An emergency vehicle dynamic path planning method based on accident influence causality inference

Abstract

The application discloses an emergency vehicle dynamic path planning method based on accident influence causal inference, collects road network traffic flow information for causal inference model training, and selects the route with the shortest time consumption for an emergency vehicle to reach an accident site. By reading road network information, road network traffic flow data is obtained, target nodes, source nodes and other intersection nodes in the road network are set, and the type, location and time of the accident are determined, and corresponding accident information is generated as input. The influence value of space-time speed is obtained by the causal inference method, and the new link travel time under the time-varying period is obtained. The global optimal path planning method of the time-varying characteristics of the road network weight value is used for optimal path selection of the emergency vehicle. The application combines the machine learning method and the traditional causal effect model, realizes the dynamic path planning of the emergency vehicle, and improves the traffic efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent transportation, in particular to an emergency vehicle dynamic path planning method based on accident influence causality inference. BACKGROUND

[0002] With the increase of traffic volume, more traffic accidents occur, and China has become one of the countries with the most traffic accident casualties in the world. Traffic accidents endanger people's life and property safety, and the increasing emergency incidents and untimely emergency disposal will also cause great delay and impact on the traffic system. Emergency vehicle dynamic path planning is an important scientific and technological problem in emergency rescue and evacuation work. Due to the continuous change of road network traffic flow and other information, it is crucial to effectively develop a reasonable emergency rescue path, dynamically adjust and determine the optimal route strategy. The traditional emergency vehicle seeks the absolute shortest path, which often causes road congestion, causes unnecessary delay, and reduces the efficiency of emergency rescue. Therefore, the dynamic path planning for emergency vehicles will directly affect the efficiency of emergency rescue.

[0003] At present, there is a certain research basis for the dynamic route planning of emergency vehicles, but there are still some defects:

[0004] Firstly, the two types of analysis methods based on traffic accident influence, including the accident influence analysis method based on physics and the accident influence analysis method based on mathematical statistics, have defects. The accident influence analysis method based on physics uses traditional traffic flow theory to analyze the diffusion and dissipation law of traffic congestion, sets specific indexes and thresholds, identifies the congestion area and analyzes the accident influence degree. This method is suitable for single scene, needs to consume a lot of manpower to calibrate parameters and model scenes, and is not suitable for real-time analysis of individual events; the accident influence analysis method based on mathematical statistics can establish a traditional statistical model to analyze the influence of different independent variables on dependent variables, solve the selection bias and exogeneity assumption bias through exogenous / heterogeneous statistical model, solve the high dimension problem of data through data driven model, and realize high fitting accuracy, but cannot reveal the causal relationship.

[0005] Secondly, the static path planning method only considers the road network weight at the current time, and the optimal path obtained is the optimal path for a specific time domain, which cannot achieve the optimal solution in the whole time domain. The rolling path planning method can re-plan the path according to the new weight, but the actual driving path is spliced by the local optimal path in a specific space. However, the superposition of local optimal paths is not necessarily equal to the global optimal path, and it cannot be proved that the optimal path is maintained in the whole space.

[0006] The traditional emergency vehicle path assumes that the road network traffic state is static, which is inconsistent with the continuous change of the road network traffic state. The traditional method mainly calculates the travel time according to the static road network traffic flow data, and calculates the road resistance according to the geometric distance and road quality, and the planning method includes Dijkstra algorithm, A* algorithm, D* Lite algorithm and the like, but these algorithms can only meet the optimal result of a specific period, and cannot guarantee the optimal solution in the whole time domain. Therefore, in recent years, a time-varying road network dynamic planning method considering the time-varying road network has appeared, but the actual driving path determined by the method is spliced from the local optimal path in a specific space, and the optimal solution in the whole space cannot be guaranteed.

