Intelligent network connection automobile running safety risk early warning method for fixed route driving

Abstract

The invention relates to the technical field of intelligent networked automobiles, and particularly discloses an intelligent networked automobile running safety risk early warning method for a fixed route, which comprises the following steps: acquiring single-day multi-frame historical running data of an intelligent networked automobile; dividing the single-day multi-frame historical operation data into single-day multi-pass historical operation data; determining a vehicle driving state corresponding to each frame of historical operation data in single-trip historical operation data of a single day; determining the scene category of the same position point on the fixed route according to the vehicle driving state corresponding to the same accumulated driving mileage in the historical operation data of multiple shifts of the single day; updating the scene category of each position point; and when the intelligent networked automobile runs to each position point on the fixed route, performing early warning on the operation safety risk of the intelligent networked automobile according to the final scene category of each position point on the fixed route. According to the invention, the complexity of the research and judgment model can be effectively reduced, and the research and judgment instantaneity and accuracy are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent networked vehicles, and more specifically, to an operation safety risk warning method for an intelligent networked vehicle traveling on a fixed route. Background Art

[0002] Intelligent connected vehicles are the development direction of the global automotive industry. my country and major developed countries in the world are actively promoting the technology and industrial development of intelligent connected vehicles.

[0003] At present, in terms of industrial scale, the demonstration application of intelligent networked vehicles carried out in various places has gradually become normalized, and fixed-route operating vehicles such as unmanned minibuses and unmanned logistics vehicles have been deployed on a large scale. In terms of technical solutions, the mainstream vehicle operation risk warning is based on trajectory prediction, grid network occupation and other methods, focusing only on vehicle collision risks, traffic efficiency and other aspects. There is no observation angle from the perspective of safe and civilized driving awareness. There is a lack of efficient mining methods for operation safety incidents such as long-term parking in straight lanes, low accuracy of platform parking positions, and failure to slow down before the stop line at intersections. In terms of platform tools, most of the intelligent networked vehicle supervision platforms deployed in various places only support real-time data viewing and historical data statistics functions, and lack effective mining and warning functions for potential operation safety incidents, resulting in insufficient platform supervision capabilities, and unable to fully support the work of relevant units in risk prevention, emergency response and other aspects, and give full play to the social benefits that platform tools can bring. Summary of the invention

[0004] In order to address the deficiencies in the prior art, the present invention provides an operational safety risk warning method for intelligent connected vehicles traveling on fixed routes, so as to make full use of the massive real-time vehicle operation data received by the supervision platform, mine operational safety events and issue warnings.

[0005] As a first aspect of the present invention, a method for early warning of operating safety risks of an intelligent networked vehicle traveling on a fixed route is provided, comprising the following steps:

[0006] Step S1: when the intelligent network-connected vehicle is cyclically traveling along a fixed route, multiple frames of historical operation data of the intelligent network-connected vehicle in a single day are obtained; wherein each frame of historical operation data includes a timestamp, longitude and latitude, driving speed, driving acceleration, angular velocity, steering wheel angle, and accumulated driving mileage;

[0007] Step S2: dividing the multi-frame historical operation data of the intelligent connected vehicle in a single day into the historical operation data of multiple shifts of the intelligent connected vehicle in a single day by a two-dimensional mean shift clustering algorithm; wherein the same accumulated mileage in the historical operation data of multiple shifts of the intelligent connected vehicle in a single day corresponds to the same position point on the fixed route;

[0008] Step S3: Determine the vehicle driving state corresponding to each frame of historical operation data in the historical operation data of a single trip of the intelligent connected vehicle on a single day by using a three-dimensional Gaussian mixture model clustering algorithm, wherein the vehicle driving state is divided into “straight driving, turning, and parking”;

[0009] Step S4: determining the route scene category corresponding to the same cumulative mileage according to the vehicle driving status corresponding to the same cumulative mileage in the historical operation data of multiple trips of the intelligent connected vehicle in a single day, that is, determining the scene category of the same position point on the fixed route; wherein the route scene categories are divided into “starting point, straight road, curve, intersection, platform”;

[0010] Step S5: using a voting mechanism to update the scene category of each location point on the fixed route to obtain a final scene category of each location point on the fixed route;

[0011] Step S6: When the intelligent network-connected vehicle travels to each position point on the fixed route, an early warning of the operation safety risk of the intelligent network-connected vehicle is issued according to the final scene category of each position point on the fixed route.

