A method and system for deducing the spatiotemporal distribution of key vehicles on bridges

Abstract

The present invention provides a method and system for deducing the spatiotemporal distribution of key vehicles on a bridge. The method first obtains the historical trajectory data of key vehicles passing through the target area, and based on the pre-established road network topology, uses a multi-order Markov chain to construct a dynamically updated vehicle driving path deduction model, and then places the historical trajectory data of the key vehicles into the vehicle driving path deduction model, and combines the current road section of the key vehicles to predict the maximum probability driving path of the key vehicles in real time, and then predicts the spatiotemporal distribution information of multiple key vehicles traveling to the target bridge based on the maximum probability driving paths of multiple key vehicles. If the spatiotemporal distribution information meets the preset risk assessment criteria, the spatiotemporal distribution information is pushed to the risk assessment module. Compared with the prior art, the method of the present invention realizes the accurate prediction of the future driving path of key vehicles, can dynamically update the predicted vehicle path according to the change of vehicle position, and has stronger foresight performance.

Description

Technical Field

[0001] The invention belongs to the technical field of traffic data analysis, and in particular relates to a method and system for deducing the spatiotemporal distribution of key vehicles on a bridge. Background Art

[0002] In bridge capsizing incidents, the main load-bearing bridges are single-column pier bridges, among which vehicle overload is the main cause of the overturning of single-column pier bridges, especially large trucks, buses, and two-passenger and one-hazardous heavy-duty vehicles. When these vehicles are traveling on the single-column pier bridge, the single-column pier bridge is susceptible to the effect of overload, and its lateral stability is poor, the anti-overturning ability will be further reduced, and when they are on the bridge at the same time, the risk of overturning of the single-column pier bridge increases sharply.

[0003] In the existing technology, the bridge overturning risk warning method mainly uses high-precision sensors to monitor the state changes of key parts of the bridge, such as displacement, inclination, etc., and uses data analysis and modeling techniques (such as historical data analysis, finite element model analysis, and data mining and machine learning algorithms) to predict the potential risks of the bridge. When the analysis data exceeds the preset threshold, the warning signal is immediately triggered so that emergency response measures can be taken in time, thereby effectively reducing the risk of bridge overturning. Although this method can accurately provide real-time warnings, it lacks predictability and cannot provide early warnings before the vehicle reaches the bridge. The above-mentioned key vehicles are of large mass and are difficult to stop directly when driving at high speeds, and it is still difficult to avoid the occurrence of bridge overturning incidents. Summary of the invention

[0004] The present invention provides a method and system for deducing the spatiotemporal distribution of key vehicles on a bridge, which are used to solve the problems existing in the prior art.

[0005] In a first aspect, to achieve the above-mentioned purpose, the present invention provides a method for deriving the spatiotemporal distribution of key vehicles on a bridge, comprising:

[0006] S10: Acquire historical trajectory data of key vehicles passing through the target area;

[0007] S20: Based on the pre-established road network topology, a multi-order Markov chain is used to build a dynamically updated vehicle driving path deduction model;

[0008] S30: placing the historical trajectory data of the key vehicle into the vehicle driving path deduction model, and combining the current road section of the key vehicle to predict the most probable driving path of the key vehicle in real time;

[0009] S40: predicting the time-space distribution information of the multiple key vehicles traveling to the target bridge according to the maximum probability driving paths of the multiple key vehicles;

[0010] S50: If the time-space distribution information meets the preset risk assessment standard, the time-space distribution information is pushed to the risk assessment module.

[0011] Through the above technical scheme, the method of the present invention constructs a dynamically updated vehicle driving path deduction model based on a multi-order Markov chain, places the historical trajectory data of key vehicles into the vehicle driving path deduction model, and obtains a personalized deduction path (maximum probability driving path) for each key vehicle. Compared with the existing technology, the method of the present invention has stronger predictability and realizes accurate prediction of the future driving path of key vehicles.

[0012] Preferably, step S20 includes:

[0013] S21: Define the vehicle segment status, the order of the multi-order Markov chain, the prediction step size and the state transition probability of the key vehicle in each road segment in the road network topology;

[0014] S22: Using a multi-order Markov chain, the state transition probability of the key vehicle from the current section to the next section in the road network topology is calculated;

[0015] S23: constructing a first state transition probability matrix according to the state transition probability;

[0016] S24: Calculating a second state transition probability matrix of the key vehicle under all vehicle depot states according to the first state transition probability matrix;

[0017] Among them, the vehicle driving path deduction model at least includes a multi-order Markov chain, a first state transfer probability matrix and a second state transfer probability matrix.

[0018] Preferably, the expression of the state transition probability in step S22 is:

[0019]

[0020] in, Indicates the starting point of the path is And the key vehicle moves from the current section to the front section Road sections Drive to the next section The probability of Calculate the starting point of the path in the data set for the state transition probability From the current section of the vehicle to the previous section Road sections The number of times, The driving path contains The number of paths.

[0021] Preferably, step S30 includes:

[0022] S31: Obtain the current state of the multi-order Markov chain in the vehicle driving path deduction model and the current state link corresponding to the road section where the key vehicle is currently located;

[0023] S32: After defining the Markov state set, the current state of the multi-order Markov chain is transmitted to the Markov state set for storage, and the current state of the multi-order Markov chain is initialized;

[0024] S33: Using the Markov state set and the second state transition probability matrix, the probability of the key vehicle going to each road section in the road network topology is predicted, and the maximum probability driving path of the key vehicle is screened out.

