2023 Vol. 41, No. 4

Display Method:
2023, 41(4): .
Abstract(387) HTML (334) PDF(24)
A Review of Safety Studies on Lane Change Decision-makings for Connected Automated Vehicles
CUI Bingyan, LI He, CUI Zhe, JI Haojie, GUAN Yuxin
2023, 41(4): 1-13. doi: 10.3963/j.jssn.1674-4861.2023.04.001
Abstract(635) HTML (362) PDF(85)
Safety of lane change decision-makings for connected automated vehicles (CAVs) is a key task to improve traffic safety and enhance road mobility. In this paper, the safety issues related to lane changing of CAVs are investigated. From the perspective of driving safety, the adverse impacts of extreme lane-changing behavior and emergency lane-changing behavior on traffic safety are analyzed, emphasizing the importance of risk assessment. The various risk assessment methods of lane changing are reviewed, including the use of environmental sensors, traffic conflict indicators, and vehicle-level micro-trajectory data. Identifying risks through risk assessment and taking corresponding measures can significantly reduce traffic accidents caused by dangerous lane-changing behavior. Furthermore, the methods for CAVs to make lane-changing decisions by obtaining environmental information in both traditional and vehicle to everything (V2X) environments are elaborated. Particularly focusing on the CAVs in V2X environment, the decision-making through the environment perception and recognition, targets detection, and data processing is analyzed. Reasonable recommendations are proposed for achieving safe decision-making by CAVs in V2X environment in the future. Then the existing models for decision-making of lane-changing are analyzed and categorized into four types: rule-based models, discrete choice models, artificial intelligence models, and game theory models. The status of research and application, existing problems, and prospects of decision-making models in the field of road traffic safety are systematically summarized, both domestically and internationally. In summary, despite significant research achievements in lane-changing technologies for CAVs, there are still many challenges ahead. To tackle the existing problems in research, such as ensuring safe and reliable decisions in low-level automated driving environments, making more efficient and intelligent driving decisions for CAVs in low-penetration scenarios, achieving safe decision-making in situations with incomplete information, and improving the optimization of the algorithms for lane-changing decision-making, feasible solutions are proposed accordingly.
A Method for Evaluating the Safety over the Takeover Process of the Level 3 Automated Vehicles Based on IAHP-EWM-LDM
LI Zhenlong, PAN Mengniu, QU Yansong, ZHAO Xiaohua, GONG Jianguo, WANG Qiuhong
2023, 41(4): 14-23. doi: 10.3963/j.jssn.1674-4861.2023.04.002
Abstract(284) HTML (116) PDF(35)
In the Level 3 autonomous driving stage, the driver needs to respond and take over the vehicle when the system sends a takeover request. Therefore, to accurately assess the safety of the takeover process of Level 3 autonomous vehicles, the safety evaluation index system of the takeover process of autonomous driving is constructed. In this paper, a 4×2×2 takeover scenario factor is used to design a driving simulation test, and a driving simulator is used to collect various types of driving data. Based on the coefficient of variation method and Spearman correlation discriminant method, 13 security evaluation indicators are obtained from the analysis of 3 aspects, such as risk perception, risk avoidance manipulation and takeover performance. The subjective weights of the indicators are obtained using an improved hierarchical analysis that characterizes the experience of the experts, and the subjective weights of the indicators are obtained using entropy weights that reflect the characteristics of the data. To combine the advantages of the two methods, a composite weight incorporating both subjective and objective weights is obtained using the grade maximization method. The combined weights of risk perception, risk avoidance manipulation, and takeover performance are calculated to be 0.259, 0.475, and 0.271, which are used to construct the security evaluation index system of the takeover process. In this paper, the system is applied to comprehensively evaluate 655 takeover processes obtained from driving simulation tests, and they are classified into 3 categories of A, B and C takeover processes according to the evaluation results. Comparing the scores of the 3 types of takeover processes in 3 aspects: risk perception, risk avoidance manipulation and takeover performance, it is found that the A-type takeover process performs well in three aspects, the C-type takeover process performs poorly in risk avoidance manipulation and takeover performance, and the B-type takeover process performs intermediary between the A-type and C-type. Different types of takeover process have a better degree of differentiation in each indicator. The indicator system is constructed that effectively combines expert experience and indicator characteristics. The evaluation index system constructed in this paper effectively combines expert experience and index characteristics. It can provide theoretical support for a more comprehensive, reasonable and scientific evaluation of the safety in the process of automatic driving takeover.
