The Underwater Sonar Moving Target (USMT) dataset is a forward-looking sonar (FLS) image dataset for low-pixel underwater moving-target perception and segmentation. It contains 21 temporally ordered sequences and 7,136 annotated rectangular B-scan FLS images collected during AUV experiments in a controlled lake environment. Each sonar image is accompanied by a pixel-level annotation in LabelMe JSON format and a TXT file containing the corresponding sonar acquisition metadata. The annotations were generated using X-AnyLabeling with the SAM 2 video model and subsequently reviewed and refined manually. The dataset can support research on underwater intelligent perception, sonar image segmentation, and temporal modeling of moving targets.

A project for a study on LLM-designed cooperative control strategies for connected and automated vehicles at unsignalized intersections. It includes the Python source code, SUMO simulation files, processed scenario data, experiment logs, and documentation needed to reproduce and inspect the main computational experiments.

This repository implements RADGCN, a relation-aware graph neural network for network-wide traffic state estimation (traffic kriging) when only a subset of road sensors is observed. It accompanies the manuscript Relation-diffusion augmented network-wide traffic state estimation.
Rather than learning adaptive relations from zero-filled features at unobserved nodes, RADGCN first learns functional relations among nodes with reliable observations. It then diffuses these relations across the directed road graph and uses the resulting relation kernel alongside forward and backward physical diffusion to reconstruct traffic speeds at unobserved nodes.

This repository provides the code and supporting files for reproducing the experiments reported in “Distilling Vision-Language Models for Explainable Vehicle Collision Prediction.” It includes data-processing scripts for generating collision-prone and collision-free video clips from the MM-AU and CCD datasets, dataset manifests and annotations, VLM fine-tuning and knowledge-distillation scripts, and evaluation tools. The code was developed in a Linux environment using Python 3.10 and CUDA 11.8 or later. The required dependencies are listed in `requirements.txt`.

Multi-V2X is a large-scale, multi-modal, multi-penetration-rate dataset for cooperative perception under vehicle-to-everything (V2X) environment. Multi-V2X is gathered by SUMO-CARLA co-simulation and supports tasks including 3D object detection and tracking. Multi-V2X provides RGB images, point cloud from CAV and RSU with various CAV penetration rates (up to 86%). This package also includes the source code for the Which2comm framework proposed in the associated paper.

The code is structured into two subdirectories, image_code and video_code, which respectively hold the full implementation of data generation, preprocessing, model training, inference and evaluation for the image and video modalities, with each subdirectory equipped with a separate README file that provides detailed environment setup and usage instructions. The runtime environment for this code is configured with Python 3.10, torch==2.6.0+cu126, and torchvision==0.21.0+cu126

This replication package comprises two complementary components: (1) a MATLAB‑based DAS simulation framework that generates synthetic acoustic signals and spatiotemporal traffic maps from microscopic vehicle trajectories, and (2) a Python implementation of physics‑informed neural networks (PINNs) for traffic state restoration, covering 16 model variants with different architectures (ResUNet and FNO) and physical constraints (LWR, LWR+FD, and ARZ). The package includes simulation input files (trajectory data and SUMO road network), a tutorial video for the simulation workflow, pre‑trained weights for two‑channel inputs, and sample training data. The full dataset and additional three‑channel model weights are available via a separate download link. 

This repository provides a lightweight reproduction of an autonomous-driving intelligence evaluation workflow. It evaluates vehicle interactions from five dimensions, combines ten normalized objective metrics with LLM-based subjective scores, and trains an attention-based neural network to approximate the final intelligence score.

The repository includes 705 indexed Waymo interaction samples and a lightweight Waymo/trajdata cache. It can therefore reproduce the main workflow without downloading the complete Waymo or InterHub dataset.

Replication package access:  https://github.com/2004yan/MapKD_new.

Data description: We adopt the publicly available nuScenes as our dataset.

Code description:  torch==1.9.1+cu111 torchvision==0.10.1+cu111 torchaudio==0.9.1

Simulation software description: Configuring Python 3.8 on Ubuntu 20.04

Experiment design description: In the first stage, the teacher and coach model is jointly pretrained.In the second stage, the student model is trained from scratch under the proposed Teacher-Coach-Student pipeline. The teacher and the coach are kept frozen, while providing intermediate guidance at the feature and output level. The student is supervised with both hard labels and soft logits. We train our model on the nuScenes dataset , using a BEV range of 60 m × 30 m. The voxel resolution is set to 0.15 m. Input images are resized to 128 × 352. We train for 30 epochs with a batch size of 8, using 4 GPUs and 20 workers.We use the Adam optimizer with a learning rate of , weight decay of , and gradient clipping at 5.0.