[0025] Preferably, before step S40, the method includes:

[0026] S400: Construct an average travel speed prediction model for each road section in the road network topology;

[0027] S401: Obtain the time points at which the phases of all traffic lights in the road network topology alternate.

[0028] Preferably, step S40 includes:

[0029] S41: Determine whether the most probable driving path of each key vehicle passes through the target bridge. If so, execute step S42. If not, end the prediction.

[0030] S42: Calculate the relative distance between the current position of the key vehicle tested in step S41 and the target bridge;

[0031] S43: Determine whether the relative distance is less than a preset distance threshold, if so, execute step S44, if not, end the prediction;

[0032] S44: Calculate the driving speed of the key vehicle detected in step S43 traveling through each section of the maximum probability driving path according to the average driving speed prediction model;

[0033] S45: After converting the maximum probability driving path into a vehicle driving path link, the node waiting time required for the key vehicle detected in step S43 to travel from the starting point to each node in the vehicle driving path link is calculated according to the driving speed and the time points at which each phase of all traffic lights in the road network topology alternates;

[0034] S46: Predicting the time-space distribution information of the key vehicle detected in step S43 traveling to the target bridge based on the driving speed, node waiting time and section information of the target bridge.

[0035] Preferably, the road section information of the target bridge includes the bridge length and the bridge road section node where the target bridge is located in the maximum probability driving path, and step S46 includes:

[0036] S461: Obtain bridge length and bridge section nodes;

[0037] S462: Calculate the bridge passing time required for the key vehicle to enter and exit the target bridge according to the driving speed, node waiting time, bridge section nodes and bridge length;

[0038] S463: Based on the bridge length and the bridge passing time, the temporal and spatial distribution information of multiple key vehicles traveling to the target bridge is predicted.

[0039] Preferably, the calculation expression of step S462 is:

[0040]

[0041]

[0042] in, is the time when the key vehicle enters the target bridge, is the time when the key vehicle leaves the target bridge. Node for key vehicles entering the bridge section speed, The speed of the key vehicle when entering the bridge. is the bridge length, is the node waiting time, Bridge section node Distance to the start of the bridge.

[0043] Preferably, before step S10, the method of the present invention further comprises:

[0044] S100: obtaining original trajectory data of key vehicles and performing data cleaning on the original trajectory data;

[0045] S101: Performing a topological check on the road network according to the road network data;

[0046] S102: performing trajectory matching on the original trajectory data after data cleaning and the road network after topology check to obtain a road network matching result of the trajectory;

[0047] S103: According to the road network matching result, the position of the trajectory point is corrected to the road network to obtain the historical trajectory data of the key vehicle.

[0048] In the second aspect, in order to achieve the purpose of the invention, the present invention also provides a bridge key vehicle time-space distribution deduction system, using the above bridge key vehicle time-space distribution deduction method, the system of the present invention comprises:

[0049] A data acquisition module, used to acquire historical trajectory data of key vehicles passing through the target area;

[0050] The model building module is used to build a dynamically updated vehicle driving path deduction model based on the pre-established road network topology using a multi-order Markov chain;

[0051] The path prediction module is used to place the historical trajectory data of key vehicles into the vehicle driving path deduction model, and combine the current road section of the key vehicles to predict the most probable driving path of the key vehicles in real time;

[0052] A spatial distribution prediction module is used to predict the temporal and spatial distribution information of multiple key vehicles traveling to the target bridge according to the maximum probability driving paths of multiple key vehicles;

[0053] The risk assessment module is used to receive the time-space distribution information when the time-space distribution information meets the preset risk assessment standard.

[0054] Compared with the prior art, the present invention has at least the following advantages:

[0055] (1) The method of the present invention is based on a multi-order Markov chain, taking into account factors such as the starting point of the vehicle's driving path, the vehicle's historical driving route, and the prediction step size, and establishes a personalized path prediction model for each key vehicle. This model can not only accurately predict the future driving path of the key vehicle, but also dynamically update the vehicle's predicted path according to the change in the vehicle's position.

[0056] (2) In order to solve the problem that the future spatiotemporal distribution of vehicles on bridges is difficult to predict, the method of the present invention takes into account the average speed of vehicles passing through different sections at different time points and the waiting time of vehicles at traffic lights when they arrive at intersections, and establishes a spatiotemporal distribution deduction model for key vehicles on bridges based on time-varying operation characteristics. This model can identify the spatiotemporal distribution of vehicles on bridges and can realize the identification of high-risk working conditions where multiple key vehicles pass through single-column pier bridges at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0058] Figure 1 The present invention is a flowchart of a method for deriving the spatiotemporal distribution of key vehicles on a bridge in accordance with a first embodiment of the present invention.

[0059] Figure 2 It is the current time-space distribution diagram of key vehicles in the time-space distribution deduction method of key vehicles on a bridge in the first embodiment of the present invention.

[0060] Figure 3 This is a diagram of the deduction result of the distribution of key vehicles on a bridge after step S40 in the method for deducing the spatiotemporal distribution of key vehicles in the first embodiment of the present invention.