A Study on Longitudinal Collision Risk of Airplanes during Paired Approach Under the Influence of Positioning Error
LU Fei, ZHAO Erli, LIANG Xianyun
2023, 41(4): 24-32. doi: 10.3963/j.jssn.1674-4861.2023.04.003
Abstract(287) HTML (125) PDF(24)
Studying longitudinal collision risk of paired approach on closely spaced parallel runways (CSPRs) is crucial for assessing its safety, where positioning errors directly influence the longitudinal collision risk during the process. Given the lack of consideration on actual data fitting for positioning error distribution in previous studies, this study aims investigate the longitudinal collision risk during paired approach under the influence of actual data-fitted positioning error. According to the implementation process of paired approach, a kinematic model for the longitudinal spacing between aircrafts before and after pairing is established. In terms of positioning error during flight, statistical data of actual aircraft positioning errors are utilized to fit the distribution. Next, utilizing Automatic Dependent Surveillance-Broadcast (ADS-B) data, the longitudinal positioning error during the final approach phase is analyzed and fitted to identify the best-fitting distribution, that is, normal distribution. The collision risk between the aircraft fuselages in paired approach and the collision risk between the wake turbulence of lead aircraft and the fuselage of trailing aircraft are studied separately, and integral intervals for each collision risk model are determined. Based on the normal distribution and the movements of the paired aircrafts during paired approach, an assessment model for the longitudinal collision risk is established. Finally, data about the B737-800 aircraft at Shanghai Hongqiao Airport in December 2020 are collected for a case study. Simulations are conducted to analyze the changes in collision risk of fuselage Px1 and collision risk of wake turbulence Px2 over time under the initial longitudinal separations of 926 m and 2 778 m. Further, the relationship between different initial longitudinal separations and Px1 / Px2 or the maximum value of overall longitudinal collision risk. The results indicate that: ①when the initial longitudinal separation is 926 m, Px1 gradually decreases while Px2 increases over time, and Px1 is significantly greater than Px2. ②When the initial longitudinal separation is 2 778 m, the results are the opposite. ③ Px1 decreases while Px2 increases as the initial longitudinal separation increases. ④The overall longitudinal collision risk between the lead and trailing aircrafts decreases first and then increases with increasing initial longitudinal separation; ⑤when the initial longitudinal separation is smaller than 2 136 m, the longitudinal collision risk is primarily determined by the collision risk between the fuselages of lead and trailing aircrafts; when the initial longitudinal separation is larger than 2 136 m, it is determined by the collision risk between the wake turbulence of lead aircraft and the fuselage of trailing aircraft.
An Analysis of Driving Behavior on Short Distance Section between Tunnel and the Exit of Main Roadway
TANG Hao, TANG Zhongze, ZHANG Chi, WEI Bin, ZHANG Kunlun, YANG Kun
2023, 41(4): 33-43. doi: 10.3963/j.jssn.1674-4861.2023.04.004
Abstract(286) HTML (178) PDF(31)
A short distance section is a section between the tunnel and the mainline exit of a mountainous highway whose length is lower than the normative value because of the limitation of geographical and investment factors. In order to analyze the driving characteristics of this area thus to enhance the theoretical base for mountain road design and traffic control, high-definition driving videos are collected by drones in 7 mountain roads (e.g., Qinling service area) with short distance sections in China. The high-precision speed and trajectory data of vehicles across the entire region have been extracted. The SIFT algorithm is used for video stabilization. The YOLOv5 and DeepSORT algorithms are adopted for vehicle detection and tracking. The Savitzky-Golay filter is utilized to filter the data. Finally, high precision driving data can be obtained based on the above methods. It is verified that the accuracy of speed can reach more than 95% and the error of trajectory is less than 20 cm. Next, the driving characteristics are analyzed from a variety of perspectives, such as clearance distance, vehicle type, lane distribution, and others. The results show that: ①the driving characteristics on short distance section are very different from those on usual road section that the speed distribution does not follow a normal distribution; ②generally the outgoing vehicles would be steady 10 to 20 m ahead of the commencement of the fading phase; ③the speed of trucks is smoother as truck drivers can identify the exit road conditions more quickly than the car drivers because of the larger angle of view; ④approximately 20 meters after the starting point of the transition section, the cars in the inner lane enter the deceleration lane with a lateral speed of 1.1 to 1.4 m/s, and when the mainline has a leftward curve it is most favorable to drive out; ⑤the clearance distance has the highest influence on driving behaviors, the traffic volume affects the most among traffic flow factors while the direction of curve deflection and deflection angle affect the most among the road geometry factors.
