Merging roadways are a major source of conflict and congestion, and can increase risk of collision, and fuel consumption. Coordinating merging of connected and automated vehicles (CAVs) in on-ramp scenarios can improve traffic efficiency, increase safety, and reduce the negative environmental impacts. In our previous work, a multi-player game-based centralized optimization algorithm was proposed to achieve global optimization of merging sequences and the decentralized non-linear model predictive controller was proposed for tracking optimal trajectory in the vehicle lower-level. This paper addresses the problem of developing a longitudinal and lateral cooperative motion control for CAVs merging at on-ramps. The vehicle longitudinal and lateral kinematics was decoupled by feedback linearization method to be linear uncorrelation. Quadratic of longitudinal and lateral virtual accelerations are used as the optimal objective to reduces the fuel consumption. A decentralized optimization of longitudinal and lateral cooperative control method was proposed and the analytical solution considering the strict input constraint was derived. Efficiency of the proposed cooperative method was validated by simulation. The proposed decentralized merging control system can improve traffic efficiency and reduce fuel consumption with the potential for real-time application.

Recent advances in LiDAR 3D detection have demonstrated the effectiveness of Transformer-based frameworks in capturing the global dependencies from point cloud spaces, which serialize the 3D voxels into the flattened 1D sequence for iterative self-attention. However, the spatial structure of 3D voxels will be inevitably destroyed during the serialization process. Besides, due to the considerable number of 3D voxels and quadratic complexity of Transformers, multiple sequences are grouped before feeding to Transformers, leading to a limited receptive field. Inspired by the impressive performance of State Space Models (SSM), in this paper, we propose a novel Unified Mamba (UniMamba), which seamlessly integrates the merits of 3D convolution and SSM in a concise multi-head manner, aiming to perform "local and global" spatial context aggregation efficiently and simultaneously. Specifically, a Uni-Mamba block is designed which mainly consists of spatial locality modeling, complementary Z-order serialization and local-global sequential aggregator. The spatial locality modeling module integrates 3D submanifold convolution to capture the dynamic spatial position embedding before serialization. Then the efficient Z-order curve is adopted for serialization both horizontally and vertically. Furthermore, the local-global sequential aggregator adopts the channel grouping strategy to efficiently encode both "local and global" spatial inter-dependencies using multi-head SSM. Additionally, an encoder-decoder architecture with stacked UniMamba blocks is formed to facilitate multi-scale spatial learning hierarchically. Extensive experiments are conducted on three popular datasets: nuScenes, Waymo and Argoverse 2. Particularly, our UniMamba achieves 70.2 mAP on the nuScenes dataset.

Geometric constraints between feature matches are critical in 3D point cloud registration problems. Existing approaches typically model unordered matches as a consistency graph and sample consistent matches to generate hypotheses. However, explicit graph construction introduces noise, posing great challenges for handcrafted geometric constraints to render consistency. To overcome this, we propose HyperGCT, a flexible dynamic Hyper-GNN-learned geometric ConstrainT that leverages high-order consistency among 3D correspondences. To our knowledge, HyperGCT is the first method that mines robust geometric constraints from dynamic hypergraphs for 3D registration. By dynamically optimizing the hypergraph through vertex and edge feature aggregation, HyperGCT effectively captures the correlations among correspondences, leading to accurate hypothesis generation. Extensive experiments on 3DMatch, 3DLoMatch, KITTI-LC, and ETH show that HyperGCT achieves state-of-the-art performance. Furthermore, HyperGCT is robust to graph noise, demonstrating a significant advantage in terms of generalization.

As the most widely used in-vehicle network, the controller area network bus lacks effective encryption and authentication mechanisms, exposing it to significant security threats. Voltage-based intrusion detection systems (IDSs), which detect malicious frames and locate attackers by establishing voltage fingerprints, have attracted widespread attention from researchers. However, the voltage signals of frames are vulnerable to temperature variations, leading to false positives and false negatives in voltage-based IDS. To address this, researchers have proposed the scheme that involves frequently updating voltage fingerprints and the temperature compensation-based scheme. Unfortunately, frequent updates to voltage fingerprints have been shown to be vulnerable to poisoning attacks, while the temperature compensation-based approach requires knowledge of the sender node's temperature. To this end, we propose utilizing robust voltage features to develop a robust voltage-based IDS without knowing sender node's temperature. Experiments conducted on both the prototype and real vehicle show that, compared to existing mainstream voltage-based IDS, our system demonstrates superior robustness under temperature variations. To the best of our knowledge, we are the first to detect intrusion by collecting voltage signals and extracting their features on resource-constrained device. Our system also includes an alarm module, which can trigger a buzzer to alert the driver when the intrusion is detected.

