交通信息与安全

Development of Crash Prediction Models Involving Heavy-duty Trucks over Long Downhill Segments Considering Multi-mode Failure Probability

YIN Yanna, WEN Huiying

2022, 40(3): 1-9. doi: 10.3963/j.jssn.1674-4861.2022.03.001

Abstract(2240) HTML (1506) PDF(106)

Abstract:
A crash prediction model is developed, in order to explore the relationship between multi-mode failure probability and heavy-duty truck crashes over long downhill road sections. A model for multi-mode failure probability prediction is developed to study the probability of different types of failures associated with heavy-duty trucks, such as skidding, rollover, insufficient sight distance, and braking failure, on the long downhill sections. The single-mode failure probability is simulated using a Monte Carlo method and the multi-mode failure probability of the system is studied by a wide bound method. Three crash prediction models including a Poisson model, a random-effect Poisson model, and a random-parameter Poisson model are developed, considering multi-mode failure probability as one of the explanatory variables together with other impact factors. The models are used to link the multi-mode failure probability with the crashes of heavy-duty trucks. The optimal crash prediction model is selected through the goodness-of-fit for accurately modeling the relationship between crashes of the trucks and their multi-mode failure probability. The method is verified by a 10-year data of heavy-duty truck crash and road design of 71 long downhill sections in the Washington State, the United States. The results show that there is little difference in the goodness of fit between the random-effect Poisson model and random-parameter Poisson model, and both of them are better than the Poisson model. It is found that radius of the horizontal curves, grades and superelevation rates are not significant in leading to the crashes, when compared with the multi-mode failure probability. Study results show that, the elasticity of multi-mode failure probability (0.239) is much greater than that of the radius of horizontal curve and superelevation (0.097 and 0.002) respectively; heavy-duty truck crashes and multi-mode failure probability are approximately linearly correlated, and the intercept of the model is found to be other than "0". The above results indicate that the multi-mode failure probability can be used for road safety analysis, but it is not equivalent to the crash probability, which may be used as a basis for improving road design.

An Analysis of Influential Factors of Crashes at Tunnels and Open Sections of Mountainous Freeways

ZHANG Xuan, TANG Jinjun, HUANG Helai, CHANG Fangrong, WANG Jie, YUAN Shuanglin

2022, 40(3): 10-18. doi: 10.3963/j.jssn.1674-4861.2022.03.002

Abstract(1079) HTML (367) PDF(84)

Abstract:
Freeway tunnels tend to have higher accident rates, due to their special engineering structure and complex traffic environments, compared to regular segments. In order to study the differences in mechanisms and factors influencing severity of crashes in tunnels and regular open sections on freeways, a total of 1 537 crashes taking place on Shaohuai Freeway from 2011 to 2016 are collected for the analysis. A binary Logit model considering the heterogeneity is used to explain the impacts of various risk factors on the likelihood of the locations of traffic crashes and to investigate the factors influencing severity of crashes taking place at tunnel and open sections. Statistical analysis results show that crashes associated with drowsy driving and unsafe following distance are more likely to occur in the tunnel sections, and the crash probability is 2.373 and 2.482 times higher than that in the open sections, respectively. In the tunnel sections, downhill (slope more than 2%), summer, and speeding are positively correlated to the likelihood of injury crashes, and the probability that crashes take place at downhill (slope more than 2%) sections is 3.397 higher than uphill (slope more than 2%) sections in resulting in serious accidents. It is also found that the probability of serious crashes taking place in the summer is 3.951 higher than that in the autumn; and the probability of the speeding behaviors is 4.242 higher than other inappropriate behavior. In the open sections, speeding and fatigue driving are likely to be associated with injury crashes, and the probability that speeding results in an injury crash is 2.713 higher than other inappropriate behavior. It is also found that the probability that the fatigue driving leading to an injury crash is 4.802 higher than other inappropriate behavior. The above results show that the factors influencing the crash propensity and severity at the two types of sections are different. The conclusions of this paper can be used to formulate road safety improvement plans over tunnel and open section segments of freeways.

