Vehicle Trajectory Prediction Method for Task Offloading in Vehicular Edge Computing

Real-time computation tasks in vehicular edge computing (VEC) provide convenience for vehicle users. However, the efficiency of task offloading seriously affects the quality of service (QoS). The predictive-mode task offloading is limited by computation resources, storage resources and the timeliness of vehicle trajectory data. Meanwhile, machine learning is difficult to deploy on edge servers. In this paper, we propose a vehicle trajectory prediction method based on the vehicle frequent pattern for task offloading in VEC. First, in the initialization stage, a T-pattern prediction tree (TPPT) is constructed based on the historical vehicle trajectory data. Secondly, when predicting the vehicle trajectory, the vehicle frequent itemset with the largest vehicle trajectory support is found in the vehicle frequent itemset of the TPPT. Finally, in the update stage, the TPPT is updated in real time with the predicted vehicle trajectory results. Meanwhile, based on the proposed prediction method, the strategies of task offloading and optimization algorithm are designed to minimize energy consumption with time constraints. The experiments are carried out on real-vehicle datasets and the Capital Bikeshare datasets. The results show that compared with the baseline T-pattern method, the accuracy of the prediction method is improved by more than 10% and the prediction efficiency is improved by more than 6.5 times. The vehicle trajectory prediction method based on the vehicle frequent pattern has high accuracy and prediction efficiency, which can solve the problem of vehicle trajectory prediction for task offloading.

Research on an Adaptive Method for the Angle Calibration of Roadside LiDAR Point Clouds

Light Detection and Ranging (LiDAR), a laser-based technology for environmental perception, finds extensive applications in intelligent transportation. Deployed on roadsides, it provides real-time global traffic data, supporting road safety and research. To overcome accuracy issues arising from sensor misalignment and to facilitate multi-sensor fusion, this paper proposes an adaptive calibration method. The method defines an ideal coordinate system with the road's forward direction as the X-axis and the intersection line between the vertical plane of the X-axis and the road surface plane as the Y-axis. This method utilizes the Kalman filter (KF) for trajectory smoothing and employs the random sample consensus (RANSAC) algorithm for ground fitting, obtaining the projection of the ideal coordinate system within the LiDAR system coordinate system. By comparing the two coordinate systems and calculating Euler angles, the point cloud is angle-calibrated using rotation matrices. Based on measured data from roadside LiDAR, this paper validates the calibration method. The experimental results demonstrate that the proposed method achieves high precision, with calculated Euler angle errors consistently below 1.7%.

Hierarchical Clustering Algorithm for Multi-Camera Vehicle Trajectories Based on Spatio-Temporal Grouping under Intelligent Transportation and Smart City

With the emergence of intelligent transportation and smart city system, the issue of how to perform an efficient and reasonable clustering analysis of the mass vehicle trajectories on multi-camera monitoring videos through computer vision has become a significant area of research. The traditional trajectory clustering algorithm does not consider camera position and field of view and neglects the hierarchical relation of the video object motion between the camera and the scenario, leading to poor multi-camera video object trajectory clustering. To address this challenge, this paper proposed a hierarchical clustering algorithm for multi-camera vehicle trajectories based on spatio-temporal grouping. First, we supervised clustered vehicle trajectories in the camera group according to the optimal point correspondence rule for unequal-length trajectories. Then, we extracted the starting and ending points of the video object under each group, hierarchized the trajectory according to the number of cross-camera groups, and supervised clustered the subsegment sets of different hierarchies. This method takes into account the spatial relationship between the camera and video scenario, which is not considered by traditional algorithms. The effectiveness of this approach has been proved through experiments comparing silhouette coefficient and CPU time.

