEA), a computing-in-memory (CIM) accelerator supporting self-attention in transformer applications. To trade off between energy efficiency and algorithm accuracy, HARDSEA features an algorithm-architecture-circuit codesign. A product-quantization-based scheme dynamically facilitates self-attention sparsity by predicting lightweight token relevance. A hybrid in-memory computing architecture employs both high-efficiency analog ReRAM-CIM and high-precision digital SRAM-CIM to implement the proposed new scheme. The ReRAM-CIM, whose precision is sensitive to circuit nonidealities, takes charge of token relevance prediction where only computing monotonicity is demanded. The SRAM-CIM, utilized for exact sparse attention computing, is reorganized as an on-memory-boundary computing scheme, thus adapting to irregular sparsity patterns. In addition, we propose a time-domain winner-take-all (WTA) circuit to replace the expensive ADCs in ReRAM-CIM macros. Experimental results show that HARDSEA prunes BERT and GPT-2 models to 12%–33% sparsity without accuracy loss, achieving$13.5\times $–$28.5\times $speedup and$291.6\times $–$1894.3\times $energy efficiency over GPU. Compared to state-of-the-art transformer accelerators, HARDSEA has$1.2\times $–$14.9\times $better energy efficiency at the same level of throughput.

Aiming at the problems that most of the existing researches use a single indicator as a reward function for evaluating the behavior of intelligences as well as ignoring the efficiency of data sampling in the experience pool, which leads to unstable training process and slow convergence. In this paper, we propose a traffic signal control method based on the D4QN_PEN (Distributional Double Dueling Deep Q-learning with Priority Experience Replay and NoisyNet) deep reinforcement learning model. The method adopts a multi-indicator weighted average approach to define the reward function and makes the following improvements: (1) Use the Priority Experience Replay technique to give priority to the samples and increase the probability of occurrence of valuable samples. (2) Enhance the exploration ability of the model by introducing noise parameters through NoisyNet technique. (3) En-hance the model performance by utilizing Double Q Network, Dueling DQN and Distributional DQN. Additionally, the high-dimensional real-time traffic information of the intersection is converted into a state matrix consisting of vehicle positions, speeds and waiting times by variable-length discrete traffic coding, and the traffic state is abstractly characterized by deep learning to achieve adaptive signal control. To validate the proposed algorithm, it is verified on the simulation platform SUMO, and compared with the traditional DQN and Double Dueling DQN algorithms, the pro-posed D4QN_PEN algorithm is able to shorten the waiting time and queue length of the vehicles, and has a better control effect.

Internet ride sharing allows multiple passengers to share a trip in the same vehicle, enabling cost sharing as well as reducing traffic congestion. However, existing technological limitations and uncertainties in the service (e.g., uncertainty in driver and passenger locations) make it difficult to achieve accurate and efficient real-time responses for ride-sharing matching. Balancing the optimal solutions of drivers, passengers, and platforms, dynamically matching passengers and drivers, and planning optimal paths are complex challenges. Therefore, this paper proposes a greedy algorithm based on the nearest match insertion operation to synthesize the interests of platforms, drivers and passengers. Compared with static one-time matching, this algorithm can effectively realize dynamic matching of drivers and passengers, meet real-time demand, provide drivers with optimal driving paths, and improve the scheduling efficiency of the platform. In this thesis, a dynamic carpooling optimization model is constructed and used to design comparison experiments with the traditional greedy algorithm. This study helps improve the efficiency of the ride-hailing system and enhance the passenger experience, providing valuable references for the promotion and application of dynamic carpooling models in smart cities.