Classifying Road Accidents and Forecasting Level of Risk Based on a Combined PCA-LPP and DBSCAN Method
XIN Yi, LI Gang, DENG Youwei, ZHANG Shengpeng, ZHOU Pan, LIU Yiyang
2023, 41(4): 44-54. doi: 10.3963/j.jssn.1674-4861.2023.04.005
Abstract(333) HTML (191) PDF(44)
Road traffic accidents are one of the major problems causing large numbers of casualties and property losses worldwide. By classifying road traffic accidents and predicting risk levels, it becomes possible to identify high-risk vehicles and reduce the probability of accidents and casualties. Traffic accidents are often influenced by multiple factors such as environment, weather, road conditions, and infrastructure, but existing accident impact analysis methods lack comprehensive research on traffic accident data. Therefore, this paper proposes a traffic accident classification model that incorporates an improved dimensionality reduction algorithm called PCA-LPP, which measures the similarity between data of different levels to achieve secondary dimensionality reduction. The model utilizes a large-scale traffic accident dataset and applies the DBSCAN algorithm to partition the accident data into risk areas. By training the spatial representations of different risk levels iteratively, the model could assess the risk levels in simulated vehicle environments. Experimental results demonstrate the effectiveness of the proposed approach. Comparative experiments on large-scale traffic data reduced to different dimensions show that the PCA-LPP algorithm achieves higher correlation between the reduced features and sample categories compared to traditional PCA. Moreover, when handling complex and sporadic traffic accident data, the density-based DBSCAN clustering algorithm achieves a purity of 0.942 9, a Rand index of 0.946 2, and a mutual information index of 0.678 4. Comparing these results with traditional algorithms like K-means and spectral clustering, DBSCAN consistently outperforms them in various evaluation metrics. Additionally, visual analysis of the classification results indicates that the proposed model reduces the influence of noisy data. Finally, an ablation experiment confirms that the PCA-LPP algorithm with secondary dimensionality reduction achieves the highest evaluation metrics. The confusion matrix of the prediction results shows that the model achieves precision rates of 85.77%, 70.78%, and 80.65% for different risk levels, further validating its effectiveness and practicality.
A Method for Predicting the Type and Severity of Freeway Accidents Based on XGBoost
GAO Xuelin, TANG Houjun, SHEN Jiaping, XU Chengcheng, ZHANG Yujie
2023, 41(4): 55-63. doi: 10.3963/j.jssn.1674-4861.2023.04.006
Abstract(334) HTML (146) PDF(39)
Freeway accidents are frequent, and previous studies have failed to adequately reveal the effect of dynamic traffic flow on accident type and severity. This study focuses on a prediction method for types and severity of freeway accidents based on real-time traffic flow data. Traffic flow characteristics, including volume, density, and speed, are extracted from freeway gantry data. Simultaneously, temporal features and spatiotemporal non-uniformity features are considered. These data are then matched with accident data to constitute the full dataset for modeling. The model based on the extreme gradient boosting tree (XGBoost) algorithm is developed to predict the occurrence of accidents and accident types, and also to assess accident severity. Two types of accidents (i.e., rear-end collisions and other types of accidents) are considered and two levels of accident severity (i.e., injury or fatal accidents and proper-ty-damage-only accidents) are distinguished. The results indicate that: ①a higher risk of traffic accidents is associated with significant speed difference between upstream and downstream traffic, low speeds, high traffic volumes with frequent merging and diverging conditions; ②rear-end accidents are more likely to occur in situations with lower speeds, high traffic volumes with merging and diverging flows, and significant speed difference between upstream and downstream traffic; ③accidents involving rear-end collisions may result in higher severity when they occur on road segments with lower traffic volumes or occur during weekends or nighttime. The Area Under Curve (AUC) of the XGBoost-based models for accident types prediction and accident severity prediction reached 0.76 and 0.88 respectively. Compared with other commonly used algorithms such as Sequential Logistic, Gaussian Naive Bayes, Linear Support Vector Machine (SVM), Random Forest, and Neural Network, the XGBoost-based model demonstrates an average improvement of 0.08 and 0.24 in AUC values for predictions of accident types and accident severity. These results indicate that the XGBoost-based model exhibits better prediction performance. The research findings provide a reliable way for state warning of real-time traffic flow on freeway segments, which could be useful for improving driving safety.
An Analysis of Linguistic Characteristics and the Effectiveness of Safety Slogans for Preventing Drunk Driving
YUAN Yang, LI Donghe, LI Kexin, LI Peiling, NING Peishan, HU Guoqing
2023, 41(4): 64-71. doi: 10.3963/j.jssn.1674-4861.2023.04.007
Abstract(259) HTML (145) PDF(21)
To address the uncertainty regarding the effectiveness of traffic safety slogans in enhancing public awareness of traffic safety, a quantitative analysis of Chinese traffic safety slogans for preventing drunk driving and their effectiveness is conducted, aiming to analyze the linguistic characteristics and their effectiveness. Using Python 3.7 software, a total of 1 828 traffic safety slogans about drunk driving prevention, collected from the Baidu search engine between September 2019 to September 2020, are analyzed. The linguistic characteristics of these slogans were categorized and analyzed from six aspects, including person expression, emotion, rhetoric, rhyme, context, and length. The primary linguistic characteristics found in these slogans include the use of the other person expression (87.7%), negative emotion (45.4%), lack of rhetoric (49.9%), presupposed states (52.4%), and medium length (65.2%). Through an online questionnaire survey, the effectiveness of slogans with different linguistic characteristics is evaluated in terms of public attention, understanding, and acceptance. By employing the Chi-square test, the differences in linguistic characteristics of the slogans are analyzed and a generalized linear regression model is developed to identify significant linguistic factors affecting public attention, understanding, and acceptance. The results of generalized linear regression show as follow. ①Public attention to slogans is primarily influenced by the person expression and length of the slogans. Specifically, the slogans using the first- and other-person expression attract greater public attention compared to slogans using the second person expression (b =0.24, 0.49). Longer slogans (18 words or more) aroused more public attention compared to shorter slogans (less than 12 words) (b =0.26). ② Public acceptance of slogans is mainly influenced by the use of person expression. Slogans using the first-person expression receive higher public acceptance than those using the second person expression (b =0.31). ③There is no statistically significant relationship between public understanding of slogans and six linguistic characteristics of slogans. The results indicate that Chinese traffic safety slogans designed to prevent drunk driving exhibit diverse linguistic characteristics. The different linguistic characteristics affects the public attention and acceptance of the slogans. In the future, when designing traffic safety slogans, it is recommended to use the first person or other person expression and to consider increasing the length of the slogans to enhance the effectiveness of slogans in enhancing public awareness of traffic safety.