[0012] Furthermore, in the process of the intelligent network-connected vehicle cyclically driving along a fixed route, obtaining the multi-frame historical operation data of the intelligent network-connected vehicle for a single day also includes:

[0013] Downsampling the multi-frame historical operation data of the intelligent network-connected vehicle in a single day to obtain the multi-frame historical operation data of the intelligent network-connected vehicle after downsampling in a single day; wherein the number of frames of the historical operation data of the intelligent network-connected vehicle in a single day is reduced to 1 Hz by setting interval sampling;

[0014] According to the correspondence between "timestamp-driving speed" in the multiple frames of historical operation data of the intelligent connected vehicle after downsampling on a single day, the historical operation data with a driving speed lower than 80% of the maximum driving speed of the day in the multiple frames of historical operation data of the intelligent connected vehicle after downsampling on a single day are eliminated.

[0015] Furthermore, the method of dividing the multi-frame historical operation data of the intelligent networked vehicle in a single day into the multi-trip historical operation data of the intelligent networked vehicle in a single day by using a two-dimensional mean shift clustering algorithm also includes:

[0016] Compare the timestamps of adjacent frames of historical operation data of the intelligent connected vehicle on a single day in sequence. If the timestamp difference between the current frame historical operation data and the next frame historical operation data of the intelligent connected vehicle on a single day is greater than 300 seconds for the first time, the current frame historical operation data is considered to be the last frame historical operation data of the first shift historical operation data of the intelligent connected vehicle on a single day. Then, the first shift operation time of the intelligent connected vehicle on a single day is calculated according to the timestamp of the last frame historical operation data in the first shift historical operation data and the timestamp of the first frame historical operation data. Half of the first shift operation time is used as the radius parameter of the two-dimensional mean shift clustering algorithm.

[0017] The two-dimensional mean shift clustering algorithm is used to determine the geometric center of the historical operation data of subsequent shifts of the intelligent connected vehicle in a single day. Based on the timestamps corresponding to the geometric centers, the first frame of historical operation data and the last frame of historical operation data in the subsequent shifts are determined by the timestamp comparison method mentioned above, thereby dividing the historical operation data of multiple shifts of the intelligent connected vehicle in a single day.

[0018] Furthermore, the two-dimensional mean shift clustering algorithm comprises the following steps:

[0019] (1) Initialize all frame historical running data as the class center point;

[0020] (2) traverse each cluster center point, calculate the time dimension mean of all points within the radius parameter range, and replace the timestamp of the original cluster center point;

[0021] (3) De-duplicate the cluster center points;

[0022] (4) Repeat steps (2) and (3) until the number of cluster centers no longer changes.

[0023] Furthermore, the vehicle driving state corresponding to each frame of historical operation data of a single trip of a smart connected vehicle on a single day determined by the three-dimensional Gaussian mixture model clustering algorithm also includes:

[0024] By using the three-dimensional Gaussian mixture model clustering algorithm, according to the corresponding relationship of "driving speed-driving acceleration-angular velocity" of each frame of historical running data in the historical running data of a single trip of the intelligent networked vehicle on a single day, the vehicle driving state corresponding to each frame of historical running data in the historical running data of a single trip of the intelligent networked vehicle on a single day is determined;

[0025] Among them, the calculation formula of the angular velocity ω in each frame of historical running data is:

[0026]

[0027] Among them, ω represents the angular velocity, v represents the center speed of the vehicle's rear axle, L represents the vehicle's wheelbase, and δ represents the front wheel deviation angle.