A Method for Quantitatively Analyzing Risks Associated with the Operation of Urban Buses Considering Chained Conflicts

LI Xiying, LIANG Jingru, HAO Tenglong

2022, 40(3): 19-29. doi: 10.3963/j.jssn.1674-4861.2022.03.003

Abstract(817) HTML (334) PDF(63)

Abstract:
A quantitative method for analyzing risks associated with the operation of buses in mixed traffic environment is studied by extracting data about traffic conflict and identifying a set of chain-conflicts. Regarding data collection, aerial video data are adopted based on which features of objects are extracted using YOLOv4 network. In this way, the trajectories with accurate attributes for buses and other related vehicles can be obtained. Regarding the identification of vehicle conflicts, the relative locations between pairs of vehicles which are likely to collide laterally in different lanes are set up as constraints. Based on the classic car-following model, the dynamic relationship of vehicle pairs in adjacent lanes is studied and added. With this, the classic time-to-collision(TTC)model is extended to a two-dimensional TTC model, which can identify lateral conflicts as well. Next, according to the Stimulus-Re-sponse Theory, the temporal and spatial scope caused by each conflicting vehicle who continuously disturb regional traffic is calibrated to study interrelationships between conflicted vehicles, and a time-series conflict tree model is established. With this, chained conflicts can be identified and the causal relationship between continuous risks can be traced using the conflict tree model. The risks of urban buses under different traffic settings are quantified from the following three aspects: ①the frequency of the conflicts is analyzed based on the two-dimensional TTC model; ②on this basis, the severity of conflicts is analyzed combined with cumulative frequency method; ③the probability and scope of conflicts are analyzed through ratio of chain conflicts and the length of conflict tree. Aerial video data of Guangzhou Bridge Road are collected for a case study. The results show that urban buses in frequently congested sections have high conflict risks, which reveal have high rate of severity and regional aggregation. The conflict frequency of buses in congested traffic flow exceeds 9 times per vehicle per minute. The average rate of serious conflicts of buses is 33.39%, which is much higher than the corresponding rate of regular passenger vehicles(i.e., 16.61%). The rate of regional chain-conflicts caused by buses is 30.75%, which is twice than that of cars(i.e., 14.67%)

A Method for Assessing the Risks of Freeway Segments with Combined Horizontal and Vertical Curves

HU Liwei, ZHANG Chengjie, ZHAO Xueting, LIU Fan, LYU Yifan, XUE Yu

2022, 40(3): 30-41. doi: 10.3963/j.jssn.1674-4861.2022.03.004

Abstract(821) HTML (330) PDF(155)

Abstract:
Freeway segments with combined horizontal and vertical curves might not satisfy the safety specifications of Highway Alignment Design(hereafter referred as HAD)even though each of them does so. To assess the risks of such road sections and improve road safety, a safety assessment method is proposed by using the extendable cloud theory and the ideal point method(IPM). Firstly, a road safety evaluation system is developed, which includes 15 indicators from the following five perspectives: drivers, roads, traffic, environment, and others, and each indicator is coded as one of five levels. Secondly, subjective and objective weights of the indicators are determined by the analytic hierarchy process(AHP)and entropy weight method(EWM), respectively, which are combined into one using the ideal point method.Thirdly, the risk levels of each indicator are classified based on the specifications, considering the fuzz boundaries of the qualitative indicators, and the qualitative indicators are quantified based on the principle of equal ratio. Finally, the membership evaluation matrix is developed, a comprehensive assessment vector is calculated, and the level of risk of road section is determined by the maximum membership principle. To demonstrate the proposed method, three cases from the Yunnan Province are used, and the results show that the proposed method not only provides compliant outcomes with the traditional fuzzy comprehensive assessment method but also offers more information. Specifically, the expected value of fuzzy grade eigenvalue for comprehensive assessment, i.e. E_xr, reflects the safety level of the road sections; Confidence Factor θ reveals the reliability level of the result. In the studied cases, E_xr of Section Y is higher than that of Section C, showing that Section Y is safer than Section C; Confidence Factor θ are all under 0.05, showing that the results are reliable. These results reveal the potential of the proposed method for road safety assessment.

A Comparative Analysis of Heterogeneous Effects of Various Factors on Accident Severity at Sharp Curve Sections of Mountainous Highway

ZHAO Huaxiang, DU feixiang, FU Kaihua, SU Yu, YANG Wenchen

2022, 40(3): 42-50. doi: 10.3963/j.jssn.1674-4861.2022.03.005

Abstract(1072) HTML (414) PDF(52)