An Analysis of the Impact of Time Delay of Fusion Modes for Point Clouds from Cooperative Road Vehicle Systems on Autonomous Driving
YE Qing, ZHAO Cong, ZHU Yifan, YU Shanchuan
2023, 41(4): 72-79. doi: 10.3963/j.jssn.1674-4861.2023.04.008
Abstract(277) HTML (103) PDF(26)
The rapid development of the new generation of communication technologies provides a foundation for cooperative perception between autonomous vehicles (AVs) and road. This advancement holds the potential to significantly enhance the perception capabilities of AVs in complex scenarios. Previous studies have explored different information fusion modes for cooperative perception, but neglected to analyze the balance between perception accuracy and communication delay. Aiming at the delay characteristics of point cloud fusion in cooperative perception of AVs, a delay impact analysis framework is proposed based on simulation, concentrating on three fusion modes: pre-fusion, feature fusion, and post-fusion. Considering the time lag of cooperative perception results caused by communication delay, the Extended Kalman Filter algorithm is used to make predictive compensation for cooperative perception results with delay. The novel metrics, namely Lag Compensation Error and equivalent time delay, are proposed for comprehensive evaluation of the impact of different fusion modes on cooperative perception results. Based on perception results from various point cloud fusion modes, a model is established to fit the relationship between average perception accuracy and translation error distribution. Utilizing the distribution characteristics of translation errors, this model serves as the basis for generating simulated trajectories with perception errors and subsequently the evaluating of cooperative perception performance. Finally, leveraging the TrajNet++ pedestrian trajectory dataset, 180 000 numerical simulations are conducted across 1 200 trajectories with various point cloud fusion modes and different delay parameters. The results demonstrate that the shorter trajectory lengths and the higher target speeds amplify the impact of delay on cooperative perception accuracy. In comparison to post-fusion with a 100 ms delay as the benchmark, the equivalent or superior cooperative perception accuracy is feasible when the feature-fusion delay is below 500 ms and the front-fusion delay is below 700 ms. In complex scenarios involving sudden target appearances or high-speed targets, it is recommended to choose low-delay, low-accuracy post-fusion modes. Conversely, it is advisable to consider feature fusion or pre-fusion modes with high delay and high accuracy. This study can provide a basis for the selection of point cloud fusion modes for cooperative perception of autonomous driving.
Cooperative Positioning of Vehicle Fleets Using Road Probability Field
WANG Zhongqi, CHEN Wei, DU Luyao, LEI Zhen, LEI Ting
2023, 41(4): 80-87. doi: 10.3963/j.jssn.1674-4861.2023.04.009
Abstract(199) HTML (94) PDF(11)
Scalar field methods are widely used in the coordinated positioning of groups of unmanned automatic vehicles (UAVs) and submarines. However, there are difficulties in applying similar scalar fields such as magnetic anomaly fields and water depth fields in vehicle fleet scenarios. To address this issue, a vehicle fleet collaborative positioning method based on road probability field and vehicle motion model is proposed an open-source database is used to obtain electronic maps, and the electronic maps were buffered, rasterized, and processed with mathematical morphology to construct road probability field. At the same time, a vehicle motion model is established based on GNSS technology, using the relative positions between vehicles in the fleet as collaborative information, taking the value of the predicted position of the vehicle in the road probability field as a weight calculation criterion, and using particle filtering localization algorithms to continually update the predicted trajectory of the vehicle. The new method establishes the road probability field as a scalar field, using the road probability value corresponding to the vehicle position as an important basis for determining the vehicle position, and applying the geographic spatial information contained in the electronic map to the collaborative positioning of the vehicle fleet. Unlike traditional scalar field methods, road probability fields do not require new specialized measurements and can be generated using the massive electronic map resources already available, and vehicles do not require new sensors. A vehicle motion model is designed based on the application scenario, and the trajectory is continuously optimized using the road probability field during the dynamic process of vehicle driving, reflecting the difference with traditional fleet positioning methods that pay more attention to single time point positioning. Comparative tests were conducted on different buffer widths and vehicle numbers in real and simulation experiments. The results show that using positioning error as the criterion for determining the positioning effect, compared to the classic extended Kalman filtering method using vehicle motion models, proposed method achieves improvements of 49.6% and 49.8% in simulated and real scenarios, respectively. Compared with the fleet collaborative localization method based on empty road probability field, this method has improved by 59.5% and 50.3% in simulation and real scenarios, respectively. This study provides a new method for cooperative positioning by constructing a road probability field and utilizing a vehicle motion model. Compared to traditional methods, this method improves the accuracy and reliability of positioning and has important application prospects.