Traffic delay and congestion frequently occur in off-ramp areas, but few researches focus on generating the microscopic trajectory plan for individual connected and automated vehicles (CAV) to instruct their real-time acceleration/deceleration rate and lane-changing maneuvers at off-ramps. This paper proposes a trajectory planning model for vehicle platoons at off-ramp based on mixed-integer nonlinear programming (MINLP). The model can generate systematical optimal trajectories for CAVs passing the off-ramp safely and efficiently. Aiming at optimizing the overall traffic efficiency, we developed a series of constraints to model the behavior of vehicles running in the trajectory area. Especially, the impact of the lane-changing of vehicles is carefully designed to further improve the efficiency and safety by coordinating the behavior of different vehicles on the main road into several vehicle platoons and simultaneously generates trajectories for all vehicles in a platoon. The experiments show that the model can improve the traffic efficiency of pure CAV flow on the off-ramp with safety and effectively solve traffic congestion at the off-ramp.

Intersections are typical bottle necks in urban traffic system. Unsignalized intersection technology, together with the Connected and Autonomous vehicles (CAVs), has a great potential to alleviate traffic congestion and has gained rapidly increasing interest. This work presents a Mixed Integer Nonlinear Programming (MINLP) model to generate the CAV trajectories that maximize average speed, so as to guide CAVs to pass an unsignalized intersection in a safe and efficient manner. Specifically, potential conflict points between CAVs from different lanes are analyzed. Upon that, the realistic constraints, e.g. vehicle kinematic limits and collision avoidance under different conflict scenarios, are established. To carry out collision avoidance mechanism and control the time at which CAVs pass the conflict point, we reformulate the arc-shaped CAVs' trajectory at an unsignalized intersection into a straight line. Numerical examples are conducted on several representative applications and demonstrate the effectiveness of proposed mathematical model.

This paper studies the short-term passenger flow prediction of urban rail transit for optimally adjusting the real-time departure of rail trains. Aiming at the problem that the traditional deep learning model does not consider the spatial-temporal information enough, the short-term passenger flow prediction model of urban rail transit based on CNN-LSTM with attention mechanism is proposed. Firstly, the stations are divided into seven categories according to the significant difference of daily passenger flow in urban rail stations so as to further analyze the distribution pattern of daily inbound and outbound passenger flow in different categories of stations; secondly, the short sequence feature abstraction ability of CNN is used to extract the spatial characteristics of historical passenger flow in each time period in different categories of stations; finally, the attention mechanism is used to assign different weights to the extracted characteristic information, and the temporal characteristic information is obtained from the LSTM comprehensive short-term sequence to realize the short-term passenger flow prediction of urban rail transit. Experiments show that the prediction model has the encouraging prediction performance and accuracy.

The severe rodent damage in Gurbantunggut desert forest may result in eco-environment destruction and desertification acceleration. However, little is known concerning the level of rodent hazard, spatial distribution and regularity of outbreak. This paper approached to use high-resolution imagery (TDOM and DEM) provided by Unmanned Aerial Vehicles (UAV) for rodent pests monitoring in desert forest. The distribution of rat-holes was extracted from TDOM images, then rat-hole density were calculated for evaluating the rodent hazard degree in the study area. Lastly, topographic factors of slope gradient, slope aspect, elevation were calculated using DEM. Spatial and statistical analysis methods were used for it. The results demonstrated that Rhombimys opimus mainly distributed in the elevation of 432.5~437.5m, low slope areas of 0~15°, low RDLS of 0~0.25m, and shady slope of fixed and semi-fixed sand with well-growth plants by “island” mousehole groups. This paper validated the availability of UAV technique for dynamic monitoring rodent pests in desert forest, and laid a critical foundation for further study.

The research presents in this paper develops an Ant Colony System (ACS) based system optimal route choice strategy to ensure the rational traffic flow assignment for urban traffic network. In this work, the traffic flow and impedance function of each road section are calculated firstly, and then the individual traveler's route choice behaviors on network nodes are simulated based on applying the pseudo-random state transition rule, route and road section pheromone update formula, which implement the synthesizing of static prior knowledge, dynamic traffic state and the randomness of route choice. This paper's findings reveal that the designed strategy is in a position to reflecting the overlay and delay effect of route choice under different Origin-Destination (OD) demands. In addition, the findings can obtain better network equilibrium comparing with the incremental assignment method, and will benefit for achieving the route guidance system with time varying traffic conditions.