Trajectory Dataset Research Articles

An Asymmetric Selective Kernel Network for Drone-Based Vehicle Detection to Build a High-Accuracy Vehicle Trajectory Dataset

An Asymmetric Selective Kernel Network for Drone-Based Vehicle Detection to Build a High-Accuracy Vehicle Trajectory Dataset

Four-Dimensional Aircraft Trajectory Prediction with a Generative Deep Learning and Clustering Approach

Medium- and long-term four-dimensional (4D) aircraft trajectory prediction (TP) is a critical technology in air traffic management (ATM). This paper addresses the issue of existing medium- and long-term TP methods that are difficult to accurately fit aircraft trajectory data distributions. We propose a 4D TP method based on K-medoids clustering and conditional tabular generative adversarial networks (CTGAN), called C-CTGAN. Comparative experiments with four long short-term memory (LSTM)-based models and the original CTGAN model show that the proposed model’s TP accuracy is significantly higher than others when predicting medium- and long-term trajectories. When using the trajectory datasets without holding and a prediction time span of 10 min, compared to the convolutional neural network (CNN)-LSTM model, the C-CTGAN model reduces the mean absolute errors (MAEs) of core trajectory parameters, such as latitude, longitude, geometric altitude, and ground speed, by 69.89, 15.00, 74.07, and 84.21%, respectively. Compared to the original CTGAN model, the MAE is reduced by 20.43, 39.09, 31.98, and 17.07%, respectively. When using the trajectory datasets with holding, compared to the CNN-LSTM model, the C-CTGAN model shows MAE reductions of 14.08, 23.68, 31.46, and 2.86%, respectively. Compared to the original CTGAN, the reduction is 34.88, 2.69, 23.16, and 73.91%, respectively.

A COMBINATION OF DEEP LEARNING AND DENSIIY METHOD IN ANOMALOUS HUMAN TRAJECTORY DETECTION

Abnormal human trajectories in working places are often associated with problems such as terrorism, violent attacks and fire. Therefore, detecting anomalous human trajectories can improve safety and security in working areas. In this work, a novel framework of abnormal trajectory detection is proposed based on combining deep learning and density method. In particular, a Long Short-Term Memory-Autoencoder (LSTM-AE) is first applied to learn informative representations of normal trajectories. Then, the density of trajectory representation in the latent space and reconstruction error of trajectory are used to detect anomalies. A novel metric is also proposed to determine the anomaly scores of trajectories. The proposed framework is evaluated using two real trajectory datasets: the MIT Badge and the sCREEN datasets. The experimental results show that our work effectively detects anomalies, achieving a f1-score of 81.08% on the MIT Badge dataset and 89.57% on the sCREEN dataset.

Aggressive Driving Behavior Recognition based on K-means Clustering

Accurate identification of aggressive driving behavior is crucial for improving traffic safety and promoting social responsibility, and it is also a key component in the realization of intelligent transportation systems. This paper proposes a method for aggressive driving behavior recognition and data modeling based on K-means clustering. First, the NGSIM (Next Generation Simulation) vehicle trajectory dataset is used for preprocessing the raw data, and effective behavioral feature representations are obtained through feature extraction. Next, Principal Component Analysis (PCA) is applied to reduce the dimensionality of the data, extracting three main components to simplify the data structure. Subsequently, the K-means clustering algorithm is employed to classify the reduced data and identify potential patterns of aggressive driving behavior. Experimental results demonstrate that the proposed method effectively identifies aggressive driving behavior, providing technical support for improving traffic system safety and raising awareness of social responsibility.

VECTOR: Velocity-Enhanced GRU Neural Network for Real-Time 3D UAV Trajectory Prediction