In intelligent transportation systems, highly reliable information exchange via New Radio (NR) Vehicles-to-Everything (V2X) communications is pivotal for ensuring road safety and enhancing traffic efficiency. However, the open nature of wireless channels renders NR V2X communications highly vulnerable to interference, thereby presenting opportunities for potential attackers to exploit. This paper proposes a method based on Gaussian Mixture Models (GMM) for detecting potential jamming attacks in NR V2X communications. This method detects jamming employing parameters calculated by stochastic geometry, without relying on abundant datasets. By utilizing the expectation-maximization algorithm for iterative, this method is able to identify jamming attacks based on the measured power. Extensive numerical analysis has demonstrated that our proposed method exhibits superior accuracy compared to existing schemes. Besides, this method is general, and capable of detecting jamming attacks under unknown jammer characteristics and varying vehicle densities.

Vehicular platoon communication demands high reliability and low latency to ensure safe and coordinated operations. However, the Semi-Persistent Scheduling (SPS) protocol in Cellular Vehicle-to-Everything (C–V2X) Mode 4 often results in persistent packet collisions in resource contention scenarios, greatly undermining the stability of the platoon. This paper proposes a Echelon-based Collaborative Resource Allocation (ECRA) protocol that conceptualizes the platoon structure as a three-tier communication hierarchy and implements refined resource management strategies. ECRA introduces four mechanisms to enhance the reliability of platoon communication. Specifically, a vehicle role-aware resource evaluation mechanism achieves differentiated assessment and allocation of resource quality by considering the functional importance of vehicles within the platoon. A hybrid error classification mechanism effectively differentiates communication errors within the platoon by integrating deterministic decision making with fuzzy logic. A echelon-based error response mechanism provides customized resource selection strategies and error response mechanisms for vehicles of different roles, ensuring efficient allocation of communication resources according to importance. A echelon-based waiting window mechanism dynamically optimizes the detection frequency based on vehicle roles, thereby balancing real-time requirements with system overhead. Finally, a theoretical model of packet collision probability and average delay is developed to quantify improvements in reliability and latency. Simulation results demonstrate that the proposed ECRA enhances platoon communication reliability and reduces latency more effectively than traditional SPS and other existing solutions. In particular, ECRA exhibits superior robustness in dynamic environments, especially under high-density and complex interference conditions.

Deep neural networks (DNNs) are potent in LiDAR-based 3D object detection (LiDAR-3DOD), yet their deployment remains daunting due to their cumbersome parameters and computations. Knowledge distillation (KD) is promising for compressing DNNs in LiDAR-3DOD. However, most existing KD methods transfer inadequate knowledge between homogeneous detectors, and do not thoroughly explore optimal student architectures, resulting in insufficient gains for compact student detectors. To this end, we propose a category knowledge-driven compression framework to achieve efficient LiDAR-based 3D detectors. Firstly, we distill knowledge from two-stage teacher detectors to one-stage student detectors, overcoming the limitations of homogeneous pairs. To conduct KD in these heterogeneous pairs, we explore the gap between heterogeneous detectors, and introduce category knowledge-driven KD (CaKD), which includes both student-oriented distillation and two-stage-oriented label assignment distillation. Secondly, to search for the optimal architecture of compact student detectors, we introduce a masked category knowledge-driven structured pruning scheme. This scheme evaluates filter importance by analyzing the changes in category predictions related to foreground regions before and after filter removal, and prunes the less important filters accordingly. Finally, we propose a modified IoU-aware redundancy elimination module to remove redundant false positive samples, thereby further improving the accuracy of detectors. Experiments on various point cloud datasets demonstrate that our method delivers impressive results. For example, on KITTI, several compressed one-stage detectors outperform two-stage detectors in both efficiency and accuracy. Besides, on WOD-mini, our framework reduces the memory footprint of CenterPoint by 5.2× and improves the L2 mAPH by 0.55$\%$%.

