Replication package for STAR: Structure-Aware Representation Learning for Efficient and Robust UAV-based Geo-Localization

This replication package includes the source codes of the pipeline, source data and results data, demo video, and the complementary material package (mentioned in the paper).  

Model and experimental data for “Prior knowledge-assisted reinforcement learning for vehicle platoon control in cut-in scenarios.” The co-simulation model uses TruckSim and MATLAB/Simulink, and the shared data support the main experimental results presented in the paper.

The dataset is obtained from miniature-vehicle experiments at intersections. It contains the experimental results used to evaluate multi-vehicle cooperation under fully connected and mixed-traffic conditions, including vehicle motion data and the main performance indicators reported in the manuscript.

This replication package contains code necessary to reproduce the results reported in the manuscript "Infrastructure-Assisted Cooperative Decision Model With Priority Awareness at Unsignalized Intersections" (COMMTR-2026-0005).

Hardware: Intel Core i7-11700F processor and an NVIDIA GeForce RTX 3060 Ti 

Replication code associated with the manuscript “Vehicle-Dynamics-Aware Motion Planning for Pothole-Hazard Mitigation” (Journal of Intelligent and Connected Vehicles, Manuscript ID: JICV-2026-0051.R2). The package includes the vehicle-dynamics simulation environment, reinforcement-learning training, comparison baselines, paired evaluation, and statistical reporting scripts required to reproduce the numerical experiments.

ROSE is a PyTorch/MMDetection3D-based framework for robust roadside Camera–LiDAR 3D object detection under adverse weather. The code implements physics-guided cross-modal augmentation (RISA), teacher–student self-supervised coupling, adaptive training analysis, and tools for training, evaluation, and visualization on DAIR-V2X-style datasets.

The replication package for the paper “A multi-agent asynchronous cooperative on-ramp merging strategy with decision-making priorities in mixed traffic” is available. It contains source code, configurations, pretrained checkpoints, simulation outputs, and scripts for the proposed attention-order framework. The package reproduces the proposed model simulations, learning-curve analysis, ordering agreement, safety evaluation, and finite-stage-game Stackelberg Equilibrium.

This code-only replication package accompanies the revised manuscript “Crash Risk Prediction Guided by Large Language Models Using Pre-Crash Trajectories” submitted to Communications in Transportation Research (Manuscript ID: COMMTR-2026-0063.R1).

The package contains the implementation of the lightweight temporal graph student, offline LLM-guided knowledge distillation and fine-tuning workflows, evaluation and deployment benchmarking scripts, configuration files, environment specifications, a documented data interface, and a synthetic end-to-end smoke test. It also includes a Replication Explanatory File and a SHA-256 integrity manifest.

The basic code of RAVE (End-to-end Hierarchical Visual Localization with Rasterized and Vectorized HD Map)

This replication package contains all data and code necessary to reproduce the results reported in the manuscript "Physics-Informed Platooning with Learning-Augmented Calibration and Compensation: A Real-World Study" (COMMTR-2026-0163).

Hardware: Intel Core i5-9400F, 32GB RAM
Software: Python 3.8+, PyTorch 1.12+, d3rlpy 2.0+, Ubuntu 20.04/ROS Noetic

This paper introduces a novel deep learning framework that enhances vehicle re-identification (ReID) accuracy by integrating visual and temporal data. Vehicle ReID, which identifies target vehicles from large volumes of traffic data, is essential for continuous tracking in large-scale monitoring scenarios involving multiple Unmanned Aerial Vehicles (UAVs). UAV-based monitoring, while offering a comprehensive bird’s-eye view (BEV), faces key challenges: loss of uniquely identifiable features and reliance on visual data, which struggles with vehicles of similar appearance. To overcome these issues, our approach incorporates traffic-oriented features based on shockwave theory to model predictable vehicle travel times. Methods have been tested with data from one of the largest drone experiments with 10 drones monitoring 20 intersections for one week in the city of Songdo in Seoul Area. Experimental results demonstrate a 36.8\% improvement in ReID accuracy over traditional methods, highlighting the potential of UAV-based solutions for robust and scalable traffic monitoring.

This replication package includes all materials necessary for readers to understand and reproduce the analyses presented in the manuscript. As this study is a review paper that does not involve empirical datasets, experimental code, or algorithmic simulations, the materials provided focus on enabling transparent replication of the literature search and bibliometric procedures used in the study. All literature used in the analysis is fully cited within the review paper, ensuring complete traceability of the sources.

This paper proposes a lightweight vision and language large model based approach for pedestrian crossing intention prediction.We introduce a two-stage tuning strategy designed to enhance the model’s explicit and implicit reasoning capabilities. In our experiments, we first use Gemini to generate chain-of-thought annotations from the raw data. We then fine-tune a lightweight Qwen 2.5 model on the resulting CoT dataset using the proposed two-stage procedure. Experiments are conducted on two public datasets, PIE and JAAD. This replication package includes all materials required for readers to understand and reproduce the analyses reported in the paper.

This paper introduces latency-aware trajectory prediction, a new task that explicitly accounts for latency and repurposes it as a useful signal. We present LatenAux, a consolidated auxiliary learning paradigm that first decouples prediction into two tasks: a primary task that predicts valid-horizon trajectories from historical data, and an auxiliary task that utilizes latency-inclusive observations. By allowing the auxiliary branch access to latency-crafted inputs, LatenAux is then committed to transferring latency-aware knowledge to the primary branch via a progressive feature alignment strategy. Extensive experiments on two large-scale real-world datasets demonstrate the effectiveness and superiority of LatenAux, showing that it consistently supports latency-aware modeling and delivers more accurate and reliable trajectory forecasts.

Accurate short-horizon trajectory prediction is crucial for safe and reliable autonomous driving. However, existing vision-language models (VLMs) often fail to accurately understand driving scenes and generate trustworthy trajectories. To address this challenge, this paper introduces KEPT, a knowledge-enhanced VLM framework that predicts ego trajectories directly from consecutive front-view driving frames. KEPT integrates a temporal frequency–spatial fusion (TFSF) video encoder, which is trained via self-supervised learning with hard-negative mining, with a k-means & HNSW retrieval-augmented generation (RAG) pipeline. Retrieved prior knowledge is added into chain-of-thought (CoT) prompts with explicit planning constraints, while a triple-stage fine-tuning paradigm aligns the VLM backbone to enhance spatial perception and trajectory prediction capabilities. This replication package includes all materials required for readers to understand and reproduce the analyses reported in the paper.

All the materials source coding and results coding of  Vehicle-Infrastructure Cooperative General Object Detection through Feature Flow and Differentiable Pose-based Spatial Alignment

This replication package contains the dataset (CTSU) and source code used in TrafficPerceiver.