[0061] Figure 4 It is a structural diagram of the time-space distribution deduction system of key vehicles on bridges in the second embodiment of the present invention. DETAILED DESCRIPTION

[0062] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. According to the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0063] In the description of the present invention, it should be noted that the terms "center", "up", "down", "left", "right", "vertical", "horizontal", "inside", "outside", etc. indicate directions or positional relationships based on the directions or positional relationships shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction. Therefore, they cannot be understood as limitations on the present invention.

[0064] This embodiment discloses a method and system for deducing the spatiotemporal distribution of key vehicles on a bridge, which are used to solve the problems existing in the prior art.

[0065] Embodiment 1

[0066] like Figure 1 As shown, this embodiment discloses a method and system for deducing the spatiotemporal distribution of key vehicles on a bridge, including:

[0067] S10: Acquire historical trajectory data of key vehicles passing through the target area.

[0068] In this embodiment, the target area includes bridge sections and other sections, and its specific road network range can be adjusted according to the predicted actual demand.

[0069] It needs to be explained that key vehicles mainly refer to key vehicles such as buses, large trucks, two passenger and one dangerous goods vehicles. The routes or driving logic of these key vehicles are almost fixed. They all have frequently traveled routes and their road changes are very regular during driving. Therefore, route prediction can be carried out based on the historical trajectory data of key vehicles and mathematical modeling.

[0070] Before step S10, the method of the present invention further includes:

[0071] S100: obtaining original trajectory data of key vehicles and performing data cleaning on the original trajectory data;

[0072] S101: Performing a topological check on the road network according to the road network data;

[0073] S102: performing trajectory matching on the original trajectory data after data cleaning and the road network after topology check to obtain a road network matching result of the trajectory;

[0074] S103: According to the road network matching result, the position of the trajectory point is corrected to the road network to obtain the historical trajectory data of the key vehicle.

[0075] It should be explained that the above-mentioned original trajectory data is the unprocessed GPS vehicle historical trajectory data of the key vehicle.

[0076] In this embodiment, the data cleaning method may adopt the Kerman filtering method. The road network data includes information such as road number, speed limit, length, start and end points, etc.

[0077] Through the above steps S100-S103, the method of the present invention first obtains the original trajectory data of buses, large trucks, two passenger and one dangerous goods vehicles, and then filters them to remove abnormal values, such as erroneous coordinate points caused by unstable signals. At the same time, the GPS data (original trajectory data) is smoothed using similar methods such as Kalman filtering to reduce the impact of noise. Next, the road network is topologically checked according to information such as road number, speed limit, length, starting and ending points. Then, the preset road network matching model is used to match the original trajectory data after data cleaning with the road network after topological checking to obtain the road network matching result of the trajectory. Finally, the road network matching result corrects the position of the trajectory point to the road network to obtain the historical trajectory data of key vehicles.

[0078] Compared with the prior art, the historical trajectory data of key vehicles obtained by the method of the present invention is more reliable.

[0079] After the above step S10, the method of the present invention further comprises:

[0080] S20: Based on the pre-established road network topology, a multi-order Markov chain is used to construct a dynamically updated vehicle driving path deduction model.

[0081] In this embodiment, the pre-established road network topology structure includes the connection relationship of each road section and the intersection information.

[0082] Furthermore, the above step S20 includes:

[0083] S21: Define the vehicle segment status, the order of the multi-order Markov chain, the prediction step size and the state transition probability of the key vehicles in each road segment in the road network topology.

[0084] In this embodiment, the above definition of vehicle segment status is to take each segment in the road network topology as a state. , then there are Since the driving paths of buses, large trucks, two passenger and one dangerous goods vehicles have historical similarities, more historical path information can be mined through multi-order Markov chains, thereby improving the accuracy and practicality of predictions. However, too high an order may result in too few sample data and increased calculations. Therefore, it is necessary to define an appropriate order. In order to have enough time to give advance warning to key vehicles before they enter the bridge, a sufficient prediction step length is required, that is, the number of sections that the current vehicle will travel in the future. Due to the uncertainty of the vehicle's driving path, a prediction step length that is too long will easily lead to a decrease in prediction accuracy, and a prediction step length that is too short will easily lead to a decrease in the time for advance warning. Therefore, it is necessary to comprehensively consider the prediction accuracy and the advance warning time and set the prediction step length. The state transition probability refers to the probability that the key vehicle will travel from the current section to another section. For the multi-order Markov chain model, it is necessary to calculate the transition probability from the current state and its previous n-1 states to the next state, that is, the probability of the vehicle's current section and its previous n-1 states to the next state. -1 The probability of traveling from one segment to another.

[0085] S22: Use a multi-order Markov chain to calculate the state transition probability of the key vehicle from the current section to the next section in the road network topology.

[0086] It should be explained that, since the probabilities of different sections traveling to other sections are different, the probabilities of different sections traveling to other sections need to be calculated separately. When calculating the state transition probability, the present invention takes into account the topological relationship of the road network topology and the continuity of the vehicle's driving path, and takes into account the different factors of the routes selected by key vehicles starting from different starting points, and then combines the historical trajectory matching data to accurately calculate the state transition probability.