A Method for Detecting Personnel at Vessel Bridge and Evaluating Level of Activities Based on Channel State Information
ZHANG Yifan, CHEN Mengda, WANG Lu, CHEN Cong, LIU Kezhong, CHEN Mozi
2023, 41(4): 88-100. doi: 10.3963/j.jssn.1674-4861.2023.04.010
Abstract(235) HTML (115) PDF(5)
The personnel on bridge consist of regularly scheduled officers on watch and additional person for look-out, captain, and pilots in specific circumstances. The activity level of the personnel on bridge is one of the crucial indicators to assess their work status. Traditional computer vision-based personnel detection methods show reduced accuracy when confronted with challenges such as multiple obstructions on the ship's bridge, insufficient light conditions during nighttime or adverse weather conditions. To address this issue, a detection and activity evaluation method based on ordinary commercial Wi-Fi devices is proposed. Due to the dynamic multipath and strong signal noise caused by the ship's material and structural characteristics and the changing motion states, the function of Wi-Fi devices is interfered. To mitigate these challenges, a duty high-correlation data (DHCD) selection module and a multi-layer feature extraction module based on channel state information (CSI) are designed. The DHCD selection module analyses the CIS characteristics in different navigation and duty situations and compares the channel variations when 0-5 people on the ship's bridge. The fuzzy C-means clustering algorithm is employed to extract the most responsive channel information to the behavior of personnel on bridge while eliminating the information sensitive to signal noise. The multi-layer feature extraction module calculates various features, including amplitude, phase dispersion, multi-link fusion dispersion, and variation index for denoised CSI data as the foundation for activity evaluation. The activity evaluation module is designed primarily based on the requirements for the on-duty personnel on bridge. The Support Vector Machine algorithm is utilized to determine the number of bridge personnel, while the Criteria Importance through the Intercriteria Correlation method is used to obtain the weight for basic parameters. Combining the headcount information and weight information, the activity level of bridge personnel is evaluated. The results indicate that the multi-layer features using the DHCD selection module and multi-layer module processing improve the accuracy of detecting the number of bridge personnel to 89.6%, representing a 7.1% increase compared to directly using raw data. In low-light conditions such as nighttime, rainy, or foggy weather, the accuracy of computer vision-based methods decreases from 96.2% under normal light to 60.3%. In contrast, the detection accuracy of proposed method remains stable. Therefore, the CSI-based detection and activity evaluation method enriches the detection algorithm for bridge personnel and can effectively identify whether the personnel meet the basic requirements for safe duty.
A Study on the Shy Away Effects of Left-turn Vehicles at Urban Intersections
QIN Xiaolin, DU Zhigang, CHEN Ying, LIU Qingyi
2023, 41(4): 101-110. doi: 10.3963/j.jssn.1674-4861.2023.04.011
Abstract(194) HTML (85) PDF(11)
When a traffic barrier extends into the boundary of inner lane constructions at an urban road intersection, it may result in lateral deviations of vehicles and traffic risk. To examine how traffic barriers at urban planar intersections affect the avoidance behavior of left-turning vehicles, unmanned aerial vehicles (UAVs) are employed to record vehicle videos at three planar intersections. Subsequently, information of vehicle trajectories, speeds, and accelerations is extracted. The distributions of vehicle offsets and velocities in various lanes at intersection exits are analyzed. Additionally, the avoidance behavior of left-turning vehicles is examined. Results show that: ①vehicles not in the middle lane exhibit a greater tendency to shift toward the middle lane, while those in the middle lane maintain a more stable trajectory; ②after traveling 20 meters, drivers achieve a stable level of vehicle control, resulting in smooth trajectories and maintaining safe lateral distance with traffic barriers; ③over 85% of vehicles in the left lane keep away from the traffic barrier, exhibiting an average lateral offset of 0.278 meters, and about 60% of vehicles in the right lane away from the right side, showing an average offset distance of 0.116 meters; ④significant disparities exist in the velocity distribution of left-turning vehicles at various lanes of the interaction exit: the peak velocity, mean lateral acceleration, and mean longitudinal acceleration of left-turning vehicles in the left and right lanes are lower than those observed in the middle lane. Based on these findings, recommendations are made to improve the operation levels of urban road intersections: ①enlarge the width of the center divider and improve roadside clearance to align with the desired right-of-way width of the left lane; ②increase the lateral distance between rigid facilities and drivers; ③optimize rigid facilities in the exits about the flexibility to mitigate their psychological impact and lessen the driving burden; ④enhance the presence of roadway guide lines and reflective facilities to ensure the uninterrupted and uniform implementation of guiding elements and to boost drivers' perception of direction and speed, consequently to reduce the safety risks associated with over- or under-shy away effects.