Outlier Vehicle Trajectory Detection Using Deep Autoencoders in Santiago, Chile

In the last decade, a large amount of data from vehicle location sensors has been generated due to the massification of GPS systems to track them. This is because these sensors usually include multiple variables such as position, speed, angular position of the vehicle, etc., and, furthermore, they are also usually recorded in very short time intervals. On the other hand, routes are often generated so that they do not correspond to reality, due to artifacts such as buildings, bridges, or sensor failures and where, due to the large amount of data, visual analysis of human expert is unable to detect genuinely anomalous routes. The presence of such abnormalities can lead to faulty sensors being detected which may allow sensor replacement to reliably track the vehicle. However, given the reliability of the available sensors, there are very few examples of such anomalies, which can make it difficult to apply supervised learning techniques. In this work we propose the use of unsupervised deep neural network models based on stacked autoencoders to detect anomalous routes in vehicles within Santiago de Chile. The results show that the proposed model is capable of effectively detecting anomalous paths in real data considering validation given by an expert user, reaching a performance of 82.1% on average. As future work, we propose to incorporate the use of Long Short-Term Memory (LSTM) and attention-based networks in order to improve the detection of anomalous trajectories.

A Comparative Study of Frequent Pattern Mining with Trajectory Data

Sequential pattern mining (SPM) is a major class of data mining topics with a wide range of applications. The continuity and uncertain nature of trajectory data make it distinctively different from typical transactional data, which requires additional data transformation to prepare for SPM. However, little research focuses on comparing the performance of SPM algorithms and their applications in the context of trajectory data. This study selected some representative sequential pattern mining algorithms and evaluated them with various parameters to understand the effect of the involved parameters on their performances. We studied the resultant sequential patterns, runtime, and RAM consumption in the context of the taxi trajectory dataset, the T-drive dataset. It was demonstrated in this work that a method to discretize trajectory data and different SPM algorithms were performed on trajectory databases. The results were visualized on actual Beijing road maps, reflecting traffic congestion conditions. Results demonstrated contiguous constraint-based algorithms could provide a concise representation of output sequences and functions at low min_sup with balanced RAM consumption and execution time. This study can be used as a guide for academics and professionals when determining the most suitable SPM algorithm for applications that involve trajectory data.

Risky Driving Behavior Recognition Based on Vehicle Trajectory

This paper proposes a measurement of risk (MOR) method to recognize risky driving behavior based on the trajectory data extracted from surveillance videos. Three types of risky driving behavior are studied in this paper, i.e., speed-unstable driving, serpentine driving, and risky car-following driving. The risky driving behavior recognition model contains an MOR-based risk evaluation model and an MOR threshold selection method. An MOR-based risk evaluation model is established for three types of risky driving behavior based on driving features to quantify collision risk. Then, we propose two methods, i.e., the distribution-based method and the boxplot-based method, to determine the threshold value of the MOR to recognize risky driving behavior. Finally, the trajectory data extracted from UAV videos are used to validate the proposed model. The impact of vehicle types is also taken into consideration in the model. The results show that there are significant differences between threshold values for cars and heavy trucks when performing speed-unstable driving and risky car-following driving. In addition, the difference between the proportion of recognized risky driving behavior in the testing dataset compared with that in the training dataset is limited to less than 3.5%. The recognition accuracy of risky driving behavior with the boxplot- and distribution-based methods are, respectively, 91% and 86%, indicating the validation of the proposed model. The proposed model can be widely applied to risky driving behavior recognition in video-based surveillance systems.

Improve Aggressive Driver Recognition Using Collision Surrogate Measurement and Imbalanced Class Boosting

Real-time recognition of risky driving behavior and aggressive drivers is a promising research domain, thanks to powerful machine learning algorithms and the big data provided by in-vehicle and roadside sensors. However, since the occurrence of aggressive drivers in real traffic is infrequent, most machine learning algorithms treat each sample equally and prone to better predict normal drivers rather than aggressive drivers, which is our real interest. This paper aims to test the advantage of imbalanced class boosting algorithms in aggressive driver recognition using vehicle trajectory data. First, a surrogate measurement of collision risk, called Average Crash Risk (ACR), is proposed to calculate a vehicle’s crash risk. Second, the driver’s driving aggressiveness is determined by his/her ACR with three anomaly detection methods. Third, we train classification models to identify aggressive drivers using partial trajectory data. Three imbalanced class boosting algorithms, SMOTEBoost, RUSBoost, and CUSBoost, are compared with cost-sensitive AdaBoost and cost-sensitive XGBoost. Additionally, we try two resampling techniques with AdaBoost and XGBoost. Among all algorithms tested, CUSBoost achieves the highest or the second-highest Area Under Precision-Recall Curve (AUPRC) in most datasets. We find the discrete Fourier coefficients of gap as the key feature to identify aggressive drivers.