Abstract:
To identify contributing factors and their heterogeneity effects onto accident severity at sharp curve sections of mountainous highway, fifteen potential factors are selected from the following areas including driver, vehicle, road, and environment conditions based on the data from 1 067 accidents on a two-lane highway in mountain areas. Then, a binary Logit model, and a random parameters binary Logit model are used to analyze severity of three typical types of accidents including rear-end, head-on, and side collision. Results show that there are significant differences in the effect of impact factors on crash severity of three types of accidents at sharp curve sections as follows: ①For rear-end collisions, the significant variables of crash severity are motorcycle, night, cornering, age of drivers, and different seasons. Motorcycle and winter are heterogeneous influence factors obeying a normal distribution with a mean value of 2.716 and -1.495, a variance of 1.564 and 2.116. The probability of resulting in a casualty accident is 95.72% and 23.58%, respectively. ②For head-on collisions, the significant variables of crash severity are truck, motorcycle, overtaking of drivers, curve corners, and curve lengths in turn. The probability of casualty accidents with the truck increases by 108.8%. The motorcycle and curve lengths are heterogeneous influencing factors obeying a normal distribution, with a mean value of 6.941 and -0.004, a variance of 9.901 and 0.003. Consequently, the probability of casualty accident is 76.11% and 9.18%, respectively. ③For side collisions, the significant variables of crash severity are motorcycle, age of drivers, and corner with entrance in turn. The motorcycle and corner with entrance are heterogeneous influencing factors obeying a normal distribution, with a mean value of 5.211 and -1.408, and a variance of 5.111 and 2.146. Consequently, the probability of casualty accident is 88.87% and 25.47%, respectively. ④Compared with the traditional binomial Logit models, the accuracy of the random parameter binary Logit models for predicting crash severity of the rear-end, head-on, and side collision are increased by 2.85%, 4.15%, and 6.76%, respectively. With the proposed model, the heterogeneous effects of several factors can be quantitatively captured, and therefore, it can be used for improved severity analysis of road accidents.

A Safety-oriented Optimization Model for Train Skip-stop Strategy of Oversaturated Metro Lines

TAO Lefeng, SHI Jungang, YANG Jing, YANG Xiaoguang

2022, 40(3): 51-59. doi: 10.3963/j.jssn.1674-4861.2022.03.006

Abstract(922) HTML (392) PDF(31)

Abstract:
In order to alleviate the extreme congestion of oversaturated passenger flow of metro lines during peak hours, an optimization problem of skip-stop strategy for metro trains is studied from the respective of safety, which aims to minimize both risk of passenger congestion and their waiting time. Due to varying passenger demands over time, the number of waiting passengers under the train skip-stop strategy is estimated at each station by considering multiple constraints, including skip-stop operation, train tracking, and dynamic loading of passengers, and a specific evaluation function is formulated to measure the risk of passenger congestion. Based on an optimization method for the traditional train skip-stop strategy only considering passenger waiting time, a safety-oriented optimization model is proposed by integrating risk of passenger congestion into its objective function. Due to nonlinear characteristics of the proposed model, a variable neighborhood search algorithm (VNS) is designed to improve computation efficiency, where three types of novel neighborhood solutions are presented, and a penalty function is set for constraint violations. Taking Beijing Batong metro line as a case study, the proposed optimization model for train skip-stop strategy is tested for the downstream direction with 42 operating trains during morning peak hours and a part of off-peak hours (from 07:00 to 10:40 am). The experiment results show that the proposed algorithm can find high-quality train skip-stop schemes within 5 min, which can significantly relieve passenger congestion and improve service quality. Compared with the scenario where trains stop at all stations, the maximum number of waiting passengers with the train skip-stop strategy decreases from 5 299 to 2 495 over all the stations, and the risk of passenger congestion is reduced by 98.7%. At the same time, the average passenger waiting time decreases from 9.49 min to 9.15 min, reduced by 3.6%.

A Twin Delayed Deep Deterministic Policy Gradient Method for Collision Avoidance of Autonomous Ships

LIU Zhao, ZHOU Zhuangzhuang, ZHANG Mingyang, LIU Jingxian

2022, 40(3): 60-74. doi: 10.3963/j.jssn.1674-4861.2022.03.007

Abstract(1106) HTML (411) PDF(76)