A Method for Predicting Long-term 4D Trajectory of Airplanes Based on Informer
FENG Xia, SUN Qiqi, ZUO Haichao
2023, 41(4): 111-121. doi: 10.3963/j.jssn.1674-4861.2023.04.012
Abstract(211) HTML (84) PDF(15)
Prediction of long-term 4D trajectory is an important foundation for trajectory-based operation, which is significant for improving safety of air transportation system and optimizing airspace. The existing methods for predicting long-term 4D trajectory do not fully consider implicit association among trajectory data with a long sequence. To address this problem, a long-term 4D trajectory prediction model based on Informer model with the self-attention mechanism is developed. To extract the global feature from trajectory data, enhance data independence and the capability to learn the feature of time series, a global timestamp module is added into the data embedding layer. Moreover, the layered timestamps, such as trajectory point sequences, are utilized to overcome the inherent time scale limitation of the Informer model. To better capture the implicit correlations between non-adjacent temporal sequence points, a self-attentive mechanism is employed to extract the features of trajectory data, and a probabilistic sparse method is applied to reduce the computational complexity of the self-attentive mechanism to O(LlogL). Additionally, a distillation mechanism is incorporated into the encoder to reduce the computational dimensions and the number of network parameters. To avoid the error accumulation arising from traditional step-by-step prediction models and improve the accuracy of trajectory prediction, a fully connected layer is applied to adjust the dimensions of the predicted data, achieving one-step generative output. After three-time spline interpolation, the pre-processed historical 4D trajectory data are inputted to the trajectory prediction model along with the data presenting the feature of time-series. Through iterative training of the model, the trajectory prediction results are generated and output. Study results show that, the Informer-based model outperforms the LSTnet method when predicting the trajectory of 4D features simultaneously. The root mean square error and Euclidean distance error is 0.2185 and 15.980 km, respectively, which is a reduction of 1.48% and 2.44% compared to that of the LSTnet network. In addition, when predicting the trajectory features separately, the Euclidean distance error of the Informer-based model is 13.248 km, with a reduction of 3.11% compared to the LSTnet network and a reduction of 34.99% compared to the traditional LSTM network.
A Method for Measuring Visibility under Foggy Weather for Expressways Based on Siamese Network
TANG Wei, FANG Jianan, ZHANG Long, YANG Xiaodong, LI Guoqiang
2023, 41(4): 122-131. doi: 10.3963/j.jssn.1674-4861.2023.04.013
Abstract(208) HTML (86) PDF(12)
Accurately identifying highway visibility levels in foggy weather from surveillance video is important for intelligent highway supervision. Aiming at the problems of low accuracy, slow rate, and weak generalization of the current visibility recognition methods on highways, a visibility recognition method based on Siamese network is proposed, focusing on the optimization of the image feature extraction module and the fog visibility level recognition. The image feature extraction module adopts the improved VGG16 network as the backbone network. In order to enhance the ability of the network to extract important features from the global information of the image, a convolution block attention module is added to the five blocks of the VGG16 network to emphasize the effective features and suppress the useless features. To improve the generalization ability and training rate of the network, a filter response normalization layer is added after the convolutional layer of the network to remove the differences between the dimensional data. In order to solve the redundancy problem of network weight parameters and prevent overfitting, global average pooling is used to compress the output feature map directly into a 1×1 vector instead of the first two fully connected layers in the VGG16 network. A Siamese network is adopted as the main framework of the fog visibility level recognition module, and the effective features extracted by the image feature extraction module are propagated forward. The distance measurement method is utilized in the contrastive loss function to assess the similarity between input image pairs in a high-dimensional space for fog visibility level recognition. Experiments are conducted based on a dataset of actual foggy images collected from August 2022 to January 2023 on highways in Shaanxi Province. The experimental results show that the recognition accuracy of the proposed method is 90.3%, which is an improvement of 20.4%, 18.9%, and 18.0% compared to the single networks AlexNet, ResNet50, and VGG16, respectively. It is also an improvement of 16.2%, 11.0%, and 5.4% compared to the Siamese networks Simaese-AlexNet, Simaese-ResNet50, and Simaese-VGG16, respectively, which constructed based on single networks as benchmark models. In conclusion, this method exhibits a high accuracy, which contributes to enhancing the intelligent supervision capabilities for foggy weather conditions on highways.