Determination of Vehicle Trajectory through Optimization of Vehicle Bounding Boxes Using a Convolutional Neural Network

In this manuscript, a new method for the determination of vehicle trajectories using an optimal bounding box for the vehicle is developed. The vehicle trajectory is extracted using images acquired from a camera installed at an intersection based on a convolutional neural network (CNN). First, real-time vehicle object detection is performed using the YOLOv2 model, which is one of the most representative object detection algorithms based on CNN. To overcome the inaccuracy of the vehicle location extracted by YOLOv2, the trajectory was calibrated using a vehicle tracking algorithm such as a Kalman filter and intersection-over-union (IOU) tracker. In particular, we attempted to correct the vehicle trajectory by extracting the center position based on the geometric characteristics of a moving vehicle according to the bounding box. The quantitative and qualitative evaluations indicate that the proposed algorithm can detect the trajectories of moving vehicles better than the conventional algorithm. Although the center points of the bounding boxes obtained using the existing conventional algorithm are often outside of the vehicle due to the geometric displacement of the camera, the proposed technique can minimize positional errors and extract the optimal bounding box to determine the vehicle location.

A Method for Extracting Road Boundary Information from Crowdsourcing Vehicle GPS Trajectories

Crowdsourcing trajectory data is an important approach for accessing and updating road information. In this paper, we present a novel approach for extracting road boundary information from crowdsourcing vehicle traces based on Delaunay triangulation (DT). First, an optimization and interpolation method is proposed to filter abnormal trace segments from raw global positioning system (GPS) traces and interpolate the optimization segments adaptively to ensure there are enough tracking points. Second, constructing the DT and the Voronoi diagram within interpolated tracking lines to calculate road boundary descriptors using the area of Voronoi cell and the length of triangle edge. Then, the road boundary detection model is established integrating the boundary descriptors and trajectory movement features (e.g., direction) by DT. Third, using the boundary detection model to detect road boundary from the DT constructed by trajectory lines, and a regional growing method based on seed polygons is proposed to extract the road boundary. Experiments were conducted using the GPS traces of taxis in Beijing, China, and the results show that the proposed method is suitable for extracting the road boundary from low-frequency GPS traces, multi-type road structures, and different time intervals. Compared with two existing methods, the automatically extracted boundary information was proved to be of higher quality.

A Data Cleaning Method for Big Trace Data Using Movement Consistency

Given the popularization of GPS technologies, the massive amount of spatiotemporal GPS traces collected by vehicles are becoming a new kind of big data source for urban geographic information extraction. The growing volume of the dataset, however, creates processing and management difficulties, while the low quality generates uncertainties when investigating human activities. Based on the conception of the error distribution law and position accuracy of the GPS data, we propose in this paper a data cleaning method for this kind of spatial big data using movement consistency. First, a trajectory is partitioned into a set of sub-trajectories using the movement characteristic points. In this process, GPS points indicate that the motion status of the vehicle has transformed from one state into another, and are regarded as the movement characteristic points. Then, GPS data are cleaned based on the similarities of GPS points and the movement consistency model of the sub-trajectory. The movement consistency model is built using the random sample consensus algorithm based on the high spatial consistency of high-quality GPS data. The proposed method is evaluated based on extensive experiments, using GPS trajectories generated by a sample of vehicles over a 7-day period in Wuhan city, China. The results show the effectiveness and efficiency of the proposed method.