This paper addresses the challenge of predicting 3D trajectories for Unmanned Aerial Vehicles (UAVs) in real-time, a critical task for applications like aerial surveillance and defense. Current prediction models primarily leverage only position data, which may not provide the most accurate forecasts for UAV movements and usually fail outside the position domain used in the training phase. Our research identifies a gap in utilizing velocity estimates and first-order dynamics to better capture the dynamics and enhance prediction accuracy and generalizability in any position domain. To bridge this gap, we introduce a trajectory prediction scheme using sequence-based neural networks with Gated Recurrent Units (GRUs) to forecast future velocity and positions based on historical velocity estimates instead of position measurements. This approach is designed to improve the predictive capabilities over traditional methods that rely solely on recurrent neural networks (RNNs) or transformers, which can struggle with scalability in this context. Our methodology employs both synthetic and real-world 3D UAV trajectory data, incorporating diverse patterns of agility, curvature, and speed. Synthetic data are generated using the Gazebo robotics simulator and PX4 Autopilot, while real-world data are sourced from the UZH-FPV and Mid-Air drone racing datasets. We train the GRU-based models on drone 3D position and velocity samples to capture the dynamics of UAV movements effectively. Quantitatively, the proposed GRU-based prediction algorithm demonstrates superior performance, achieving a mean square error (MSE) ranging from 2×10−8 to 2×10−7. This performance outstrips existing state-of-the-art RNN models. Overall, our findings confirm the effectiveness of incorporating velocity data in improving the accuracy of UAV trajectory predictions across both synthetic and real-world scenarios, in and out of position data distributions. Finally, we open-source our 5000 trajectories dataset and a ROS2 package to facilitate the integration with existing ROS-based UAV systems.

Map-Matching Algorithm Based on Hidden Markov and Constraint Value Pruning

Map matching is the process of matching global positioning system (GPS) trajectory data with map data. Its purpose is to determine the actual route of the moving object. Because of factors such as positioning devices and the environment, the GPS trajectory data obtained may not be accurate. In location-based services, map matching can be used to address the accuracy and reliability issues of GPS location data. There are currently many map-matching algorithms, for example, spatio-temporal-based (STD) matching algorithm, improved interactive voting-based map-matching (IIVMM) algorithm, and turning-point-based (TPB) offline map-matching algorithm, but existing algorithms have some shortcomings, such as low matching accuracy on complex roads or low sampling rates, and routing calculation time bottlenecks. Therefore, this paper proposes a fast matching algorithm based on constrained value pruning that is suitable for complex roads. The algorithm considers the multi-directional features within the road and uses secondary calculations to determine candidate points, improving the accuracy of candidate point selection. Additionally, two pruning strategies based on constrained values are introduced to reduce the majority of the routing calculation process and improve the matching efficiency. Finally, comparative experiments are conducted on a real trajectory dataset. The results show that, compared with STD, IIVMM, and TPB algorithms, the algorithm accuracy is improved by about 2% to 4%, and the running time is reduced by about 30%.

Agricultural Machinery Movement Trajectory Recognition Method Based on Two-Stage Joint Clustering

To address the challenges posed by the large scale of agricultural machinery trajectory data and the complexity of actual movement trajectories, this paper proposes a two-stage joint clustering method for agricultural machinery trajectory recognition to enhance accuracy and robustness. The first stage involves trajectory clustering, where the spatial distribution characteristics of agricultural machinery trajectories are analyzed, and the position coordinates and the number of neighboring points of trajectory points are extracted as features. The silhouette coefficient method is used to determine the optimal number of clusters k for the K-Means algorithm, thus reducing the data scale. The second stage focuses on trajectory recognition, where a list of Eps and Minpts parameters is generated based on the statistical properties of the trajectory dataset. The Genetic Algorithm is employed for parameter optimization to determine the optimal DBSCAN parameters, enabling precise identification of field operation trajectories and road travel trajectories. Experimental results show that this method achieves mean values of 91.55% for Accuracy, 95.41% for Precision, 89.86% for Recall, and 92.41% for F1-score on a sample dataset of 337 trajectories, representing improvements of 12.8%, 5.13%, 7.79%, and 6.84%, respectively, over the traditional DBSCAN algorithm. Additionally, the Runtime of the two-stage joint clustering method is approximately 30% shorter than that of single-stage clustering. Compared with mainstream deep learning models such as LSTM and Transformer, this method delivers comparable recognition accuracy without the need for labeled data training, significantly reducing recognition costs. The proposed method achieves accurate and robust recognition of agricultural machinery trajectories and holds broad application potential in practical scenarios.