[0028] Furthermore, the three-dimensional Gaussian mixture model clustering algorithm comprises the following steps:

[0029] (1) Initialize the algorithm parameters: set the category to 3, the covariance matrix type to share the same general covariance matrix, the maximum number of iterations to 500, and the convergence threshold to 0.001;

[0030] (2) For each data point, the posterior probability of the data point belonging to the three vehicle driving states of “going straight, turning, and stopping” is calculated based on the current parameter estimation value;

[0031] (3) Update the center point, covariance matrix and weight coefficient of various vehicle driving states according to the posterior probability;

[0032] (4) Repeat steps (2) and (3) until the log-likelihood function reaches the convergence threshold or the algorithm reaches the maximum number of iterations.

[0033] Furthermore, the method of determining the route scenario category corresponding to the same cumulative mileage according to the vehicle driving state corresponding to the same cumulative mileage in the historical operation data of multiple shifts of the intelligent connected vehicle in a single day, that is, determining the scenario category of the same position point on the fixed route, also includes:

[0034] When the driving states of all vehicles corresponding to the same cumulative driving mileage are all "parked", the route scene category corresponding to the same cumulative driving mileage is determined to be "platform", that is, the scene category of the same position point on the fixed route is determined to be "platform"; wherein the first "platform" on the fixed route is classified as the "starting point";

[0035] When the driving states of all vehicles corresponding to the same cumulative driving mileage are all "straight driving", the route scene category corresponding to the same cumulative driving mileage is determined to be "straight road", that is, the scene category of the same position point on the fixed route is determined to be "straight road";

[0036] When the driving states of all vehicles corresponding to the same cumulative driving mileage are all "turning", it is determined that the route scene category corresponding to the same cumulative driving mileage is "curve", that is, the scene category of the same position point on the fixed route is determined to be "curve";

[0037] When all vehicle driving states corresponding to the same accumulated mileage include "parking" and "going straight", the route scene category corresponding to the same accumulated mileage is determined to be "intersection", that is, the scene category of the same position point on the fixed route is determined to be "intersection".

[0038] Furthermore, the method of updating the scene category of each location point on the fixed route by using a voting mechanism to obtain a final scene category of each location point on the fixed route also includes:

[0039] The route range covered by a 30-meter length on a fixed route is taken as the interval window. When the scene categories of multiple location points in the same interval window are different, a voting mechanism is used to confirm the scene category of the central location point in the same interval window. If a tie occurs, the scene category of the central location point is updated to the scene category of the next location point in the same interval window.

[0040] Furthermore, when the intelligent network-connected vehicle travels to each position point on the fixed route, the warning of the operation safety risk of the intelligent network-connected vehicle according to the final scene category of each position point on the fixed route also includes:

[0041] When the intelligent networked vehicle drives to an "intersection" scene on a fixed route, the automatic driving control strategy of the intelligent networked vehicle is analyzed by comprehensively analyzing the braking, throttle opening and vehicle speed change trends 30 meters before the stop line of the intersection, and an early warning is issued when the automatic driving control strategy presents an operational safety risk;

[0042] When the intelligent networked vehicle travels to a "straight road" scenario on a fixed route, the average speed of the road section, the parking time and the lane selection are comprehensively analyzed to determine whether the intelligent networked vehicle has entered a minimum risk state, and an early warning is issued when the intelligent networked vehicle has not entered a minimum risk state;

[0043] When the smart connected vehicle drives to a "platform" scene on a fixed route, the longitude and latitude information of the platform where the smart connected vehicle stops is integrated, and the variance value of the longitude and latitude information of the platform where the smart connected vehicle stops is analyzed. When the variance value is greater than a threshold, it is considered that the smart connected vehicle's ability to stop at the platform is weak and an early warning is issued;

[0044] The abnormal driving behavior of the intelligent networked vehicle is analyzed by combining the acceleration and longitude and latitude information of the intelligent networked vehicle on the entire fixed route; the abnormal driving behavior of the intelligent networked vehicle with an acceleration greater than 3m / s 2 Or less than -5m / s 2 It can also issue early warnings for sudden acceleration / deceleration situations, and for situations where the position of the smart connected vehicle deviates from the fixed route by more than 10.5 meters.