Abstract:
In order to meet the requirements of developingautonomous navigation of intelligent ships and solve the problems of low learning efficiency, weak generalization ability and poor robustness ofdecision-making methods for collision avoidance based on reinforcement learning, an autonomous collision avoidance method based on Twin Delayed Deep Deterministic Policy Gradient(TD3)algorithmis proposed based on the high-dimensional characteristics of the information processed in the process of collision avoidanceand continuity nature of ship maneuvers, also considering the rationality and real-time progress of decision-making. The collision risk of a given ship is calculated by considering geographical location of the ship and the other ships nearby. The state space of intelligent collision avoidance model for autonomous ships is developed from the perspective of the global point of view. The continuous decision-making and action space of the ship is designed according to the maneuvering characteristics of encountered ships. An intelligent collision avoidance model is developed considering factors such as orientation of target ship, course keeping, collision risk, the COLREGs and good seamanship. Based on the TD3 algorithm, the ship autonomous collision avoidance network model is designed based on the state space structure, combining Long Short Term Memory(LSTM)networks and 1D-convolutional networks, and a network model is designed by using Actor-Critic structure.The network training is stabilized by means of clipped double q-learning, target strategy smoothing, and delayed policy updates.The developed collision avoidance model is trained and updated with random scenarios by usingframe skipping, dynamic increase of batch size, and iterative update times.In order to solve the problem of weak generalization ability of the model, a training process of random shipencounter scenariosbased on TD3 is proposed to achievemulti-scenario migration for theapplications of the model. A simulationis carried out to verify the model, then compared with the modified Artificial Potential Field(APF)method. The results show that the proposed method has high learning efficiency, fast and stable convergence. The trained model is applicable for the ships to passa safe distance in both two-ship and multi-ship encounter scenarios. In a complex encounter scenario, the success rate of collision avoidance reaches 99.233% when avoiding 2~4 target ships, 97.600% when 5~7 target ships, 94.166% when 8~10 target ships, is higher than that of the modified APF algorithm. The proposed method can effectively respond to the uncoordinated actions of incoming ships, with real-time performance, as well as safe and reasonable decision-making.The course change is fast, stable, and the vibration is small, also the path for avoiding collisions is smooth, which has better performance than the modified APF method.

A Method for Optimizing Geometric Design and Signal Timing for Contraflow Left-turn Lanes with Double-exits

SONG Lang, WANG Jian, YANG Binyu, ZHU Yong

2022, 40(3): 75-85. doi: 10.3963/j.jssn.1674-4861.2022.03.008

Abstract(1133) HTML (457) PDF(44)

Abstract:
To solve the problem of ineffective match between the lengths of contraflow left-turn lanes(i.e., dynamic borrowed exit lines through pre-signal control)with single-exit(i.e., one pre-signal exit only)and corresponding traffic demand, a method for optimizing the geometric design and signal timing for contraflow left-turn lanes with double-exits(i.e., two pre-signal exits)is proposed after analyzing the design of contraflow left-turn lanes with single-exit. Based on the observed maneuvers of vehicles in the contraflow left-turn lanes and queuing behaviors of vehicles in left-turn lanes with single-exit and double-exits, their capacity and delay estimation models are developed, respectively. The conventional design methods for signal timing of single-exit and double-exits are integrated into a unified optimization model by introducing a dummy variable to indicate whether each pre-signal exit is enabled. Considering the constraints of coordination between main and pre-signals, saturation and traffic wave transfer, the objective of the model is to minimize the delay per vehicle. Through this model, the basis for designing the length of contraflow left-turn lanes can be obtained. Finally, a case study is carried out, the results indicate that: ①the improvement of intersection capacity is largest when the length of contraflow left-turn lanes is 80 m.②When the capacity of the roads meets traffic demand, the shorter the length of contraflow left-turn lanes, the more significant the reduction of traffic delay at the intersection.③If longer contraflow left-turn lanes are adopted to maintain road capacity, the benefit of contraflow left-turn lanes with double-exits are better than lanes with single-exit.④Considering traffic delay, capacity, and other factors, the length of contraflow left-turn lanes with single-exit is suggested to be 40 to 60 m, while lanes with double-exits should be set around 80 m. ⑤The contraflow left-turn lanes with double-exits are able to control pre-signal exits as requested, which is flexible and suitable for various traffic scenarios.

A Signal Control Method for Bus Priority Considering the Delay of Non-priority Vehicles in a Connected-vehicle Environment

TAN Baihong, QIU Zhijun, ZHANG Yi, HE Shuxian

2022, 40(3): 86-95. doi: 10.3963/j.jssn.1674-4861.2022.03.009

Abstract(879) HTML (375) PDF(33)