Medark: a map-matching error detection and rectification framework for vehicle trajectories

The widespread use of Global Navigation Satellite System (GNSS) trackers has significantly enhanced the availability of vehicle tracking data, providing researchers with critical insights into human mobility. Map matching, a key preprocessing step in movement analysis, matches vehicle tracking data to road segments but often introduces errors that can affect subsequent analyses. Existing map-matching methods, categorized into classic spatially generalizable methods and region-specific deep-learning-based methods, both have limitations. Region-specific deep learning methods, while more accurate, do not transfer well across different geographical regions. Moreover, the temporal adaptability of both approaches—their ability to handle GNSS signals of varying sampling intervals—has not been thoroughly examined. To overcome these limitations, we introduce Medark, a novel framework for detecting and rectifying errors in classic map-matching methods while preserving spatial generalizability. The proposed model is trained using a transfer-learning approach with synthetic trajectories generated in Ann Arbor and Los Angeles at various sampling intervals and a real vehicle trajectory dataset from Ann Arbor. Our experimental results validate the effectiveness of Medark. This framework can be integrated with any map-matching method to improve accuracy and produce high-quality trajectories for further analysis.

Vehicle trajectory dataset from drone videos including off-ramp and congested traffic – Analysis of data quality, traffic flow, and accident risk

Vehicle trajectory data have become essential for many research fields, such as traffic flow, traffic safety, and automated driving. To make trajectory data useable for researchers, an overview of the included road section and traffic situation as well as a description of the data processing methodology is necessary. In this paper, we present a trajectory dataset from a German highway with two lanes per direction, an off-ramp and congested traffic in one direction, and an on-ramp in the other direction. The dataset contains 8,648 trajectories and covers 87 ​min and an ∼1,200 ​m long section of the road. The trajectories were extracted from drone videos using a posttrained YOLOv5 object detection model and projected onto the road surface via three-dimensional (3D) camera calibration. The postprocessing methodology can compensate for most false detections and yield accurate speeds and accelerations. The trajectory data are also compared with induction loop data and vehicle-based smartphone sensor data to evaluate the plausibility and quality of the trajectory data. The deviations of the speeds and accelerations are estimated at 0.45 ​m/s and 0.3 ​m/s2, respectively. We also present some applications of the data, including traffic flow analysis and accident risk analysis.

Masked Graph Autoencoder for Self‐Supervised Transportation Mode Recognition

ABSTRACTTransportation mode recognition reveals the traveling patterns of urban residents and plays an important role in travel analysis and city planning. Research has been done to identify transportation modes using deep learning methods. However, large‐scale and far‐ranging labeled trajectory data are hard to acquire, which hinders the training and application of supervised models. In this study, we propose a self‐supervised method for trajectory transportation modes recognition. First, we construct sequence and dependency graph for trajectory points. Next, we apply mask and remask strategy to graph node features to enhance its learning ability in semantic information. Notably, in order to learn features without labeled data, we introduce graph autoencoder into the framework. This method is able to learn the sequence and dependency feature reconstruction of the travel paths. It can obtain underlying semantic information with the mask and remask strategy. We applied our model to GPS trajectory dataset from the Microsoft GeoLife Project, achieving an average accuracy of 68.26% in identifying the transportation modes among seven categories. The result indicates that the proposed method offers a solution to identifying transportation modes without abundant labeled data. This method can also help to extract space–time features at different scales from trajectory data, which could be applied to various downstream tasks.

An approach for accurately extracting vehicle trajectory from aerial videos based on computer vision

Vehicle trajectory data holds valuable information for advanced driving development and traffic analysis. While unmanned aerial vehicle (UAV) offer a broader perspective, the detection of small-scale vehicles in video frames still suffers from low accuracy or is even missed. This study proposes a comprehensive technical framework for accurate vehicle trajectory extraction, encompassing six main components: video stabilization, vehicle detection, vehicle tracking, lane marking detection, coordinate transformation, and data denoising. To mitigate video jitter, the SURF and FLANN stabilization algorithms are utilized. An enhanced detector based on You Only Look Once X (YOLOX) is employed for multi-target vehicle detection, incorporating a shallow feature extraction module within the detection head to improve the performance for low-level and small-scale features. Efficient Channel Attention (ECA) modules are integrated before the neck to further boost the expressiveness. Additionally, a sliding window inference method is applied at the input stage to prevent compression of high-resolution video frames. The Savitzky-Golay filter is used for trajectory noise reduction. Verification results demonstrate that the improved YOLOX achieves a mean average precision (mAP) of 88.7 %, an enhancement of 5.6 % over the original model. When compared to advanced YOLOv7 and YOLOv8 models, the proposed method increases mAP@50 by 7.63 % and 1.07 %, respectively. The Mostly Tracked (MT) trajectories metric reaches 98.9 %, and the root-mean-square error of one-sided localization is approximately 0.05 m. These results confirm that the proposed framework is an effective tool for high-accuracy vehicle trajectory data collection in traffic studies. Additionally, a vehicle trajectory dataset has been developed and is publicly accessible at www.cqskyeyex.com.