Three-dimensional (3D) point clouds are essential for representing traffic scenes in autonomous driving, yet incompletion often occurs due to sensor angle limitations and signal occlusions. Point cloud completion focuses on generating the missing parts of incomplete point cloud shapes, but still shows limitations in preserving and restoring the details of the shapes. To address these challenges, we propose a novel point cloud completion method, termed PCC-GAN, which leverages an improved generative adversarial network to predict and correct missing portions of point clouds. The proposed architecture features two primary components: a generator and a discriminator. The generator includes a feature extension module that uses local feature interpolation to learn and refine the input point cloud features, thereby enhancing learning efficiency. Meanwhile, the discriminator employs a self-attention mechanism to capture long-range dependencies and extract detailed contextual information, allowing it to evaluate the local accuracy of the generated features effectively. The performance of the PCC-GAN model is assessed using public datasets, and its robustness is tested across point cloud sets with varying levels of missing proportion. Results indicate that PCC-GAN significantly improves the shape reconstruction of lidar point clouds, demonstrating strong predictive capabilities and robustness.

A deep reinforcement learning-based method for planning the lane-changing trajectory of connected and automated vehicles (CAVs) is proposed to increase traffic efficiency and reduce fuel consumption. The long-short-term-memory-based twin delayed deep deterministic policy gradient (LSTM-TD3) algorithm is implemented and trained to achieve the optimal longitudinal and lateral lane-changing trajectory. The instantaneous fuel consumption and the desired speed and acceleration difference are used as reward and penalty terms. The effectiveness of the algorithm was verified through real data based typical lane-changing scenarios using CARLA software. The results indicate that the proposed LSTM-TD3-based lane-changing planning method reduced fuel consumption by 6.36% compared to TD3, 9.84% compared to LSTM-DDPG, and 26.31% compared to DDPG. Compared to TD3, LSTM-DDPG, and DDPG, the completion time for lane-changing is reduced by 0.18s, 0.15s and 0.2s, respectively. The success rate of trajectory planning has also increased compared to other algorithms. Furthermore, the results demonstrate the potential of deep reinforcement learning technologies in the control and applications of CAVs.

Knowledge distillation (KD) can compress deep neural networks (DNNs) by transferring the knowledge of the redundant teacher model to the resource-friendly student model, where cross-layer KD (CKD) conducts KD between each stage of students and the multiple stages of teachers. However, previous CKD schemes select the coarse-grained stagewise features of teachers to teach students, leading to improper channel alignment. Also, most of these methods conduct uniform distillation for all the knowledge, limiting students to focus more on important knowledge. To address these problems, we propose a dense KD (DenseKD) in this article, dubbed as DenseKD. First, to achieve more accurate feature alignment in CKD, we construct the learnable dense architecture to make each channel of student flexibly capture more diverse channelwise features from teacher. Moreover, we introduce region importance to investigate the region's guiding potential, it distinguishes the influence of different regions by the variation of representations of teacher models. In addition, to make students pay more attention to useful samples in KD, we calculate sample importance by the loss of teacher models. Consistent improvements over state-of-the-art approaches are observed in experiments on multiple vision tasks. For example, in the classification task, DenseKD achieves 72.30% accuracy of ResNet-20 on CIFAR-100, which is higher than the results of previous CKD methods. In addition, in the object detection task, DenseKD gains 2.84% mean average precision (mAP) improvements of Faster R-CNN with ResNet-18 against vanilla KD.

Lidar-based 3D Detection is one of the significant components of Autonomous Driving. However, current methods over-focus on improving the performance of 3D Lidar perception, which causes the architecture of networks becoming complicated and hard to deploy. Thus, the methods are difficult to apply in Autonomous Driving for real-time processing. In this paper, we propose a high-efficiency network, SwiftPillars, which includes Swift Pillar Encoder (SPE) and Multi-scale Aggregation Decoder (MAD). The SPE is constructed by a concise Dual-attention Module with lightweight operators. The Dual-attention Module utilizes feature pooling, matrix multiplication, etc. to speed up point-wise and channel-wise attention extraction and fusion. The MAD interconnects multiple scale features extracted by SPE with minimal computational cost to leverage performance. In our experiments, our proposal accomplishes 61.3% NDS and 53.2% mAP in nuScenes dataset. In addition, we evaluate inference time on several platforms (P4, T4, A2, MLU370, RTX3080), where SwiftPillars achieves up to 13.3ms (75FPS) on NVIDIA Tesla T4. Compared with PointPillars, SwiftPillars is on average 26.58% faster in inference speed with equivalent GPUs and a higher mAP of approximately 3.2% in the nuScenes dataset.