[0045] The method for early warning of operational safety risks of intelligent connected vehicles traveling on fixed routes provided by the present invention has the following advantages: it does not rely on prior information such as high-precision maps, vehicle driving routes, and operating shifts, and can automatically estimate road section information only through the vehicle's one-day driving status data, including five road section area categories of "starting point, straight road, curve, intersection, and platform". Furthermore, the vehicle's real-time operating status data can be combined to make targeted analysis and judgment on the vehicle's operating safety status, which can effectively reduce the complexity of the judgment model and improve the real-time and accuracy of the judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the present invention, but do not constitute a limitation of the present invention.

[0047] Figure 1 A flow chart of the method for early warning of operational safety risks of an intelligent networked vehicle traveling on a fixed route provided by the present invention.

[0048] Figure 2 A schematic diagram of mean shift clustering of multiple shift operation data provided by the present invention.

[0049] Figure 3 This is a schematic diagram of clustering the Gaussian mixture model of vehicle driving status provided by the present invention. DETAILED DESCRIPTION

[0050] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation method, structure, characteristics and effects of the operation safety risk warning method for intelligent networked vehicles traveling on fixed routes proposed by the present invention are described in detail below in combination with the accompanying drawings and preferred embodiments. Obviously, the described embodiments are part of the embodiments of the present invention, not all of them. Based on 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.

[0051] In this embodiment, a method for early warning of safety risks of operation of an intelligent networked vehicle traveling on a fixed route is provided. Figure 1 As shown, the operation safety risk warning method for an intelligent connected vehicle traveling on a fixed route includes:

[0052] Step S1: when the intelligent network-connected vehicle is cyclically traveling along a fixed route, multiple frames of historical operation data of the intelligent network-connected vehicle in a single day are obtained; wherein each frame of historical operation data includes a timestamp, longitude and latitude, driving speed, driving acceleration, angular velocity, steering wheel angle, and accumulated driving mileage;

[0053] It should be noted that the timestamp in the format of "year-month-day-hour-minute-second-millisecond" is converted into the timestamp in the format of "accumulated seconds" for the day.

[0054] It should be noted that the starting point of the fixed route is taken as the zero point, the Haversine formula is used to calculate the displacement distance between frames, and the cumulative mileage corresponding to each frame of historical operation data is obtained by integration.

[0055] Preferably, the step of obtaining multi-frame historical operation data of the intelligent network-connected vehicle for a single day during the cyclic driving of the intelligent network-connected vehicle along a fixed route further includes:

[0056] Downsampling the multi-frame historical operation data of the intelligent network-connected vehicle in a single day to obtain the multi-frame historical operation data of the intelligent network-connected vehicle after downsampling in a single day; wherein the number of frames of the historical operation data of the intelligent network-connected vehicle in a single day is reduced to 1 Hz by setting interval sampling;

[0057] In order to enhance the data scatter point clustering effect, according to the correspondence between "timestamp-driving speed" in the multiple frames of historical operation data of the intelligent connected vehicle after downsampling on a single day, the historical operation data with a driving speed lower than 80% of the maximum driving speed of the day are eliminated.