This replication package is associated with the manuscript “Autonomous vehicle decision-making and safety evaluation based on physics-aware graphs and hybrid policy optimization” submitted to the Journal of Intelligent and Connected Vehicles under manuscript ID JICV-2026-0020. The package includes source code, configuration files, simulation scripts, processed sample data, training logs, evaluation results, generated tables, figure-generation scripts, and a Replication Explanatory File for Journal-Associated Data. These materials support reproduction of the main experiments, including online reinforcement learning in highway-v0, comparisons among PPO-MLP, Ladm-PPO, and Ladm-HPO, offline safety evaluation using LadmCritic, TTC correlation analysis, and risk-energy field visualization. External datasets are not redistributed if restricted by their original data-use policies; instead, source information and preprocessing instructions are provided.

The code and data for MILD. To access the original video of AIDE, DMD, and DriveLM, please refer to their source papers.

Ship navigation is strongly influenced by the ocean environment. Adverse sea environment not only increases fuel consumption but can also endanger navigation safety. For the challenge, a ship weather routing framework based on Deep Reinforcement Learning (DRL) is proposed.

Due to limitations on the amount of data that can be uploaded, the replication package only contains the key code.

The complete key code and data will be attached as additional attachments.

In Part I of this companion paper series, we introduced SWIFTraj, an open-source vehicle trajectory dataset collected by a UAV swarm. It provides long-distance continuous trajectories by connecting vehicle trajectories across consecutive UAV videos, with the longest trajectory exceeding 4.5 km, and covers an integrated network of freeways and connected urban roads. However, trajectory connection in UAV swarms is challenging because of video time-offset errors and irregular UAV layouts. To address these issues, this paper proposes a graph-based trajectory connection method. An undirected graph is used to represent flexible UAV layouts, an automatic time-alignment method is developed by minimizing trajectory matching costs, and cross-video vehicle association is performed using a Hungarian-algorithm-based matching table. Experiments on real-world and simulated data show that the proposed method achieves time alignment errors within three frames, about 0.1 s, and consistently high vehicle-matching F1-scores.

This paper presents a detailed description and characterization of a new open-source vehicle trajectory dataset, namely SWIFTraj, constructed from videos recorded by a swarm of 16 drones equipped with 5.4K-resolution cameras. The dataset is distinguished from existing open-source trajectory datasets in several aspects. First, it provides long-distance continuous trajectories of up to 4.5 km on a freeway, enabling in-depth investigation of traffic phenomena and their spatial and temporal evolution. Second, the data collection site covers an integrated network consisting of a long freeway corridor and parts of its connected urban network, facilitating traffic analysis and modeling from a network perspective. The dataset is publicly available at the SWIFTraj website (https://www.swiftraj.com). 

The package supports reproduction of the main numerical results reported in the study, including the MATLAB/Simulink simulations of a heterogeneous five-truck platoon, the longitudinal model predictive control with bounded RBF residual compensation, the LLM-guided scenario interpretation and deterministic validation process, and the scenario-adaptive lateral organization optimization.

The package includes MATLAB/Simulink model files, MATLAB scripts for lateral organization optimization and figure generation, Python scripts for LLM-guided scenario interpretation and safety validation, scenario input and output files, trained RBF model files, processed simulation data, and reference output figures/tables. The simulated scenarios are generic road and traffic segments and are not based on proprietary, confidential, or human-subject data. Detailed instructions for reproducing the main results are provided in the README file and the replication explanatory file.

The simulated scenarios are generic road and traffic segments and are not based on proprietary, confidential, or human-subject data.

The replication package contains processed real-world emission datasets and Python scripts used to reproduce the main analyses and figures presented in this study. The datasets include overall trip-average emissions, operating mode-specific emission results, and SCR upstream/downstream NOx data for China V and China VI heavy-duty diesel trucks tested using PEMS under real-road conditions. All analyses were conducted using Python 3.9 in PyCharm Community Edition 2023.1. Due to confidentiality restrictions, only processed and aggregated datasets are provided.