[0087] The above step S22 is specifically as follows:

[0088] S221: According to the key vehicle ID, the vehicle historical trajectory data is divided into trajectory data sets corresponding to each vehicle, and the driving path and starting point information of each key vehicle are obtained in combination with the trajectory matching result.

[0089] S222: Divide the driving path sets corresponding to the historical trajectory data of key vehicles one by one, and record them as , It is the number of driving paths of a key vehicle. The elements in the set include the starting point of each vehicle path and the road sections passed by the path.

[0090] S223: The order of multi-order Markov chains is a path sliding window, which is a set of paths for each key vehicle. Each driving path is divided into a state transition probability calculation data set, and each state transition probability calculation data set includes Continuous driving sections, The next section of a continuous driving section and the starting point of the vehicle's driving path.

[0091] S224: Calculate the starting point of the path And the front of the road section where the key vehicle is currently located Road sections Drive to the next section under the condition The calculation formula is as follows:

[0092] (1)

[0093] in, Indicates the starting point of the path is And the key vehicle moves from the current section to the front section Road sections Drive to the next section The probability of Calculate the starting point of the path in the data set for the state transition probability From the current section of the vehicle to the previous section Road sections The number of times, The driving path contains The number of paths.

[0094] It is necessary to explain that, since the vehicle driving path is continuous, according to the topological relationship of the road network, when The road section where the vehicle is currently located When not adjacent, The next step is impossible to transfer to ,Right now is 0.

[0095] After the above step S22, the method of the present invention further comprises:

[0096] S23: Construct a first state transition probability matrix according to the state transition probability.

[0097] It should be explained that according to formula (1) The probability of transferring to all other states in the state, and constructing its probability into the first state transfer probability matrix. The expression of the first state transfer probability matrix is ​​as follows:

[0098] (2)

[0099] in, for The state transition probability matrix of transferring from one state to other states, each element in the matrix represents the probability of transferring to each state.

[0100] S24: Calculate the second state transition probability matrix of the key vehicle under all vehicle depot states according to the first state transition probability matrix.

[0101] Through the above step S24, the path starting point can be calculated as The vehicle is currently on the road ahead of Road sections The state transition probability matrix is ​​calculated according to the state transition probability calculation data set, and the second state transition probability matrix under all state conditions is calculated, which is recorded as .

[0102] Through the above steps S21-S24, the method of the present invention realizes the construction of a vehicle driving path deduction model. Compared with the prior art, the vehicle driving path deduction model of the present invention is mainly composed of a multi-order Markov chain, a first state transfer probability matrix and a second state transfer probability matrix, which fully considers various conditions in the road network and has higher prediction accuracy and stronger predictability.

[0103] After the above step S20, the method of the present invention further comprises:

[0104] S30: placing the historical trajectory data of the key vehicle into the vehicle driving path deduction model, and combining the current road section of the key vehicle to predict the most probable driving path of the key vehicle in real time.

[0105] Specifically, step S30 includes:

[0106] S31: Obtain the current state of the multi-order Markov chain in the vehicle driving path deduction model and the current state link corresponding to the road section where the key vehicle is currently located.

[0107] More specifically, Figure 2 As shown, the road section where the key vehicle is currently located is obtained according to the vehicle driving path deduction model, and the road section where the key vehicle is currently located is recorded as the current state link , remember the front of the current section of the road where the key vehicle is located The state of the Markov chain composed of road segments is , the starting point of the current vehicle trajectory is , so the current Markov state can be expressed as .

[0108] S32: After defining the Markov state set, the current state of the multi-order Markov chain is transmitted to the Markov state set for storage, and the current state of the multi-order Markov chain is initialized.

[0109] More specifically, since each step of the predicted path requires updating the Markov chain state to ensure that the next step of the Markov chain state is the latest state, and each step of the predicted path may obtain multiple road sections with non-zero probability, each step of the prediction may generate multiple updated Markov chain states. The Markov chain state set is recorded as , the current Markov chain state join in , after initializing the Markov state set, it is recorded as , each element in the set includes the path composed of the predicted road segments and the probability of the vehicle traveling on the path.

[0110] S33: Using the Markov state set and the second state transition probability matrix, the probability of the key vehicle going to each road section in the road network topology is predicted, and the maximum probability driving path of the key vehicle is screened out.

[0111] It should be explained that since the method of the present invention avoids insufficient advance warning time through multi-step path prediction, the prediction of multi-step road sections through the current state needs to be carried out in stages, with each step being a stage, and the current state needs to be updated for each prediction step.

[0112] More specifically, the above step S33 is from prediction step 1 to The cyclic step-by-step forecast includes:

[0113] S331: Cyclic Markov State Set ,include:

[0114] S3311: Get the current loop The Markov states visited are:

[0115] Remember the current loop The Markov states visited are , if the current prediction is step 1, then the current state is .

[0116] S3312: Obtain the corresponding second state transition probability matrix according to the current state, specifically:

[0117] According to the current Markov state chain and the path starting point, the second state transition probability matrix under all state conditions obtained from step S24 As the state transition probability matrix corresponding to this step.

[0118] S3313: Obtain the road sections with non-zero probabilities in the second state transition probability matrix.