[0058] Step S2: dividing the multi-frame historical operation data of the intelligent connected vehicle in a single day into the historical operation data of multiple shifts of the intelligent connected vehicle in a single day by a two-dimensional mean shift clustering algorithm; wherein the same accumulated mileage in the historical operation data of multiple shifts of the intelligent connected vehicle in a single day corresponds to the same position point on the fixed route;

[0059] Preferably, the method of dividing the multi-frame historical operation data of the intelligent connected vehicle in a single day into the multi-trip historical operation data of the intelligent connected vehicle in a single day by a two-dimensional mean shift clustering algorithm further includes:

[0060] Compare the timestamps of adjacent frames of historical operation data of the intelligent connected vehicle on a single day in sequence. If the timestamp difference between the current frame historical operation data and the next frame historical operation data of the intelligent connected vehicle on a single day is greater than 300 seconds for the first time, the current frame historical operation data is considered to be the last frame historical operation data of the first shift historical operation data of the intelligent connected vehicle on a single day. Then, the first shift operation time of the intelligent connected vehicle on a single day is calculated according to the timestamp of the last frame historical operation data in the first shift historical operation data and the timestamp of the first frame historical operation data. Half of the first shift operation time is used as the radius parameter of the two-dimensional mean shift clustering algorithm.

[0061] The two-dimensional mean shift clustering algorithm is used to determine the geometric center of the historical operation data of subsequent shifts of the intelligent connected vehicle in a single day. Based on the timestamps corresponding to the geometric centers, the first frame of historical operation data and the last frame of historical operation data in the subsequent shifts are determined by the timestamp comparison method mentioned above, thereby dividing the historical operation data of multiple shifts of the intelligent connected vehicle in a single day.

[0062] Specifically, Figure 2 As shown, the two-dimensional mean shift clustering algorithm includes the following steps:

[0063] (1) Initialize all frame historical running data as the class center point;

[0064] (2) traverse each cluster center point, calculate the time dimension mean of all points within the radius parameter range, and replace the timestamp of the original cluster center point;

[0065] (3) De-duplicate the cluster center points;

[0066] (4) Repeat steps (2) and (3) until the number of cluster centers no longer changes.

[0067] Step S3: Determine the vehicle driving state corresponding to each frame of historical operation data in the historical operation data of a single trip of the intelligent connected vehicle on a single day by using a three-dimensional Gaussian mixture model clustering algorithm, wherein the vehicle driving state is divided into “straight driving, turning, and parking”;

[0068] Preferably, the method of determining the vehicle driving state corresponding to each frame of historical operation data of a single trip of a smart connected vehicle on a single day by a three-dimensional Gaussian mixture model clustering algorithm further includes:

[0069] By using the three-dimensional Gaussian mixture model clustering algorithm, according to the corresponding relationship of "driving speed-driving acceleration-angular velocity" of each frame of historical running data in the historical running data of a single trip of the intelligent networked vehicle on a single day, the vehicle driving state corresponding to each frame of historical running data in the historical running data of a single trip of the intelligent networked vehicle on a single day is determined;

[0070] Among them, the calculation formula of the angular velocity ω in each frame of historical running data is:

[0071]

[0072] Among them, ω represents the angular velocity, v represents the center speed of the vehicle's rear axle, L represents the vehicle's wheelbase, and δ represents the front wheel deviation angle.

[0073] Specifically, Figure 3 As shown, the three-dimensional Gaussian mixture model clustering algorithm includes the following steps:

[0074] (1) Initialize the algorithm parameters: set the category to 3, the covariance matrix type to share the same general covariance matrix, the maximum number of iterations to 500, and the convergence threshold to 0.001;

[0075] (2) For each data point, the posterior probability of the data point belonging to the three vehicle driving states of “going straight, turning, and stopping” is calculated based on the current parameter estimation value;

[0076] (3) Update the center point, covariance matrix and weight coefficient of various vehicle driving states according to the posterior probability;

[0077] (4) Repeat steps (2) and (3) until the log-likelihood function reaches the convergence threshold or the algorithm reaches the maximum number of iterations.