Abstract:
A connected-vehicle(CV)environment facilitates the collection of traffic data and the interactions among road users; therefore, it can contribute to more accurate evaluation of travel demand and traffic control. This paper investigates a signal control method at a single intersection for bus priority based on the weights for and delay distributions of bus and the other, non-priority vehicles. First, the arrival rates are calculated based on the trajectory data of connected buses and non-priority vehicles in the intersection, and the corresponding probability function of each phase is developed according to the distribution pattern of vehicle arrivals, based on which the probability of arrival rate is calculated using a maximum likelihood estimation model. Second, the wave speed of queuing, discharge, and departure are calculated respectively, using a traffic flow shock wave model. Third, the model specification for bus delay is carried out using the time-distance diagram of the shock-wave velocity, based on the fact that the number of buses in the traffic flow is less than regular vehicles while their weights are higher. Meanwhile, the model specification for non-priority vehicles is carried out using the Fixed Number Theory based on vehicles' arrival rate, considering the number of non-priority vehicles in traffic flow is larger while the weight of non-priority vehicle is lower, and most of them are not connected. The weighted delay of the intersection is calculated based on the number of passengers in vehicles. Finally, a mixed integer linear programming model is established to minimize the weighted delay, whose solution will then be used for optimizing signal control systems. To check the validity of the proposed method, a case study of the intersection of North Checheng Road and Dongfeng Avenue in the City of Wuhan is carried out. Traffic flow data of buses and non-priority vehicles at the intersection in different periods are collected, and an simulation experiment is accomplished based on Simulation of Urban Mobility(SUMO)Package. Experimental results show that the average delays for buses reduce by 25.63%, 25.25%, and 18.32%, under the scenario of low, medium, and high traffic flow rate, respectively. Compared with those before optimization, the average delays for non-priority vehicles in a single cycle under the same scenarios reduce by 8.80%, 4.68%, and 1.99%, respectively; and the average weighted delay in a single cycle under the same scenarios reduce by 20.98%, 9.39%, and 12.70%, respectively. The above results show that the proposed method is suitable and performs better in different traffic settings.

An Optimal Scheduling Method of AGVs at Automated Container Terminal Considering Conflict Avoidance

DING Yi, YUAN Hao, FANG Huaijin, TIAN Yu

2022, 40(3): 96-107. doi: 10.3963/j.jssn.1674-4861.2022.03.010

Abstract(1161) HTML (432) PDF(100)

Abstract:
Scheduling of automated guided vehicles (AGVs) is crucial for improving the operational efficiency of automated container terminals. In this paper, a two-stage optimization model is proposed for task allocation and path planning of AGVs with the consideration of the following factors: the remaining power supply of the AGV, multiple loads, and the characteristics of automated port layout. During the first stage of the optimization, a task allocation model is used to minimize the total operation time of AGVs, while a path planning model is used to optimize the operation paths of AGVs in the second stage, which will prevent the conflicts between AGVs. A new simulated annealing algorithm is developed to solve the proposed task allocation model. To guarantee an acceptable running time of the algorithm and the quality of the solution, the time cost of the task and the number of AGVs are prioritized in the process of improving the solution algorithm. A path planning algorithm based on time-space network is designed to solve the path planning problem, which discretizes the work area into a grid network and adds revised time information to develop an updated time-space network. It searches for the shortest path based on the network, while detecting conflicts and adjusting routes to avoid collisions and congestion of paths. Under the congestion scenarios, where there is no feasible solution for path planning due to unbalanced task assignment, the cost of AGV tasks will be recalculated based on conflict avoidance and their tasks will be reassigned again. Simulation experiment and comparative analysis are carried out for the case study automated container terminal (Phase Ⅳ) of the Yangshan Port. The proposed method for scheduling of AGVs is compared with a traditional path planning and obstacle avoidance model. study results show that the total operation time is reduced by 7.31% on average. The conflicts between AGVs are totally removed. The total task delay is reduced by 2 895 s, and the network congestion is reduced by 10.79%.

A Graph Study on Turning Behaviors of Older Drivers at Unsignalized Intersections

NI Dingan, GUO Fengxiang, ZHOU Yanning

2022, 40(3): 108-117. doi: 10.3963/j.jssn.1674-4861.2022.03.011

Abstract(780) HTML (305) PDF(37)

Abstract:
Considering the vision issues and slow response of older drivers, it is therefore important to study their turning behavior. Given this fact, the turning behavior of older drivers at unsignalized intersections are studied in this paper and a graph is developed to describe the changes of their driving behaviors over the time. A virtual simulation of driving scenes (including six unsignalized intersections) with different types of conflicts are developed based on a field survey. Next, older drivers and young/middle-aged drivers who meet requirements of simulation are recruited to conduct the experiment. Data regarding vehicle operation behaviors, eye movements, and physiological and psychological condition are collected to analyze the differences of behavioral characteristics between older and young/middle-aged groups of drivers under different turning scenarios. The Graph Theory is adopted to describe the features of behaviors for older and young/middle-aged drivers. Results show that for older drivers, the average of turning velocity is 20.4 km/h, fixation duration is 289.47 ms, and saccade amplitude is 3.51°. For young/middle-aged drivers, their average speed is 35.79 km/h, 247.94 ms and 4.56°, respectively. Older drivers are more nervous during the turning process since their time domain indicators (SDNN and RMSSD) and frequency domain indicators (LF/HF and TP) are lower. The graph indicates that the time of older drivers' nervousness lasts longer than young/middle-aged drivers do, and older drivers have a lower ability to collecting different information. Finally, the spatiotemporal differences from the graphs indicate that there are significant differences between the turning behavior of the two driver groups in left-turning scenarios, and the older drivers' turning behavior show a lower level of stability and safety.