[0007] The heterogeneity of traffic accidents causes different influences. The occurrence of the accident is related to many factors, but it cannot be guaranteed that each factor will have an impact on the accident, that is, causality. Therefore, the existing path planning method is not ideal in determining the weight when determining the time-varying road network, which also brings insufficient effect to the dynamic path planning. SUMMARY

[0008] The purpose of the present application is to provide an emergency vehicle dynamic path planning method based on accident influence causal inference, which combines machine learning method with traditional causal effect model to realize dynamic path planning of emergency vehicle and improve traffic efficiency.

[0009] To achieve the above functions, the present application designs an emergency vehicle dynamic path planning method based on accident influence causal inference, which executes the following steps S1-S5 for emergency rescue vehicles to complete the optimal path selection between the emergency rescue vehicle and the emergency rescue destination:

[0010] Step S1: Obtain the departure location of the emergency rescue vehicle and the location of the emergency rescue destination; set the target node, the source node and other intersection nodes in the road network;

[0011] Step S2: Collect historical road network information to obtain the traffic state information on all paths in the area road network where the departure location of the emergency rescue vehicle and the emergency rescue destination are located, including vehicle flow, vehicle density and vehicle speed;

[0012] Step S3: Select control variables, the control variables include type variable, location variable and time variable which affect the occurrence of the accident, and construct a causal inference model according to the causal relationship between each control variable;

[0013] Step S4: Based on the traffic state information obtained in step S2 and the causal inference model constructed in step S3, the influence value of the space-time speed is obtained by the causal inference method, and the updated path travel time under the time-varying period is obtained;

[0014] Step S5: the optimal path selection of the emergency rescue vehicle is completed by the global optimal path planning method of the time-varying characteristics of the road network weight.

[0015] Advantages: compared with the prior art, the advantages of the present application include:

[0016] Since the model widely applied by machine learning, deep learning and the like only solves the correlation problem and cannot solve the causality problem, serious deviation may be caused. Therefore, a new model is established by combining the processing capability of machine learning on high-dimensional data and the mining capability of causality on causality. The double machine learning method is introduced in the present application, the dynamic path planning of the emergency vehicle is realized by combining the machine learning method with the traditional causality model, and the traffic efficiency is improved. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 is a flowchart of an emergency vehicle dynamic path planning method based on accident influence causality inference according to an embodiment of the present application;

[0018] Figure 2 is a causality flowchart according to an embodiment of the present application;

[0019] Figure 3 is a causality graph according to an embodiment of the present application;

[0020] Figure 4 is a dynamic path planning flowchart according to an embodiment of the present application;

[0021] Figure 5 is a path planning result graph according to an embodiment of the present application. DETAILED DESCRIPTION

[0022] The present application will be further described below in conjunction with the drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present application, and cannot be used to limit the protection scope of the present application.

[0023] The emergency vehicle dynamic path planning method based on accident influence causality inference provided by the embodiment of the present application is described with reference to Figure 1 , for the emergency rescue vehicle, the following steps S1-S5 are executed to complete the optimal path selection between the emergency rescue vehicle and the emergency rescue destination:

[0024] Step S1: the departure location of the emergency rescue vehicle and the emergency rescue destination location are obtained; the target node, the source node and other intersection nodes in the road network are set in the road network;

[0025] Step S1 is specifically as follows:

[0026] Obtaining the departure location s and the emergency rescue destination location t of the emergency rescue vehicle, representing each intersection in the road network as a node, representing s as a source node and t as a target node, d i representing the minimum cumulative weight from the source node s to the node i, in seconds, p i is a set of path nodes passed through from the source node s to the node i; s represents the minimum cumulative weight d i is a set of known nodes, and u represents the minimum cumulative weight d i is a set of unknown nodes, and the two sets satisfy wherein is the complement of u;

[0027] defining the known node as the node adjacent to the node where the emergency rescue vehicle is located, defining the unknown node as the node not adjacent to the node where the emergency rescue vehicle is located, and the cumulative weight of the unknown node is represented by ∞; and defining the source node s of the departure location of the emergency rescue vehicle as 0.