A general image classification model for agricultural machinery trajectory mode recognition

Field-road trajectory classification is a crucial task for agricultural machinery behavior mode recognition, aiming to distinguish field operation mode and road driving mode automatically. However, the imbalanced distribution of agricultural machine trajectories brings challenges for the field-road trajectory classification task. Additionally, most existing field-road trajectory classification methods have certain shortcomings. For instance, they encounter difficulties in accurately representing the state of agricultural machinery movement using the current features. The data transformation process often leads to information loss, and the model’s generalization capabilities are limited. The performance of the models is constrained by each of these elements. To address these shortcomings, this paper introduces a general image classification model for agricultural machinery trajectory mode recognition named ATRNet. First, to address the issue of imbalanced field-road proportions in agricultural machinery trajectory data, a Conditional Tabular Generative Adversarial Network (CTGAN) is employed to generate quasi trajectories, balancing the distribution of positive and negative samples in the data. This step aims to eliminate biases during the model training process. Second, to accurately characterize the motion status of agricultural machinery, we propose a multiangle feature enhancement method to extract rich spatiotemporal features from trajectory data. Finally, different from conventional field-road trajectory classification models that primarily rely on spatial and temporal information for identifying trajectories, we present a lossless trajectory data representation paradigm. This paradigm maps each trajectory point into a “feature map” and uses an image classification model to capture latent feature representations of trajectory points for the recognition of different behavior modes of agricultural machinery. This paradigm can generalize image classification networks to the field-road trajectory classification task, providing a general vision model solution for agricultural machinery trajectory mode recognition. To validate the effectiveness of the ATRNet model, experiments were conducted on real corn and wheat harvester trajectory datasets. The results demonstrate that the proposed model achieves remarkable performance improvements over the state-of-the-art (SOTA) models. In the corn harvester trajectory dataset, ATRNet achieves an accuracy of 92.36% and an F1-score of 92.34%, surpassing existing SOTA models by 3.12% and 12.46%, respectively. Similarly, in the wheat harvester trajectory dataset, ATRNet achieves an accuracy of 92.36% and an F1-score of 92.33%, outperforming the existing optimal algorithm by 4.76% and 18.18%, respectively.

Path following control of under-actuated autonomous surface vehicle based on random motion trajectory dataset and offline reinforcement learning

To solve the path following problem in navigation tasks for under-actuated autonomous surface vehicles (ASVs), this paper proposed a path following control method which combines trajectory dataset of random ship motion and offline reinforcement learning (RM-ORL). The method does not require the reinforcement learning (RL) agent to interact with the environment while training the policy, and it can obtain training datasets with a lower cost. In RM-ORL, the irregular motion data of the ASV in open water is first collected. Then the desired path is reconstructed using the B-spline function and the path points along the motion trajectories. Thus the offline dataset will be enhanced with the motion data and the new path. Finally, the conservative Q-learning algorithm is utilized to train the path following controller. The path deviation in simulation maps, rudder data and ship motion parameters of RM-ORL, online RL and other offline RL policies trained on different datasets are compared. The simulation results illustrate that the RM-ORL achieves comparable path following accuracy to that of online RL agent and offline RL agent trained on expert data, while surpassing the one trained on online agent replay buffer data. The rudder steering amplitude of RM-ORL is also smaller than that of other policies, which verifies the effectiveness of our method applied to the path following control of under-actuated ASV.