A Study on Impact of Curb Parking on Road Capacity Using Cellular Automata
SHEN Jinxing, JIANG Wenfeng, CAO Huimin, MA Changxi
2023, 41(4): 132-142. doi: 10.3963/j.jssn.1674-4861.2023.04.014
Abstract(194) HTML (88) PDF(10)
The installation of curb parking is a crucial solution to address the imbalance between supply and demand of parking. However, unreasonable installation of curb parking spaces may fail to solve the problems of parking, and may reduce road capacity, resulting in traffic congestion and a series of related traffic issues. To investigate the impact of curb parking supply, demand, and convenience on road section capacity at a microlevel, an improved one-way two-lane cellular automata model is developed. The model proposed in this study includes three components: a car-following model, a lane-changing model, and a curb-parking model. In the car-following model, the difference between regular driving and cruising parking processes is studied by setting different slowing probabilities for the parking and non-parking road sections. In this model, the impact of curb parking convenience on road capacity is resolved by setting the duration of parking maneuvers. Based on the field survey data, the operational rules for regular driving and cruising-parking vehicles are calibrated separately, and the reliability of the model is verified. The comparison results with the existing models show that the proposed model exhibits superior fitting performance and the average absolute error index can be reduced by 53.74% to 75.71%. It indicates that the proposed model can provide a more accurate reflection of the impact of curb parking process on traffic flow of road section. Besides, as traffic flow density increases, the average absolute error index of the proposed model can be reduced by 16.39% to 52.85%. The results show that the model can avoid overestimating the negative impact of curb parking on road section capacity, making it a valuable addition to current studies. Road section capacity under the scenarios of changing parking demand, parking convenience, and supply capacity are studied in open boundary conditions. The results show that the road section capacity fluctuates significantly under all three scenarios. The most significant impact is observed from changes in parking demand. Specifically, when parking demand increases from 10% to 30% of road traffic flow, there is a 44.12% decrease in capacity. When the parking convenience decreases and the duration of parking maneuvers increases from 5 s to 15 s, the capacity decreases by 24.44%. On a 500 m road segment, when the parking supply capacity increases from 20 parking spaces to 50, a 39.39% decrease in capacity is observed.
A Short-term Traffic Flow Forecasting Model Considering Dynamic Spatio-temporal Relationship
ZHAO Zhenxing, ZENG Wei, TANG Chenjia
2023, 41(4): 143-153. doi: 10.3963/j.jssn.1674-4861.2023.04.015
Abstract(199) HTML (98) PDF(18)
A traffic flow prediction model based on dynamic spatio-temporal graph convolutional network (DySTGCN) is developed, to effectively extract the spatio-temporal features of traffic flows and improve the accuracy of traffic flow prediction. DySTGCN models not only Spatio-temporal information of traffic flow but also the influence of temporal information on spatial information. Meanwhile, a spatial structure based on temporal information, a time-varying spatial topology graph (TSG), is proposed and a deep neural network structure to efficiently calculate the TSG is designed. The structure extracts correlation features of traffic flow at different nodes and can reduce the noise through encoding and decoding. TSG reflects the real-time spatial feature of the traffic network, a stable spatial graph (SG) based on the spatial position of nodes reflects the stable spatial feature. The TSG and SG jointly guide the traffic flow prediction and depict the spatio-temporal feature more accurately to improve the prediction precision. To test the prediction effect of the model, experiments are carried out on two authoritative public data sets. The results show that TSG learned by DySTGCN can accurately reflect the correlation between traffic flows at different nodes and MAE, RMSE and WMAPE of DySTGCN are 13.40%, 10.98%, and 16.72% lower than other spatio-temporal graph convolutional network models such as STGCN, ASTGCN, verifying that dynamic spatial relation plays an important role in short-term traffic flow prediction fully. Besides, DySTGCN can extract periodic features of traffic flow and achieve continuous and uninterrupted prediction of traffic flow.
A Site Selection Model for Electric Vehicle Charging Stations Considering Queuing Time and Charging Cost
2023, 41(4): 154-162. doi: 10.3963/j.jssn.1674-4861.2023.04.016
Abstract(213) HTML (112) PDF(12)
A reasonable layout of electric vehicle charging stations plays a crucial role in reducing range anxiety, improving travel comfort, and promoting the adoption of electric vehicles. To overcome the limitations of existing studies that overlooks the consideration of queuing time and charging cost, an improved site selection model for charging stations is established with the objectives of minimizing range anxiety and charging costs. This model explicitly considers queueing and detouring behaviors in charging. The characteristics of charging behavior of electric vehicles are analyzed, and a distance constraint for allowable path deviations is introduced to establish a limit on detour distances in charging paths, thereby reducing the scale of the set of deviation paths in the road network. The characteristics of the charging station queueing system are analyzed, and an analytical expression for the average queueing time of the system is derived with constraints such as acceptable queueing time threshold and budget cost. Considering the patterns of range anxiety and the stepped electricity pricing, a site selection model for charging stations is proposed to minimize range anxiety and charging costs, and the Lingo software is used to solve the model. A case study is conducted on a partial road network in the city of Xi'an. The results show that based on the proposed model, a total queue time and a total charging cost are 5.84 h and 1 440 Yuan, respectively. Compared to the model without considering queue time and charging costs, the system queue time and the total charging cost are decreased by 1.19 h and 240 Yuan, respectively. An Analysis of the charging station budget cost B shows that when B ≤ 500 million Yuan, the total range anxiety and charging costs decrease as B increases. However, when B > 500 million Yuan, further increase in B does not result in further reduction of total range anxiety and charging costs. Under the conditions of budget costs B = 300 million, 400 million, and 500 million Yuan, respectively, the impact of path deviation distance η on the optimization objective is analyzed. As the path deviation distance η increases from 0 km to 4 km, the total range anxiety and charging costs show a decreasing trend.