[0028] Step S2: collecting historical road network information to obtain the departure location of the emergency rescue vehicle and the traffic state information on all paths in the road network in the area where the emergency rescue destination location is located, including vehicle flow, vehicle density and vehicle speed;

[0029] Step S3: selecting control variables, the control variables including type variables, location variables and time variables affecting the occurrence of the accident, and constructing a causal inference model according to the causal relationship between the control variables;

[0030] Referring to Figure 2 , the specific steps of step S3 are as follows:

[0031] Step S3.1: updating the intersection s where the emergency rescue vehicle is located at the current time and the emergency rescue destination location t;

[0032] Step S3.2: solving the travel time of the emergency rescue vehicle on each road section as the initial road resistance of each road section according to the traffic state information obtained in step S2;

[0033] Step S3.3: defining as the speed of the emergency rescue vehicle on the road section between the node i and the node j after the accident occurs for dur, and defining x (i,j) as the length of the road section between the node i and the node j, T (i,j) ∈{0,1} as whether an accident occurs on the road section between the node i and the node j, T (i,j) = 1 indicating that an accident occurs, and T (i,j) = 0 indicating that no accident occurs; wherein dur represents the time length after the accident occurs, in minutes, and the value corresponding to dur is taken as an update period;

[0034] The calculation method of the conditional average treatment effect CATE is as follows:

[0035] The calculation method of the calculation time influence effect ATE is as follows:

[0036]

[0037] In order to solve the problem of heterogeneity, a plurality of control variables are usually introduced The control variables between node i and node j in period dur are represented as Xi,j,dur; therefore, the estimated result is the conditional average treatment effect CATE:

[0038]

[0039] Step S3.4: Selecting control variables, the control variables are a kind of variables affecting the type, location and time of accidents, and the control variables are divided into basic attribute variables X basic , traffic state variables X condi and road linear variables X align ; the screening of control variables is performed, thereby reducing bias, the control variables with strong correlation are eliminated by using Pearson correlation coefficient, the control variables (non-confusing variables) that are not conducive to estimation are deleted, and a judgment basis CSVI (Conditional Shapley Value Index) for variable elimination is constructed;

[0040] The preliminary selection of control variables is as shown in Table 1:

[0041] Table 1

[0042]

[0043]

[0044] The judgment basis CSVI for variable elimination is as follows:

[0045]

[0046] In the formula, ω represents a weight factor, Xi represents a control variable, and for the control variable If T=0 or T=1, the Shap value of is approximately 0, it is considered that is irrelevant to and should be eliminated; in the formula, Shap represents a method for explaining the prediction of a machine learning model, which is based on Shapley values in game theory and aims to explain the contribution degree of different control variables to accidents;

[0047] Step S3.5: According to the causal relationship between each control variable, a causal relationship diagram is constructed, and the causal relationship diagram is referred to Figure 3 , and a causal inference model DRL (Doubly Robust Learning) is constructed;

[0048] The specific steps of step S3.5 are as follows:

[0049] Step S3.5.1: Calculate the estimated value of the accident causal effect using the do operation:

[0050]

[0051] Wherein, P(·) represents the probability of occurrence of different control variables; therefore, the estimated value of the accident causal effect CATE dur is:

[0052]

[0053] Step S3.5.2: Calculate the propensity score e(x) under different conditional variable combinations:

[0054]

[0055] Step S3.5.3: Construct inverse propensity probability weighting IPW. Inverse propensity probability weighting (IPW) is a method for dealing with selection bias in causal inference. By using IPW, the treatment effect can be estimated in the observed data with selection bias, so as to obtain an estimate closer to the result of a randomized experiment, and solve the problem of sample imbalance:

[0056]

[0057] Step S3.5.4: Calculate the estimated value CATE IPW of the inverse propensity probability weighting IPW effect as follows:

[0058]

[0059] Since the estimation of IPW is too sensitive, an additional regression model about is added, and the estimated value CATE DR of the causal inference model DRL is:

[0060]

[0061] Wherein, represents the estimation by using classification machine learning, and represents the estimation by using regression machine learning.