[0078] Step S4: determining the route scene category corresponding to the same cumulative mileage according to the vehicle driving status corresponding to the same cumulative mileage in the historical operation data of multiple trips of the intelligent connected vehicle in a single day, that is, determining the scene category of the same position point on the fixed route; wherein the route scene categories are divided into “starting point, straight road, curve, intersection, platform”;

[0079] Preferably, the method of determining the route scenario category corresponding to the same cumulative mileage according to the vehicle driving state corresponding to the same cumulative mileage in the historical operation data of multiple shifts of the intelligent connected vehicle in a single day, that is, determining the scenario category of the same position point on the fixed route, also includes:

[0080] When the driving states of all vehicles corresponding to the same cumulative driving mileage are all "parked", the route scene category corresponding to the same cumulative driving mileage is determined to be "platform", that is, the scene category of the same position point on the fixed route is determined to be "platform"; wherein the first "platform" on the fixed route is classified as the "starting point";

[0081] When the driving states of all vehicles corresponding to the same cumulative driving mileage are all "straight driving", the route scene category corresponding to the same cumulative driving mileage is determined to be "straight road", that is, the scene category of the same position point on the fixed route is determined to be "straight road";

[0082] When the driving states of all vehicles corresponding to the same cumulative driving mileage are all "turning", it is determined that the route scene category corresponding to the same cumulative driving mileage is "curve", that is, the scene category of the same position point on the fixed route is determined to be "curve";

[0083] When all vehicle driving states corresponding to the same accumulated mileage include "parking" and "going straight", the route scene category corresponding to the same accumulated mileage is determined to be "intersection", that is, the scene category of the same position point on the fixed route is determined to be "intersection".

[0084] Step S5: using a voting mechanism to update the scene category of each location point on the fixed route to obtain a final scene category of each location point on the fixed route;

[0085] Preferably, the method of updating the scene category of each location point on the fixed route by using a voting mechanism to obtain a final scene category of each location point on the fixed route further includes:

[0086] The route range covered by a 30-meter length on a fixed route is taken as the interval window. When the scene categories of multiple location points in the same interval window are different, a voting mechanism is used to confirm the scene category of the central location point in the same interval window. If a tie occurs, the scene category of the central location point is updated to the scene category of the next location point in the same interval window.

[0087] Step S6: When the intelligent network-connected vehicle travels to each position point on the fixed route, an early warning of the operation safety risk of the intelligent network-connected vehicle is issued according to the final scene category of each position point on the fixed route.

[0088] Preferably, when the intelligent network-connected vehicle travels to each position point on a fixed route, the step of issuing an early warning of the operation safety risk of the intelligent network-connected vehicle according to the final scene category of each position point on the fixed route further includes:

[0089] When the intelligent networked vehicle drives to an "intersection" scene on a fixed route, the automatic driving control strategy of the intelligent networked vehicle is analyzed by comprehensively analyzing the braking, throttle opening and vehicle speed change trends 30 meters before the stop line of the intersection, and an early warning is issued when the automatic driving control strategy presents an operational safety risk;

[0090] When the intelligent networked vehicle travels to a "straight road" scenario on a fixed route, the average speed of the road section, the parking time and the lane selection are comprehensively analyzed to determine whether the intelligent networked vehicle has entered a minimum risk state, and an early warning is issued when the intelligent networked vehicle has not entered a minimum risk state;

[0091] It should be noted that when the average speed is less than 5km / h, the parking time is longer than 15 minutes, or the low-frequency lane is chosen, it means that the intelligent connected car has not entered the minimum risk state;

[0092] When the smart connected vehicle drives to a "platform" scene on a fixed route, the longitude and latitude information of the platform where the smart connected vehicle stops is integrated, and the variance value of the longitude and latitude information of the platform where the smart connected vehicle stops is analyzed. When the variance value is greater than a threshold, it is considered that the smart connected vehicle's ability to stop at the platform is weak and an early warning is issued;

[0093] The abnormal driving behavior of the intelligent networked vehicle is analyzed by combining the acceleration and longitude and latitude information of the intelligent networked vehicle on the entire fixed route; the abnormal driving behavior of the intelligent networked vehicle with an acceleration greater than 3m / s 2 Or less than -5m / s 2 It can also issue early warnings for sudden acceleration / deceleration situations, and for situations where the position of the smart connected vehicle deviates from the fixed route by more than 10.5 meters.