A Control method of Dedicated Lanes for Mixed Use of Special Vehicles and CAVs Based on Dynamic Clear Distance

ZHAO Xin, PANG Mingbao

2022, 40(3): 118-126. doi: 10.3963/j.jssn.1674-4861.2022.03.012

Abstract(879) HTML (346) PDF(28)

Abstract:
Providing road priority to special vehicles is one of important tasks of traffic management and operation authorities. However, traditional control measures for providing road priority to special vehicle are difficult to implement, and they also tend to significantly reduce road capacity for other traffic. Therefore, a control method of dedicated lane for mixed use of special vehicles and connected automated vehicles(CAVs), is proposed to solve the above problems. First, the access rules for the dedicated lanes in the order of emergency vehicles, vehicles with secondary priority and CAVs are designed. By predicting the queue length when special vehicles arrive at a downstream intersection, the state of special vehicles at the intersection is obtained, and a dynamic clear distance model is developed to meet the demand of special vehicles with different priorities. In this model, the objective function is to minimize the speed reduction of emergency vehicles, and to balance the road capacity with traffic demand of the vehicles with secondary priority. The rules for CAVs to enter and leave the dedicated lanes are designed, and a lane-changing control model is established to solve the problem that CAVs may become obstacles to other vehicles at dedicated lanes. With the above, a dedicated lane control strategy suitable for different types of vehicles with different priority is proposed. The effectiveness of the model is validated through a set of simulated experiments. Study results show that, compared with the control methods without consideration of the priority of special vehicles, the average travel time and per capita travel time are increased by 3.9% and 2.8% respectively, but the average vehicle delay of special vehicles is reduced by more than 59.6%. Compared with the intermittent bus lane control method, the average vehicle travel time and per capita travel time are reduced by 16.7% and 14.6% respectively, and the average delay of special vehicles is reduced by 13.5% and the use rate of special lanes is increased by more than 36.3%. The best outcomes can be obtained when the CAVs penetration rate is greater than 40%. The proposed method removes some of the limitations of traditional lane control strategies when providing road priority to special vehicles, and therefore, provides"new"theoretical contribution for traffic control and management.

A Method for Detecting and Differentiating Asphalt Pavement Distress Based on an Improved SegNet Algorithm

ZHANG Zhihua, DENG Yanxue, ZHANG Xinxiu

2022, 40(3): 127-135. doi: 10.3963/j.jssn.1674-4861.2022.03.013

Abstract(809) HTML (374) PDF(121)

Abstract:
The existing SegNet methods are unable to accurately differentiate asphalt pavement distress with similar characteristics, such as cracks and sealed cracks. To solve this problem, a method of detecting asphalt pavement distress based on an improved SegNet network is proposed. In order to remove the negative impact of road markings and uneven illumination onto image quality of road surface and subsequent failure detection, a multi-scale Retinex algorithm with color restoration(MSRCR)is used to reduce the impact of road marking and uneven lighting on image quality. Through enhancing the contrast, hue and brightness of the images for road surface, the accuracy of distress recognition is improved. In order to fully use of the contextual information of the image, and overcome the issue with the SegNet network of being ineffective in segmenting and identifying subtle diseases, a residual neural network(ResNet)is introduced as the encoder, and two feature maps with a same scale obtained by a convolutional layer with a 1×1 kernel and a dilated convolutional layer with different dilation rates are fused for each feature map, generated by up-sampling in the decoder. And a closed, morphological operation is used to connect discontinuous cracks. To verify the effectiveness of the improved algorithm, it is compared with the classic semantic segmentation methods(such as SegNet and BiSeNet)over the test sets. The average intersection over Union(MIoU)and F1 score are(82.4%, 98.9%), (69.4%, 94.0%)and(80.5%, 98.1%), respectively. The three methods are compared in terms of their extraction efficiency in identifying pavement diseases using the pavement images collected at several freeway sections in the Gansu Province. The misdetection rate and false detection rate of cracks of the proposed method are 2.91%, 1.94%, respectively, which are much better than those of the SegNet(10.68%, 14.56%)and BiSeNet(6.80%, 12.62%). The above results show that the proposed method can be used to extract and identify asphalt pavement cracks and sealed cracks with a higher accuracy.