[0062] Step S4: based on the traffic state information obtained in step S2 and the causal inference model constructed in step S3, the influence value of the space-time speed is obtained by a causal inference method, and the updated path travel time under the time-varying period is obtained;

[0063] The specific steps of step S4 are as follows:

[0064] Step S4.1: based on the causal inference model DRL, the speed influence value under different periods dur is calculated, and the estimated value of the speed influence on different paths based on the accident influence is calculated, and the speed of each path under different periods dur is calculated

[0065] Step S4.2: by the real-time changes of vehicle flow and vehicle speed, the travel time of the emergency rescue vehicle under different periods dur is obtained, and the minimum cumulative weight d of the source node s to the node i is calculated i :

[0066]

[0067] wherein, is the travel time of node i and node j under period dur.

[0068] Step S5: by the global optimal path planning method of the time-varying characteristics of the road network weight, the optimal path selection of the emergency rescue vehicle is completed.

[0069] The dynamic path planning flow chart is referred to Figure 4 , and the specific steps of step S5 are as follows:

[0070] Step S5.1: at the initial moment, the emergency rescue vehicle starts from the source node s=0, and the target node is t; the minimum cumulative weight d i is known, and the minimum cumulative weight d i is unknown, and the nodes adjacent to s=0 are d i The known nodes are all ∞, and the rest of the nodes are unknown nodes;

[0071] Step S5.2: traverse the set U; search for the node a with the minimum weight from the source node s in the set U; move the node from the set U to the set S, and store the weight of the node a and the corresponding path in d a and p a , respectively, wherein d a represents the minimum cumulative weight from the source node s to the node a, and p a is the path set from the source node s to the node a;

[0072] Thus, set S = {0, a}, set U = {1, 2, …, t}, p a = {0, a}, wherein

[0073] Step S5.3: comparing the updated adjacent node weight; let the weight of the road segment from node a to adjacent node j be d(a, j), compare the size of d a + d(a, j) and d j , since node j is not adjacent to source node s, d j is ∞, d a + d(a, j) < d j ; thus, the minimum cumulative weight d j from source node s to node j needs to be updated; the updated d j is as follows:

[0074]

[0075] wherein x (0,a) is the length of the road segment between the source node and node a, x (a,j) is the length of the road segment between node a and node j; and are the speeds of the emergency rescue vehicle on the road segment between the source node and node a and on the road segment between node a and node j respectively after the accident occurs dur;

[0076] Assume that the nodes adjacent to node a in step S5.2 are b, c, d, since nodes b, c, d are not adjacent to source node s, their weights are ∞, thus the minimum cumulative weights d b , d c , d d from source node s to nodes b, c, d are updated, at this time set S = {0, a, b, c, d}, set U = {1, 2, …, t}, wherein

[0077] Step S5.4: judging whether the period dur is updated, determining the path p i ; if the minimum cumulative weight dur < d a <2dur to node a, the road segment to node a will undergo period dur update;

[0078] After step S5.1-step S5.3, the current set S = {0, a, b, c, d}, set U = {1, 2, …, t}, wherein Assume that there is a node q in set U, q is an adjacent node of one of nodes b, c, d, assume that q is an adjacent node of node b, step S5.3 obtains that node q has the minimum cumulative weight dq , and dur < d q <2dur, indicating that the path p i ={0, a, b, q} is updated to b→q; define node b as a critical parent node, node q as a critical child node, and node b and node q together constitute a critical node pair;