[0119] It should be explained that the second state transition probability matrix represents the probability matrix of the vehicle's current Markov state driving to other paths in the road network. Some sections are not adjacent to the section where the vehicle is currently located, so the probability of driving to these sections is 0. Therefore, the sections with non-zero probabilities are selected from the second state transition probability matrix, that is, the adjacent sections of the section where the vehicle is currently located, and are expressed as , each element contains the road segment number and the probability of the current state to the road segment, is the number of road segments with non-zero probability.

[0120] S3314: According to the current second state transition probability matrix, all the road sections with non-zero probabilities Construct a predicted vehicle path.

[0121] If the Markov state set Empty, join in If Not empty, from the current Markov state Get the current status of the road segment information , and from Filter paths ending with The paths with the same road segment information are recorded as , respectively Each section of Perform splicing to obtain a new prediction set =[{ , },...,{ , }].

[0122] S3315: Calculation The probability of each updated predicted path in is calculated as follows:

[0123] (3)

[0124] in, For the key vehicle in the predicted path The probability of for A road section information in is the current Markov state arrive The probability of Vehicle on path The probability of driving on.

[0125] After the above probabilities are calculated, From the collection Delete it and join in middle.

[0126] S3316: Update the Markov state according to the probability calculation result of step S3315.

[0127] Specifically, the probability of the next step to each road section in the current state is obtained, that is, the state probability transfer matrix. When predicting the next stage, the Markov state should be updated according to the information of all road sections with non-zero probabilities in the second state transfer probability matrix. , respectively, with the current Markov state chain Splice and connect the first segment Remove to maintain the order of the Markov state chain , update the Markov state set, recorded as:

[0128] (4)

[0129] Furthermore, the current Markov state From the Markov state set , and the updated Markov state set Add to the Markov state set .

[0130] Furthermore, step S33 also includes:

[0131] S332: If the prediction step length is equal to , then the loop of step S331 ends, otherwise continue to execute step S331.

[0132] S333: According to the probability calculation result of the multi-step prediction output in step S331, the maximum probability path is selected and the maximum probability path is used as the predicted driving path .

[0133] After the above step S30, the method of the present invention further comprises:

[0134] S40: According to the maximum probability driving paths of the multiple key vehicles, the temporal and spatial distribution information of the multiple key vehicles driving to the target bridge is predicted.

[0135] It needs to be explained that Figure 3 As shown, the predicted driving path of the key vehicle is obtained Finally, facing the scenario where multiple key vehicles are distributed on the bridge at the same time, which is easy to cause bridge collapse or overturning, two passenger and one dangerous goods vehicles, large trucks, and buses are long and heavy, and are easy to be distributed on the same lane when there are fewer lanes, especially single-column pier bridges, which will increase the risk of bridge collapse or overturning. Therefore, it is necessary to predict and deduce the spatiotemporal distribution of key vehicles on the bridge, combine the bridge structure analysis method, evaluate the operation risk, and then give early warnings to drivers or adopt other control strategies.

[0136] Preferably, before step S40, the method of the present invention further comprises:

[0137] S400: Construct an average driving speed prediction model for each road segment in the road network topology.

[0138] Compared with the prior art, in order to accurately predict the temporal and spatial distribution of vehicles passing through bridges, the method of the present invention takes into account the average speed of vehicles in the corresponding time periods when they travel to different road sections. The statistical time interval of the average speed of the road section is , based on the average history of the road section, combined with the date characteristics (weekdays, weekends, holidays) and time period characteristics (time slices corresponding to each time period of the day), the time period length is set to , build LSTM or other deep learning prediction models, and obtain a high-precision road section average driving speed prediction model through vehicle historical trajectory data training, so as to predict the average driving speed of key vehicles in the predicted path.

[0139] S401: Obtain the time points at which the phases of all traffic lights in the road network topology alternate.

[0140] Compared with the prior art, the method of the present invention takes into account that key vehicles may pass through intersections while traveling along the predicted vehicle path, and the traffic lights at the intersections may cause the vehicles to stop and wait for the green light time. In order to improve the accuracy of the deduction of the spatiotemporal distribution of vehicles passing through bridges, the method of the present invention takes into account the influence of traffic lights when deducing the spatiotemporal distribution of vehicles. The traffic light information data of the intersection is obtained through the signal control system, or the traffic light information data of the intersection is calculated through the vehicle trajectory data, and the time point information of each phase of all traffic lights is recorded as , It is the number of all traffic lights in the road network topology. Each traffic light information contains all the time points of the traffic lights in all phases of the day. If there is no traffic light in that phase, it is recorded as empty.

[0141] After the above steps S401 and S402, step S40 is performed, specifically:

[0142] S41: Determine whether the most probable driving path of each key vehicle passes through the target bridge. If so, execute step S42; if not, end the prediction.

[0143] In this embodiment, the target bridge is mainly a single-column pier bridge. In other embodiments, the target bridge can be selected according to actual conditions.

[0144] S42: Calculate the relative distance between the current position of the key vehicle checked in step S41 and the target bridge.

[0145] Specifically, when the vehicle's predicted driving path When a single-column pier bridge is included, it is necessary to calculate the distance from the vehicle to the bridge in the predicted driving path to determine whether the distance condition for the vehicle time-space distribution deduction is met. The specific calculation formula of step S42 is as follows:

[0146] (5)

[0147] in, To predict the distance from the key vehicle to the target bridge in the driving path, is the distance from the location of the key vehicle to the first node in the predicted driving path, It is the length of the road section between the first node in the predicted driving path of the key vehicle and the first node of the road section where the target bridge is located. Indicates the predicted driving path of the key vehicle. It is the distance between the first node of the road section where the bridge is located and the entrance point of the target bridge.