Autonomous vehicle trajectory prediction is an important component of autonomous driving assistance algorithms (ADAAs), which can help autonomous driving systems (ADSs) better understand the traffic environment, assess critical tasks in advance thus improve traffic safety and traffic efficiency. However, some existing neural network-based trajectory prediction models focus on theoretical numerical analysis and are not tested in real time, leading to doubts about the practical usability of these trajectory prediction models. To address the above limitations, this study first proposes a collaborative simulation environment integrating traffic scenario construction, driving environment perception, and neural network modeling, afterwards used the co-simulation environment for trajectory data and driving environment data collection. In addition, based on the characteristics of the collected data, a trajectory prediction model based on Bi-Encoder-Decoder and deep neural network (DNN) is proposed and pre-trained. Finally, the pre-trained completed model is embedded in the co-simulation environment and tested in real-time with different batches of data. The simulation results show that the proposed trajectory prediction model can predict trajectories well under specific training data batches, and the best performing trajectory prediction model has a prospective time of 4.9 s and a prediction accuracy of 91.55%.

Connected and automated vehicles (CAVs) can improve traffic safety and transportation network efficiency while also reducing environmental impacts. However, congestion and accidents can easily occur at merging roadways. Therefore, coordinating cooperative merging of CAVs is one of the most common traffic management problems. This paper addresses the problem of integrated longitudinal and lateral cooperative merging control with practical implications for CAVs approaching on-ramps. A hierarchical and decentralized cooperative coordination framework was developed to systematically control the merging of CAVs. The control system of each vehicle can be divided into an upper-level and lower-level. For upper-level control, an optimal control-based algorithm considering input constraints was presented to optimize fuel consumption and passenger comfort. A decision strategy was developed to optimize the start time of lateral trajectory planning. To achieve lower-level control, a Proportional-Integral (PI) controller was used for tracking the optimized longitudinal speed of the upper-level and a decentralized unified algorithm based on nonlinear model predictive control was proposed for tracking the upper-level optimal trajectory. To avoid lateral collision, the driving safety field based on vehicle size and motion state was selected as one of tracking the optimization objectives. Efficiency of the proposed framework and the algorithm was validated by CarSim/Simulink co-simulations of near-real-world vehicle scenarios. The proposed integrated merging control system can improve traffic efficiency and reduce fuel consumption compared to baseline with the potential for real-world application. Furthermore, the results demonstrate the potential applicability of cooperative control methods based on upper-level vehicle control.

Vehicle merging is one of the main causes of reduced traffic efficiency, increased risk of collision, and fuel consumption. Connected and automated vehicles (CAVs) can improve traffic efficiency, increase safety, and reduce the negative environmental impacts through effective communication and control. Therefore, to improve the traffic efficiency and reduce the fuel consumption in on-ramp scenarios, this paper addresses the global and optimal coordination of the CAVs in a merging zone. Herein, a cooperative multi-player game-based optimization framework and an algorithm are presented to coordinate vehicles and achieve minimum values for the global pay-off conditions. Fuel consumption, passenger comfort, and travel time within the merging control zone were used as the pay-off conditions. After analyzing the characteristics of the merging control zone and selecting the appropriate control decision duration, multi-player games were decomposed into multiple two-player games. An optimal merging strategy was, thereby, derived from a pay-off matrix, and minimum payoffs were predicted for a number of different potential strategies. The optimal trajectory corresponding to the predicted minimum payoffs was then utilized as the control law to coordinate the vehicles merging. The proposed control scheme derives an optimal merging sequence and an optimal trajectory for each vehicle. The effectiveness of the proposed model is validated through simulation. The proposed controller is compared with two alternative methods to demonstrate its potential to reduce fuel consumption and travel time and to improve passenger comfort and traffic efficiency.

This paper developed an integrated ISS and FPGA SoC co-verification platform named MLCV, introduced the overall structure and principle of MLCV, and described each component's function of MLCV. Processor is modeled by an ISS that interface with the source-level debugger. The peripherals are implemented in the FPGA board. Communication between ISS and FPGA is encapsulated by a co- verification wrapper. To synchronize the bus operation, the on-chip bus protocol is implemented in FPGA. As a bus master, the ISS's peripheral access information is generated by the co-verification wrapper. SoC architecture supported by MLCV is limited by the on- chip bus protocol.