[0079] Step S5.5: Calculate the minimum cumulative weight of the critical node pair; based on step S5.4, assume that the critical road segment b→q is in the first period dur and the second period dur, i.e. dur (num) =1 and dur (num) =2, the weights are d and d Assuming that the driving distance ratio of the emergency rescue vehicle on the road segment b→q is r, the driving distance ratio r in dur (num) =1 is calculated as follows:

[0080]

[0081] The minimum cumulative weight of the path p i ={0, a, b, q} is d q ; at this time, the set S={0, a, b, c, d, …, q} includes b, c, d and other adjacent node sets U={1, 2, …, t} in the ellipsis, and p i ={0, a, b, q};

[0082] The calculation formula of the minimum cumulative weight d i of the path p q ={0, a, b, q} in step S5.5 is as follows:

[0083]

[0084] In the formula, d b is the minimum cumulative weight from the source node s to node b, r is the driving distance ratio of the emergency rescue vehicle on the road segment b→q, and d are the weights of the critical road segment b→q in the first period dur and the second period dur respectively; x (0,a) is the length of the road segment between the source node and node a, and x (a,b) is the length of the road segment between node a and node b. and v are the speeds of the emergency rescue vehicle on the road segment between the source node and node a and on the road segment between node a and node b respectively after the accident occurs dur.

[0085] Step S5.6: Loop steps S5.2 to S5.5 until the set S is empty. Assume the next path selection is p i ={0,a,b,q,f}, based on step S5.3, the corresponding minimum cumulative weight d f is obtained; at this time, the set S={0,a,b,c,d,…,q,…,f}, wherein the first ellipsis means the same as step S5.5, the second ellipsis is the adjacent node of the node q, the set U={1,2,…,t}, and the last again loops steps S5.2 to S5.5, wherein the next update node α of the period dur satisfies dur α <3dur, until

[0086] The minimum cumulative weight d i of the path p f ={0,a,b,q,f} in step S5.6 is calculated according to the following formula:

[0087]

[0088] In the formula, d q is the actual cumulative weight of the path p i ={0,a,b,q}; x (q,f) is the length of the road segment between the node q and the node f, is the speed of the emergency rescue vehicle on the road segment between the node q and the node f after the accident occurs for 2dur. The path planning result diagram is shown in Figure 5 , wherein the dashed line represents the static path planning, the solid line represents the time-varying path planning, and the arrow represents the globally optimal time-varying path planning.

[0089] The embodiments of the present application are described in detail above with reference to the accompanying drawings, but the present application is not limited to the above-described embodiments, and various changes can be made within the knowledge of those skilled in the art without departing from the purpose of the present application.