[0148] S43: Determine whether the relative distance is less than a preset distance threshold. If so, execute step S44; if not, end the prediction.

[0149] Specifically, the distance from the key vehicle to the target bridge in the driving path is predicted. Distance threshold for vehicle spatiotemporal distribution deduction For comparison, if Less than , meets the deduction conditions, and can proceed to the next step. Greater than , then there is no need to perform time-space distribution deduction at present, set the time interval , regularly calculate the distance from the key vehicle to the target bridge in the predicted driving path, and judge whether the distance from the vehicle to the target bridge in the predicted driving path meets the vehicle spatiotemporal distribution deduction conditions.

[0150] S44: Calculate the driving speed of the key vehicle detected in step S43 at each section of the maximum probability driving path according to the average driving speed prediction model.

[0151] Specifically, set the forecast duration range, that is, the number of forecast periods According to the average driving speed prediction model of step S400, the daily history and the average driving speed of each section in the predicted driving path (maximum probability driving path) are input, and the average driving speed of each section in the predicted path in the future is obtained. The average driving speed during a period of time.

[0152] S45: After converting the maximum probability driving path into a vehicle driving path link, the node waiting time required for the key vehicle detected in step S43 to travel from the starting point to each node in the vehicle driving path link is calculated based on the driving speed and the time points at which the phases of all traffic lights in the road network topology alternate.

[0153] Different from the prior art, considering that the average speed of vehicles in different time periods of the route is different, and the time points at which key vehicles arrive at different routes are different, the problem of different running speeds of key vehicles when arriving at different routes is caused. The method of the present invention considers the time-varying nature of vehicle speed and the time that vehicles need to wait for traffic lights when passing through traffic lights when deducing the future spatiotemporal distribution of key vehicles, which can more accurately predict the time when vehicles pass through each node. The calculation steps of step S45 are as follows:

[0154] S451: Based on the most probable driving path of key vehicles , taking the current vehicle position as the starting point and the intersection in the path as the node, construct the vehicle driving path link, recorded as .

[0155] S452: Obtain the location, road section, time point, and average speed of the road section at the current time point of the key vehicle, which are recorded as , as the initialization information of the vehicle driving path link.

[0156] S453: Circulate the vehicle driving path link according to the driving speed and the time points of the phase alternation of all traffic lights in the road network topology Node, calculate the key vehicle detected by step S43 on the vehicle driving path link The node waiting time required from the starting point to each node.

[0157] Furthermore, the above step S453 is specifically as follows:

[0158] S4531: Get driving path link The current node information.

[0159] Specifically, remember the node currently visited in the loop , get the time point when the vehicle travels to the current node , if the current node is the starting point, then the current vehicle position, road section, time point, and the average speed of the road section at the current time point are , = , = , = .

[0160] S4532: Get the path length from the current node to the next node.

[0161] Specifically, in order to calculate the time from the current node to the next node corresponding to the road section, it is necessary to obtain the driving path link Midpoint and its next node The length of .

[0162] S4533: Obtain phase information of the path from the current node to the next node passing through the intersection.

[0163] Specifically, when a vehicle passes through a traffic light at an intersection, it needs to wait for different traffic lights at different phases, or even no waiting time, such as the right turn phase. Midpoint and its next node The road network topology relationship is used to obtain the phase of the vehicle when passing through the intersection.

[0164] S4534: Calculate the traffic light time required to wait from the current node to the next node.

[0165] Specifically, when the current node is the current position of the vehicle, there is no traffic light time between the current node and the next node, and not all intersections have traffic lights, so it is necessary to determine whether there is a traffic light time. and its next node If there is no traffic light waiting time on the path between the nodes, the traffic light waiting time is 0 when the vehicle passes the intersection. Driving path link The starting point of the vehicle, that is, the node corresponding to the vehicle's location, or the situation where there is no traffic light between adjacent intersections, such as the right turn phase. The waiting time of the traffic light when the vehicle passes the intersection is 0. The specific formula is:

[0166] (6)

[0167] According to the node and its next node The path between the nodes passes through the phase information of the intersection, based on the node and its next node The intersection between Select the corresponding intersection information and find the time point with the current node from the corresponding intersection traffic light information The latest green light time , Should be greater than , calculate the traffic light time that needs to be waited from the current node to the next node, the calculation formula is as follows:

[0168] (7)

[0169] in, is the traffic light time required to wait from the current node to the next node, The vehicle arrives at the current node time point, For the current node time point and the most recent green light time point.

[0170] S4535: Calculate the average vehicle speed when the key vehicle enters the road section.

[0171] Specifically, based on the road section between the current node and the next node, the average driving speed of the road section in each time period is predicted, combined with the time point when the vehicle enters the road section. , get the driving speed of the road section, recorded as .

[0172] S4536: Calculate the length of time it takes for the key vehicle to pass through the road section between the current node and the next node.

[0173] (8)

[0174] in, is the length of time that the key vehicle spends passing through the road section between the current node and the next node, is the length of the section between the current node and the next node, It is the average vehicle speed when key vehicles enter the road section.