A Method for Improved Air Luggage Check-in Service Based on Optimized Urban Mobile Stations

HU Xiaobing, ZHANG Xuemei, ZHOU Hang, MA Yiming

2022, 40(3): 136-145. doi: 10.3963/j.jssn.1674-4861.2022.03.014

Abstract(728) HTML (217) PDF(22)

Abstract:
To enhance the quality and competitiveness of air transport service and overcome the limitations of low service coverage, high costs, and complex site selection of traditional air terminals, this paper proposes a novel method for improved air luggage check-in service based on Urban Mobile Stations (UMS). Specifically, the proposed UMS can adapt the check-in locations to the real-time passenger positions, which is formulated as a UMS dynamic siting optimization problem over the road network. The average distance and the maximal acceptable distance from passengers to UMS are considered, incorporating the constraints on the locations of service, time-varying distribution of passengers, and the service capacity of stations. Then, a hybrid optimization algorithm satisfying the requirement of real-time computation is developed, which combines the ripple spreading algorithm (RSA) and the adaptive genetic algorithm (AGA). The RSA is used to solve the many-to-many path optimization problem of passenger and UMS stations, and the AGA is employed to optimize the UMS locations. Case studies based on the road network of Tianjin City and simulated random road networks are used for the comparison between the proposed method and the traditional method. The results show that the average distances from passengers to stations are reduced by 30.9%, the number of scenarios exceeding the maximum acceptable distance is decreased by 43.7%, and the average running time of solving the UMS optimization problem is shortened by 41.2% when using the proposed method. These facts show the advantages of the proposed UMS method, meeting the real-time passengers' demands.

An Analysis of Spatial-temporal Characteristics of Origin and Destination of Shared-bike Users

LI Fu, XU Liangjie, CHEN Guojun, ZHU Ranbo

2022, 40(3): 146-153. doi: 10.3963/j.jssn.1674-4861.2022.03.015

Abstract(766) HTML (241) PDF(48)

Abstract:
In a view of the frequent imbalance between supply and demand and uneven distribution of shared bikes over space, this paper studies the origin-destination distribution of shared-bike users and the temporal characteristics of riding demand in different areas, so as to provide theoretical support for dispatch operations of shared-bike systems. Based on riding data of users, the mean-shift algorithm is used to cluster the origin and destination points of riding, and the distribution of areas with a high riding record is obtained. Then, Spearman correlation coefficient is used to measure the similarity of temporal characteristics of riding demand. Six typical temporal characteristics of riding demand are extracted by clustering the temporal cumulative differences between the volumes of rented and returned bikes in different areas. The relationship between temporal characteristics of riding demand and land use (represented by point of interest, POI)is studied by factor analysis. The results show that the spatial distribution of aggregation areas of shared bikes is basically correlated to the spatial pattern of the urban road network in the area. There is little variation for the distribution of aggregation areas in different time periods, and the only difference is the volume of bike riding in different areas. Besides, it shows that temporal characteristics of riding demand and land use are related. Commercial land use is the dominating factor for the areas where the number of rented bicycles is less than that of returned bicycles in one day, which accounts for 40% of the total. For the areas where the number of rented bicycles is larger than that of returned bicycles in one day, residential land use is the dominating factor, accounting for 57% of the total. In areas with mixed land use, the difference between bicycle renting and returning is small and prone to fluctuate. In addition, the proportion of dominant factors of a temporal characteristics of riding demand may change between weekday and weekend, and the temporal characteristics of riding demand in a region are different between weekday and weekend.

Accessing the Impacts of Curb Parking Behavior on Traffic Flows Through Cellular Automata Models

ZHANG Yue, SUN Lishan, KONG Dewen, ZHANG Xin

2022, 40(3): 154-162. doi: 10.3963/j.jssn.1674-4861.2022.03.016

Abstract(891) HTML (296) PDF(33)