Claims

1. A dynamic route planning method for emergency vehicles based on causal inference of accident impact, characterized in that, For emergency rescue vehicles, perform the following steps S1-S5 to select the optimal route from the emergency rescue vehicle to the emergency rescue destination: Step S1: Obtain the starting location of the emergency rescue vehicle and the destination location of the emergency rescue; set the target node, source node, and other intersection nodes in the road network; Step S2: Collect historical road network information to obtain the departure location of emergency rescue vehicles and the traffic status information of all paths in the road network of the area where the emergency rescue destination is located, including vehicle flow, vehicle density, and vehicle speed. Step S3: Select control variables, including type variables, location variables, and time variables that affect the occurrence of the accident, and construct a causal inference model based on the causal relationship between each control variable; The specific steps of step S3 are as follows: Step S3.1: Update the current location of the emergency rescue vehicle at the intersection s and the location of the emergency rescue destination t; Step S3.2: Based on the traffic status information obtained in step S2, calculate the travel time of the emergency rescue vehicle on each road segment, which will be used as the initial road resistance for each road segment; Step S3.3: Definition Let x be the speed of the emergency rescue vehicle on the road segment between node i and node j after the accident occurs. (i,j) T is the length of the road segment between node i and node j. (i,j) ∈{0,1} represents whether an accident has occurred on the road segment between node i and node j, T (i,j) =1 indicates an accident has occurred, T (i,j) =0 indicates that no accident occurred; Where dur represents the time elapsed since the incident, in minutes, and the value of dur is used as an update period. Step S3.4: Select control variables and divide them into basic attribute variables X. basic Traffic state variable X condi and the linear variable X of the road align The Pearson correlation coefficient was used to screen control variables, and the CSVI criterion for variable removal was constructed. The CSVI criteria for variable exclusion are as follows: In the formula, ω represents the weighting factor. This represents the control variable. Regardless of whether T=0 or T=1 right If the Shap value is approximately 0, then it is considered that... and Irrelevant, should be eliminated; Step S3.5: Construct a causal inference model DRL based on the causal relationships between the control variables; The specific steps of step S3.5 are as follows: Step S3.5.1: Calculate the estimated value of the causal effect of the accident: in, P(·) represents the probability of different control variables occurring; therefore, the estimated value of the causal effect of the accident is CATE. dur for: Step S3.5.2: Calculate the propensity score e(x) under different combinations of condition variables: Step S3.5.3: Construct an inverse bias probability-weighted IPW to address the sample imbalance problem: Step S3.5.4: Calculate the estimated value CATE of the reverse probability-weighted IPW effect. IPW As shown in the following formula: Add a clause about The regression model, the causal inference model DRL estimate CATE DR : in, This indicates that classification machine learning is used for estimation. This indicates that regression machine learning was used for estimation; Step S4: Based on the traffic state information obtained in step S2 and the causal inference model constructed in step S3, obtain the influence value of spatiotemporal speed through the causal inference method, and obtain the updated path travel time under the time-varying period. The specific steps of step S4 are as follows: Step S4.1: Calculate the speed impact values ​​under different cycle times (dur) based on the causal inference model DRL, and calculate the estimated speed impact on different paths based on the accident impact. Calculate the speed of each path under different cycle times (dur). Step S4.2: By monitoring real-time changes in vehicle flow and speed, obtain the travel time of emergency rescue vehicles under different period dur, and simultaneously calculate the minimum cumulative weight d from source node s to node i. i : in, Let be the travel time of node i and node j under period dur; Step S5: Using the global optimal path planning method based on the time-varying characteristics of road network weights, the optimal path selection for emergency rescue vehicles is completed; The specific steps of step S5 are as follows: Step S5.1: Initially, the emergency rescue vehicle departs from the source node s = 0, with its target node being t; the initial minimum cumulative weight d... i Given a node set S = {0}, the minimum cumulative weight d i The set of unknown nodes U = {1, 2, ..., t}; the node d adjacent to s = 0. i The first node is known, and the rest are unknown nodes, all with weights of ∞. Step S5.2: Traverse set U; search for the node a with the smallest weight from the source node s in set U; move this node from set U to set S, and store the weight of node a and its corresponding path in d respectively. a and p a In, where d a p represents the minimum cumulative weight from source node s to node a. a It is the set of paths traversed from source node s to node