[0175] S4536: Calculate the time for key vehicles to arrive at the node. The specific formula is:

[0176] (9)

[0177] S4537: Determine whether to end the loop.

[0178] Determine the next node of the current node Is it the last node of the driving path link? If so, end the loop of S4531-S4536. If not, continue the loop.

[0179] After the above step S45, step S40 further includes:

[0180] S46: Predicting the time-space distribution information of the key vehicle detected in step S43 traveling to the target bridge based on the driving speed, node waiting time and section information of the target bridge.

[0181] Specifically, the above step S46 includes:

[0182] S461: Obtain bridge length and bridge section nodes;

[0183] S462: Calculate the bridge passing time required for the key vehicle to enter and exit the target bridge according to the driving speed, node waiting time, bridge section nodes and bridge length;

[0184] S463: Based on the bridge length and the bridge passing time, the temporal and spatial distribution information of multiple key vehicles traveling to the target bridge is predicted.

[0185] Preferably, the calculation expression of step S462 is:

[0186] (10)

[0187] (11)

[0188] in, is the time when the key vehicle enters the target bridge, is the time when the key vehicle leaves the target bridge. Node for key vehicles entering the bridge section speed, The speed of the key vehicle when entering the bridge. is the bridge length, is the node waiting time, Bridge section node Distance to the start of the bridge.

[0189] After the above step S40, the acquisition of the temporal and spatial distribution information can be realized. In order to realize the early warning evaluation function, the method of the present invention further includes:

[0190] S50: If the time-space distribution information meets the preset risk assessment standard, the time-space distribution information is pushed to the risk assessment module.

[0191] Through the above steps S10-S50, the method of the present invention has the following advantages:

[0192] (1) The method of the present invention takes into account the starting point of the vehicle's driving path, the vehicle's historical driving route, and the prediction step length when predicting the future driving path of key vehicles, and establishes a personalized path prediction deduction model for each key vehicle. This can not only achieve accurate prediction of the future driving path of key vehicles, but also dynamically update the vehicle's predicted path according to changes in the vehicle's position.

[0193] (2) The method of the present invention takes into account the average speed of vehicles passing through different sections at different time points and the waiting time of vehicles at traffic lights when arriving at intersections in the prediction, and establishes a spatiotemporal distribution deduction model of key vehicles on bridges based on time-varying operation characteristics. It can identify the spatiotemporal distribution of vehicles on bridges and can realize the identification of high-risk working conditions where multiple key vehicles pass through single-column pier bridges at the same time.

[0194] Finally, it should be noted that the method of the present invention is not limited to the spatial distribution prediction of key vehicles, but can also be extended to the distribution prediction of any vehicle with regular trajectories.

[0195] Embodiment 2

[0196] like Figure 4 As shown, this embodiment discloses a system for deducing the spatiotemporal distribution of key vehicles on a bridge, which uses the method for deducing the spatiotemporal distribution of key vehicles on a bridge in the embodiment. The system specifically includes:

[0197] A data acquisition module, used to acquire historical trajectory data of key vehicles passing through the target area;

[0198] The model building module is used to build a dynamically updated vehicle driving path deduction model based on the pre-established road network topology using a multi-order Markov chain;

[0199] The path prediction module is used to place the historical trajectory data of key vehicles into the vehicle driving path deduction model, and combine the current road section of the key vehicles to predict the most probable driving path of the key vehicles in real time;

[0200] A spatial distribution prediction module is used to predict the temporal and spatial distribution information of multiple key vehicles traveling to a target bridge based on the maximum probability driving paths of multiple key vehicles;

[0201] The risk assessment module is used to receive the time-space distribution information when the time-space distribution information meets the preset risk assessment standard.

[0202] Those skilled in the art can understand that all or part of the steps of implementing the above-mentioned embodiments can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above-mentioned method embodiment; and the aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk, etc. Various media that can store program codes.

[0203] For those skilled in the art, various other corresponding changes and deformations can be made according to the technical solutions and concepts described above, and all of these changes and deformations should fall within the protection scope of the claims of the present invention.

Claims

1. A method for deducing the spatiotemporal distribution of key vehicles on a bridge, characterized in that: include: S10: Acquire historical trajectory data of key vehicles passing through the target area; S20: Based on the pre-established road network topology, a multi-order Markov chain is used to build a dynamically updated vehicle driving path deduction model; S30: placing the historical trajectory data of the key vehicle into the vehicle driving path deduction model, and combining the current road section of the key vehicle to predict the most probable driving path of the key vehicle in real time; S40: predicting the time-space distribution information of the multiple key vehicles traveling to the target bridge according to the maximum probability driving paths of the multiple key vehicles; S50: If the time-space distribution information meets the preset risk assessment standard, the time-space distribution information is pushed to the risk assessment module.

2. The method for deriving the spatiotemporal distribution of key vehicles on a bridge according to claim 1 is characterized in that: The S20 includes: S21: defining the vehicle segment state, the order of the multi-order Markov chain, the prediction step length and the state transition probability of the key vehicle in each road segment in the road network topology; S22: using the multi-order Markov chain, calculating the state transition probability of the key vehicle from the current road section to the next road section in the road network topology; S23: constructing a first state transition probability matrix according to the state transition probability; S24: Calculating a second state transition probability matrix of the key vehicle under all vehicle depot states according to the first state transition probability matrix; Among them, the vehicle driving path deduction model at least includes the multi-order Markov chain, the first state transition probability matrix and the second state transition probability matrix.