A Combined Weighting-improved TOPSIS Method for Evaluating Integration of Urban Greenway-Waterfront Road-municipal Non-motorized Transport Network
FENG Zhimei, GUO Mingyang, HE Yulong, PENG Hao
2023, 41(4): 163-172. doi: 10.3963/j.jssn.1674-4861.2023.04.017
Abstract(214) HTML (96) PDF(11)
In order to solve the problem of mutual independence and poor connection between urban greenways, waterfront roads, and municipal non-motorized transport network. In a view of the lack of existing studies on the evaluation of the level of tri-networks integration, an evaluation method based on the combination of weighting-improved TOPSIS model is investigated. The traditional TOPSIS method uses ideal solutions to calculate closeness without considering outliers and actual conditions. Therefore, an outlier detection method based on Gaussian distribution is used to deal with extreme outliers, and an evaluation model is established to comprehensively assess the level of tri-network integration. Previous methods of evaluating road network are usually conducted on a single object without considering the integration of multiple objects. Therefore, considering factors of the three networks, travel characteristics of non-motorized transport network and residents' travel convenience, an evaluation system and 13 related indicators, including network connectivity, accessibility and other features, are developed from field surveys. In order to avoid the bias generated by a single weighting, this paper establishes an optimization model of weight combination so that the subjective and objective weights determined by Analytic Hierarchy Process (AHP) and entropy weighting method deviate from the combination of weights to the smallest extent. This study takes the non-motorized transport network in Chaoyang District as a case study for model verification. According to location and function, the network is segmented into 21 greenways, and the ranking for three-network integration of the greenways are obtained. Compared with previous evaluation methods, the results show that the standard deviation of closeness from the improved model is 0.278, which yields better distinction for evaluation of network integration. Besides, the proposed model can accurately identify the main factors that affect the integration of the three networks, and make optimization based on the weight of each indicator. Thus, the proposed evaluation method can be used to a reference to optimize and promote the integration of greenways, waterfront roads and municipal non-motorized transport.
A Method for Evaluating Recovery Strategies for Cascade Failures of Metro Networks
CHENG Jing, LU Qun, WU Tongzheng, WANG Yuanqing
2023, 41(4): 173-184. doi: 10.3963/j.jssn.1674-4861.2023.04.018
Abstract(228) HTML (87) PDF(20)
The effectiveness of recovery strategies plays a vital role in emergency response after cascading failures take place in a metro network, which is closely related to its operation safety. To address the cascading failures in metro networks, a method for evaluating the efficiency of recovery strategies is proposed from a system resilience perspective. A function for allocating recovery nodes is established based on the characteristics of the distribution of passenger flows at metro stations. And a model of network cascade failure with recovery strategy is developed by integrating the recovery strategy into the cascade failure process. Then, network efficiency and connectivity are used to characterize system functionality, and a system functionality curve is introduced to quantify system resilience. The effectiveness of three recovery strategies, including random recovery, importance priority recovery, and degree priority recovery, are evaluated through Python simulations which are carried out based on the metro network in the city of Xi'an. The results indicate that increasing the node recovery ratio enhances the efficacy of recovery strategies in a singular strategy effectiveness assessment. This enhancement manifests as a reduction in system damage during the resistance and recovery phases, accompanied by an accelerated recovery rate. By comparing different strategies, the strategy of importance priority recovery outperforms the degree priority recovery and the random recovery. Two resilience indicators of the importance priority recovery are 11.9% and 3.4% greater than degree priority recovery, respectively; and 7.6% and 1.2% greater than random recovery, respectively. Compared to traditional models, the proposed model exhibits better goodness of fit for the speed of propagation failure, change of system performance, and process of actual traffic cascading failure. It suggests that under the influence of cascading failures in metro networks, better recovery results can be achieved by adopting an importance priority recovery strategy and increasing node recovery proportion. The simulation results accurately represent the impact of depict disturbance on system performance, aiding decision-making for preventing and recovering from cascading failures in metro networks.