a; Therefore, set S = {0, a}, set U = {1, 2, ..., t}, p a ={0,a}, in Step S5.3: Compare and update the weights of neighboring nodes; let the weight of the road segment from node a to its neighboring child node j be d(a,j), and compare d... a +d(a,j) and d j The size of d is such that, since node j is not adjacent to the source node s, then d j For ∞, d a +d(a,j)<d j Therefore, it is necessary to update the minimum cumulative weight d from source node s to node j. j ; Updated d j As shown in the following formula: In the formula, x (0,a) x is the length of the path between the source node and node a. (a,j) Let be the length of the road segment between node a and node j; and These represent the speeds of the emergency rescue vehicles on the road segment between the source node and node a, and on the road segment between node a and node j, respectively, after the accident occurred. Assume that in step S5.2, the nodes adjacent to node a are b, c, and d. Since nodes b, c, and d are not adjacent to the source node s, their weights are ∞. Therefore, update the minimum cumulative weight d from the source node s to nodes b, c, and d. b d c d d At this point, set S = {0, a, b, c, d} and set U = {1, 2, ..., t}, where... Step S5.4: Determine if the periodic dur has been updated, and determine the path p. i If the minimum cumulative weight dur to node a is less than d a If the value is less than 2dur, then the path to node a will undergo periodic dur updates. After steps S5.1-S5.3, the current set S = {0, a, b, c, d} and the set U = {1, 2, ..., t}, where... Suppose there exists a node q in set U, where q is a neighbor of one of nodes b, c, and d. Assuming q is a neighbor of node b, step S5.3 calculates that node q has the minimum cumulative weight d. q And satisfy dur < d q <2dur indicates that the path p is followed. i ={0,a,b,q}, the periodic update path is b→q; define node b as the critical parent node and node q as the critical child node, and nodes b and q together form a critical node pair; Step S5.5: Calculate the minimum cumulative weight of the critical node pair; based on step S5.4, assume the critical segment b→q occurs in the first cycle dur and the second cycle dur, i.e. dur (num) =1 and dur (num) The weights with a value of 2 are respectively and Assuming the proportion of the emergency rescue vehicle's travel distance from b to q is r, the following formula is used to calculate the distance traveled in dur. (num) =1 of the travel distance ratio r: Then path p i The minimum cumulative weight of {0, a, b, q} is d. q At this point, the set S = {0, a, b, c, d, ..., q}, where the ellipsis also includes the set of other adjacent child nodes of b, c, and d, U = {1, 2, ..., t}, p i ={0,a,b,q}; Path p i The minimum cumulative weight d of {0, a, b, q} q The calculation formula is as follows: In the formula, d b Let be the minimum cumulative weight from source node s to node b, and r be the proportion of the distance traveled by emergency rescue vehicles on road segment b→q. and These represent the weights of the critical road segment b→q in the first and second cycle dur, respectively; x (0,a) x is the length of the path between the source node and node a. (a,b) This represents the length of the road segment between node a and node b. and These represent the speeds of the emergency rescue vehicles on the road segment between the source node and node a, and on the road segment between node a and node b, respectively, after the accident occurred. Step S5.6: Repeat steps S5.2 to S5.5 until a set is reached. Assume the next path is p i ={0,a,b,q,f}, based on step S5.3, obtain the corresponding minimum cumulative weight d. f At this point, the set S = {0, a, b, c, d, ..., q, ..., f}, where the first ellipsis has the same meaning as in step S5.5, and the second ellipsis represents the adjacent child node of child node q. The set U = {1, 2, ..., t}. Finally, steps S5.2 to S5.5 are repeated, where the next update node α of period dur satisfies dur < d. α <3dur, until Path p i The minimum cumulative weight d of {0, a, b, q, f} f The calculation formula is as follows: In the formula, d q For path p i = the actual cumulative weight of {0, a, b, q}; x (q,f) Let be the length of the road segment between node q and node f. The speed of the emergency rescue vehicle on the road segment between node q and node f 2dur after the accident.

2. The emergency vehicle dynamic route planning method based on causal inference of accident impact as described in claim 1, characterized in that, Step S1 is as follows: Obtain the starting location 's' and the destination location 't' of the emergency rescue vehicle. Represent each intersection in the road network as a node, with 's' as the source node and 't' as the destination node. i p represents the minimum cumulative weight from source node s to node i, in seconds. i It is the set of path nodes traversed from source node s to node i; S represents the minimum cumulative weight d. i Given a set of nodes, U represents the minimum cumulative weight d. i An unknown set of nodes, two sets satisfying in It is the complement of U; A known node is defined as a node adjacent to the node where the emergency rescue vehicle is located; an unknown node is defined as a node that is not adjacent to the node where the emergency rescue vehicle is located, and its cumulative weight is represented by ∞; the source node to which the departure location of the emergency rescue vehicle belongs is defined as s = 0.

Images