Abstract:
Curb parking may lead to several traffic issues, such as queue delay, slow traffic due to low-speed cruising, and reduced road capacity because of excessive parking spaces. In order to mitigate these issues, the impacts of curb parking on traffic flows are studied. Data of vehicle trajectory and speed is collected based on video recognition technique. Then, the characteristics of driving behaviors of the vehicles which use curb parking are analyzed. According to differences of driving behaviors, the process of curb parking is divided into eight steps: driving into the road, cruising for a parking space, finding a parking space, entering the space, parking, leaving the space, merging into traffic, and missing a parking space. Based on extracted data of parking and cruising behaviors of curb parking vehicles, a cellular automata model is proposed by taking multiple features into consideration, including their characteristics of car following, speed correcting, lane changing, and position updating. Time costs of both parking a vehicle and walking to destination are also considered for searching a target parking space. Compared with other vehicles, the impacts of behaviors of curb parking on the following vehicles, i.e., lane changing and lane merging, are analyzed. Besides, parameters of a simulation model are calibrated based on observed data of traffic flow, and the result shows that the degree of fit is 77.6%. Moreover, the influences of cruising speed on road capacity and delay time are analyzed by a simulation under differentiated parking intensities. The results show that delay time first increases, then decreases with the rise of traffic volume at a fixed cruising speed and parking intensity. At a low parking intensity, the impact of cruising speed on road capacity is small. In a scenario of high-volume traffic, when cruising speed declines from 30 km/h to 20 km/h, the saturation flow of outer lanes decreases by 500 veh/h, and the maximum delay time increases by 105 s.

A Short-term Prediction Model for Taxi Speed Based on XGBoost

XIAO Yu, ZHAO Jianyou, CHIGAN Du, LIU Qingyun

2022, 40(3): 163-170. doi: 10.3963/j.jssn.1674-4861.2022.03.017

Abstract(783) HTML (271) PDF(46)

Abstract:
An accurate short-term prediction for taxi speed is the premise of identifying abnormal driving behaviors of acceleration and deceleration in advance, which helps to enhance passengers'comfort and safety. A short-term prediction model is proposed to forecast real-time speed of taxis with an Extreme Gradient Boosting(XGBoost) model. The dataset of taxi speeds is divided into a training set and a test set, where a sequence of historical speed data in a time window are taken as an input variable, and the current speed data is taken as an output variable. The accuracy of the model is evaluated by a method called walk-forward validation. Based on the Bayesian algorithm, a hyperopt module is used to optimize model parameters, and a combination of optimal parameters can be obtained in a timely fashion. Experiments are carried out based on a data set of taxi GPS trajectory, which was collected in the City of Shenzhen on October 22, 2013, and the results of the proposed model are compared with those of two other models, including a non-parametric regression model and a neural network model. The results shows that the mean absolute error(MAE)and the root mean square error(RMSE)of the proposed model is 9.841 and 12.711. respectively. Due to the lack of regularity in the taxi speed sequence, the corrected R²(R² _adjusted)is 0.592, which outperform those of the non-parametric regression model and the neural network model. Besides, compared with the two other models, the proposed model has a better goodness of fit under the scenario that a taxi suddenly changes its speed in a significant way, which can be used to avoid degraded accuracy due to model overfitting.

Classification of the Level of Flight Delay Based on a VMD-MD-Clustering Method

WANG Xinglong, XU Yanfeng, JI Junrou

2022, 40(3): 171-178. doi: 10.3963/j.jssn.1674-4861.2022.03.018

Abstract(710) HTML (280) PDF(32)

Abstract:
Due to the increasing number of flights, the flight delay has been increasing in recent years. To mitigate this problem, a method for classifying flight delays is studied, which provides a theoretical basis for developing relevant measures and reducing the number of flight delays. A classification model is proposed based on six indicators from time, space, and efficiency aspects. These indicators include four numerical indicators, namely"delay time", "flying duration", "number of people affected by the delay", and"voyages affected by the delay", as well as two attribute indicators, i.e., "stopover flight or not"and"passenger capacity of delayed aircraft". Then, a method for classifying levels of flight delays is proposed, which combines the variational mode decomposition(VMD), Mahalanobis depth(MD)function, and K-means clustering, named as"VMD-MD-Clustering"(V-M-C)method. Firstly, non-normal and non-stationary multi-dimensional delay data are treated as a signal sequence with noise. Secondly, the VMD method is used to stabilize and normalize the delay data. Thirdly, the MD function is used to reduce the dimensionality of the data to one dimension(1D). Fourthly, the K-means method is applied to cluster the 1D signal data and output the level of flight delay. Finally, to evaluate the proposed method, a weighted support vector machine(SVM)is applied to analyze the classification results. The operation data collected from an airport in one month are used for validation. The validation results show that the proposed V-M-C method have an accuracy of 95.41%, which outperforms the K-means method with an accuracy of 81.9%. Study results show that the proposed V-M-C method has an enhanced accuracy and therefore, it is potentially useful for formulating flight-delay disposal plans and improving the punctuality of flight operations.

2022 Vol. 40, No. 3