3. The method for deriving the spatiotemporal distribution of key bridge vehicles according to claim 2 is characterized in that: The expression of the state transition probability in S22 is: in, Indicates the starting point of the path is And the key vehicle moves from the current section to the front section Road sections Drive to the next section The probability of Calculate the starting point of the path in the data set for the state transition probability From the current section of the vehicle to the previous section Road sections The number of times; The driving path contains The number of paths.

4. The method for deriving the spatiotemporal distribution of key bridge vehicles according to claim 2 is characterized in that: The S30 includes: S31: Obtaining the current state of the multi-order Markov chain in the vehicle driving path deduction model and the current state link corresponding to the road section where the key vehicle is currently located; S32: After defining the Markov state set, the current state of the multi-order Markov chain is transmitted to the Markov state set for storage, and the current state of the multi-order Markov chain is initialized; S33: Using the Markov state set and the second state transition probability matrix, predict the probability of the key vehicle going to each road section in the road network topology, and screen out the most probable driving path of the key vehicle.

5. The method for deriving the spatiotemporal distribution of key vehicles on a bridge according to claim 1 is characterized in that: Before S40, the method for deducing the spatiotemporal distribution of key bridge vehicles includes: S400: constructing an average travel speed prediction model for each road section in the road network topology; S401: Obtain the time points at which the phases of all traffic lights in the road network topology alternate.

6. The method for deriving the spatiotemporal distribution of key bridge vehicles according to claim 5 is characterized in that: The S40 includes: S41: Determine whether the most probable driving path of each key vehicle passes through the target bridge. If so, execute step S42. If not, end the prediction. S42: Calculating the relative distance between the current position of the key vehicle that has passed the S41 inspection and the target bridge; S43: Determine whether the relative distance is less than a preset distance threshold, if so, execute step S44, if not, end the prediction; S44: Calculating the driving speed of the key vehicle detected by S43 when traveling through each road section in the maximum probability driving path according to the average driving speed prediction model; S45: after converting the maximum probability driving path into a vehicle driving path link, the node waiting time required for the key vehicle detected in step S43 to travel from the starting point to each node in the vehicle driving path link is calculated according to the driving speed and the time points at which each phase of all traffic lights in the road network topology alternates; S46: predicting the time-space distribution information of the key vehicle detected by S43 traveling to the target bridge according to the driving speed, the node waiting time and the road section information of the target bridge.

7. The method for deriving the spatiotemporal distribution of key vehicles on a bridge according to claim 6 is characterized in that: The road section information of the target bridge includes the bridge length and the bridge road section node where the target bridge is located in the maximum probability driving path, and S46 includes: S461: Acquire the bridge length and the bridge section node; S462: Calculating the bridge passing time required for the key vehicle to enter and exit the target bridge according to the driving speed, the node waiting time, the bridge section node and the bridge length; S463: Based on the length of the bridge and the time it takes to pass through the bridge, predict and obtain the temporal and spatial distribution information of multiple key vehicles traveling to the target bridge.

8. The method for deriving the spatiotemporal distribution of key bridge vehicles according to claim 7 is characterized in that: The calculation expression of S462 is: in, The time when the key vehicle enters the target bridge; The time it takes for the key vehicle to exit the target bridge; Node for key vehicles entering the bridge section speed; The speed of the key vehicle when it enters the bridge; is the length of the bridge; Waiting time for the node; Bridge section node Distance to the start of the bridge.

9. The method for deriving the spatiotemporal distribution of key vehicles on a bridge according to claim 1, characterized in that: Before S10, the method for deducing the spatiotemporal distribution of key bridge vehicles further includes: S100: acquiring original trajectory data of the key vehicle, and performing data cleaning on the original trajectory data; S101: Performing a topological check on the road network according to the road network data; S102: performing trajectory matching on the original trajectory data after data cleaning and the road network after topology check to obtain a road network matching result of the trajectory; S103: According to the road network matching result, the position of the trajectory point is corrected to the road network to obtain the historical trajectory data of the key vehicle.

10. A system for deducing the spatiotemporal distribution of key bridge vehicles, characterized in that: The method for deducing the spatiotemporal distribution of key vehicles on a bridge according to any one of claims 1 to 9 is adopted, and the spatiotemporal distribution deduction system for key vehicles on a bridge comprises: A data acquisition module, used to acquire historical trajectory data of key vehicles passing through the target area; The model building module is used to build a dynamically updated vehicle driving path deduction model based on the pre-established road network topology using a multi-order Markov chain; A path prediction module is used to place the historical trajectory data of the key vehicle into the vehicle driving path deduction model, and combine the current road section of the key vehicle to predict the most probable driving path of the key vehicle in real time; A spatial distribution prediction module is used to predict the temporal and spatial distribution information of multiple key vehicles traveling to the target bridge according to the maximum probability driving paths of multiple key vehicles; The risk assessment module is used to receive the time-space distribution information when the time-space distribution information meets a preset risk assessment standard.

Images