Towards effective, robust and utility-preserving watermarking of GPS trajectories

Personal GPS trajectory is essential for businesses and emerging data markets due to its relevance in various data-driven methods, including traffic forecasting, accident prediction, and profiling driving behavior. Watermarking is a method that facilitates verification of data ownership and authenticity by embedding provenance information into the data. Whereas watermarking is commonly adopted in the image and audio domains, only a few initial watermarking methods exist for GPS trajectory data. GPS trajectory watermarking is particularly challenging due to the spatio-temporal data properties and easiness of data modification. As a result, existing watermarking methods often embed only minimal provenance information, lack robustness, and can fail to preserve data utility for downstream applications. In this work, we propose W-Trace - a novel, effective, robust, and utility-preserving GPS trajectory watermarking method. W-Trace transforms a GPS trajectory into a complex domain and applies the Fourier transformation to decompose the trajectory into the frequency representation. W-Trace embeds watermarks in the frequency representation and verifies them in a spatiotemporally-aware procedure. We demonstrate the effectiveness, robustness, and utility of the proposed W-Trace approach in realistic settings using real-world GPS trajectory datasets. In contrast to the baselines, the proposed W-Trace approach is robust to a wide range of trajectory modifications while preserving the GPS trajectory characteristics required for the downstream applications.

Distributed multi-object tracking under limited field of view heterogeneous sensors with density clustering

We consider the problem of tracking multiple, unknown, and time-varying numbers of objects using a distributed network of heterogeneous sensors. In an effort to derive a formulation for practical settings, we consider limited and unknown sensor field-of-views (FoVs), sensors with limited local computational resources and communication channel capacity. The resulting distributed multi-object tracking algorithm involves solving an NP-hard multidimensional assignment problem either optimally for small-size problems or sub-optimally for general practical problems. For general problems, we propose an efficient distributed multi-object tracking algorithm that performs track-to-track fusion using a clustering-based analysis of the state space transformed into a density space to mitigate the complexity of the assignment problem. The proposed algorithm can more efficiently group local track estimates for fusion than existing approaches. To ensure we achieve globally consistent identities for tracks across a network of nodes as objects move between FoVs, we develop a graph-based algorithm to achieve label consensus and minimise track segmentation. Numerical experiments with synthetic and real-world trajectory datasets demonstrate that our proposed method is significantly more computationally efficient than state-of-the-art solutions, achieving similar tracking accuracy and bandwidth requirements but with improved label consistency.

Trajectory representation learning with multilevel attention for driver identification

Massive trajectory data has originated from the development of positioning technology. Learning GPS trajectory representation to characterize a driver’s driving style is a challenging task with important applications in many areas, including autonomous driving, auto insurance, advanced driver assistance systems, urban computing, and the internet of things. Few studies have considered the interactions between different factors. In this study, we propose a novel trajectory representation method based on a multilevel attention mechanism (ATTraj2vec) and apply it to the task of driver identification. We use 1D CNN to summarize local features, including motion, spatial and temporal features. Then we utilize a multilevel attention mechanism to extract global features, aggregating the interactions of motion features with temporal and spatial features progressively. Additionally, we adopt multi-loss to optimize our model simultaneously, which consists of a softmax loss for driver classification and Siamese loss for making trajectories from the same driver more similar. Classification experimental results on a real-world automobile trajectory dataset demonstrate that our proposed model significantly outperforms existing baselines. Meanwhile, the proposed method provides significant gains in the trajectory clustering of unseen drivers. The code is available in the repository, https://github.com/limengyuan2021/ATTraj2vec,

Field-road classification for agricultural vehicles in China based on pre-trained visual model.

Field-road classification that automatically identifies the activity (either in-field or on-road) of each point in Global Navigation Satellite System (GNSS) trajectories is a critical process in the behavior analysis of agricultural vehicles. To capture movement patterns specific to agricultural operations, we propose a multi-view field-road classification method, which extracts a physical and a visual feature vector to represent a trajectory point. We propose a task-specific approach using a pre-trained visual model to effectively extract visual features. Firstly, an image is generated based on a point plus its neighboring points to provide the contextual information of the point. Then, an image recognition model, a fine-tuned ResNet model is developed using the pretraining-finetuning paradigm. In such a paradigm, a pre-training process is used to train an image recognition model (ResNet) with natural image datasets (e.g., ImageNet), and a fine-tuning process is applied to update the parameters of the pre-trained model using the trajectory point images, enabling the model to have both general knowledge and task-specific knowledge. Finally, a visual feature is extracted for a point by the fine-tuned model, thereby overcoming the limitations caused by the small-scale generated images. To validate the effectiveness of our multi-view field-road classification, we conducted experiments on four trajectory datasets (Wheat 2021, Paddy, Wheat 2023, and Wheat 2024). The results demonstrated that the proposed method achieves competitive accuracy performance, i.e., 92.56%, 87.91%, 90.31%, and 94.23% on four trajectory datasets, respectively. Extensive experiments demonstrate that our approach can consistently perform better than the existing state-of-the-art method on the four trajectory datasets by 2.99%, 4.42%, 2.88%, and 2.77% in the F1-score, respectively. In addition, we conduct an in-depth analysis to verify the necessity and effectiveness of our method.

A unified longitudinal trajectory dataset for automated vehicle

Automated Vehicles (AVs) promise significant advances in transportation. Critical to these improvements is understanding AVs’ longitudinal behavior, relying heavily on real-world trajectory data. Existing open-source trajectory datasets of AV, however, often fall short in refinement, reliability, and completeness, hindering effective performance metrics analysis and model development. This study addresses these challenges by creating a Unified longitudinal trajectory dataset for AVs (Ultra-AV) to analyze their microscopic longitudinal driving behaviors. This dataset compiles data from 14 distinct sources, encompassing various AV types, test sites, and experiment scenarios. We established a three-step data processing: 1. extraction of longitudinal trajectory data, 2. general data cleaning, and 3. data-specific cleaning to obtain the longitudinal trajectory data and car-following trajectory data. The validity of the processed data is affirmed through performance evaluations across safety, mobility, stability, and sustainability, along with an analysis of the relationships between variables in car-following models. Our work not only furnishes researchers with standardized data and metrics for longitudinal AV behavior studies but also sets guidelines for data collection and model development.

A dual-level graph attention network and transformer for enhanced trajectory prediction under road network constraints

Predicting vehicle future trajectories enables various Location-Based Services (LBS), including traffic flow monitoring, vehicle route planning, etc. Vehicles are constrained to predefined road networks. Their traveling behaviors are influenced by both their own attributes and road network characteristics hidden in other vehicles’ travel histories. Most of the existing methods fail to sufficiently leverage such collaborative trajectory data under road network constraints. This paper proposes a novel model, called HGT-RN (Hierarchical Graph Attention Network and Transformer Networks for Enhanced Trajectory Prediction under Road Network Constraints), which exploits all vehicles’ trajectories and integrates road network constraints. Specifically, HGT-RN creates a dual-level Graph Attention Network (GAT) based on global traffic flow and local individual trajectory graph to learn trajectory embeddings: (i) Weighted-directional global traffic flow graph and Weighted-Directional GAT to learn the global-level trajectory embedding by aggregating the neighbors’ embeddings based on the spatial transitional relationships among road intersections over all vehicles’ trajectories, considering the preference of current trajectory; and (ii) Local individual trajectory graph attention network to learn the local-level trajectory embedding by modeling the transitions within the current trajectory at each intersection. When incorporating the trajectory embeddings into a transformer model for future trajectory prediction, a road network topology-aware loss function is designed to improve accuracy. Extensive experiments conducted on three city-scale real-world taxi trajectory datasets demonstrate that HGT-RN significantly outperforms existing baseline models. The code is available at https://github.com/zjy9826/HGT-RN.

Empirical Studies on Merging Characteristics in Temporary Highway Work Zones with Different Traffic States and Layouts

Vehicles conduct mandatory lane changes to merge into the open lanes in temporary highway work zones, and the merging characteristics have been widely studied over the past decades. Unlike research on other traffic facilities, the data and scenarios of highway work zones are often lacking and monotonous. This study presented a vehicle trajectory dataset in highway work zones. The dataset contains ten work zone sites and three different layouts. After transforming raw videos into high-resolution trajectories, we selected eight areas and divided them into three scenarios to investigate the merging characteristics in highway work zones with different traffic states and layouts. In work zones with varying traffic conditions, we found opposite merging locations and lane speed distribution trends. In work zones with varying layouts, we found different merging location distributions, lane speed distributions, and impacts on the following vehicle’s speed. The relationship between the gap and merge speed was consistent with previous studies.