[0094] The method for early warning of operational safety risks of intelligent connected vehicles traveling on fixed routes provided by the present invention does not rely on prior information such as high-precision maps, vehicle driving routes, and operating shifts. It can automatically estimate the road section information only through the vehicle's one-day driving status data, including five road section area categories of "starting point, straight road, curve, intersection, and platform". Furthermore, the vehicle's real-time operating status data can be combined to make targeted analysis and judgment on the vehicle's operating safety status, which can effectively reduce the complexity of the judgment model and improve the real-time and accuracy of the judgment.

[0095] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technician familiar with this profession can make some changes or modifications to equivalent embodiments of equivalent changes using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the scope of the technical solution of the present invention.

Claims

1. A method for early warning of safety risks of operation of intelligent networked vehicles traveling on fixed routes, characterized in that: The following steps are involved: Step S1: when the intelligent network-connected vehicle is cyclically traveling along a fixed route, multiple frames of historical operation data of the intelligent network-connected vehicle in a single day are obtained; wherein each frame of historical operation data includes a timestamp, longitude and latitude, driving speed, driving acceleration, angular velocity, steering wheel angle, and accumulated driving mileage; Step S2: dividing the multi-frame historical operation data of the intelligent connected vehicle in a single day into the historical operation data of multiple shifts of the intelligent connected vehicle in a single day by a two-dimensional mean shift clustering algorithm; wherein the same accumulated mileage in the historical operation data of multiple shifts of the intelligent connected vehicle in a single day corresponds to the same position point on the fixed route; Step S3: Determine the vehicle driving state corresponding to each frame of historical operation data in the historical operation data of a single trip of the intelligent connected vehicle on a single day by using a three-dimensional Gaussian mixture model clustering algorithm, wherein the vehicle driving state is divided into "straight driving, turning, and parking"; Step S4: determining the route scene category corresponding to the same cumulative mileage according to the vehicle driving status corresponding to the same cumulative mileage in the historical operation data of multiple trips of the intelligent connected vehicle in a single day, that is, determining the scene category of the same position point on the fixed route; wherein the route scene categories are divided into "starting point, straight road, curve, intersection, platform"; Step S5: using a voting mechanism to update the scene category of each location point on the fixed route to obtain a final scene category of each location point on the fixed route; Step S6: When the intelligent network-connected vehicle travels to each position point on the fixed route, an early warning of the operation safety risk of the intelligent network-connected vehicle is issued according to the final scene category of each position point on the fixed route.

2. The method for early warning of safety risks of operation of intelligent networked vehicles traveling on fixed routes according to claim 1 is characterized in that: The step of obtaining multiple frames of historical operation data of the intelligent network-connected vehicle for a single day during the cyclic driving of the intelligent network-connected vehicle along a fixed route also includes: Downsampling the multi-frame historical operation data of the intelligent network-connected vehicle in a single day to obtain the multi-frame historical operation data of the intelligent network-connected vehicle after downsampling in a single day; wherein the number of frames of the historical operation data of the intelligent network-connected vehicle in a single day is reduced to 1 Hz by setting interval sampling; According to the correspondence between "timestamp-driving speed" in the multiple frames of historical operation data of the intelligent connected vehicle after downsampling on a single day, the historical operation data in which the driving speed is lower than 80% of the maximum driving speed of the day is eliminated.

3. The method for early warning of safety risks of operation of intelligent networked vehicles for fixed route driving according to claim 1, characterized in that: The method of dividing the multi-frame historical operation data of the intelligent network-connected vehicle in a single day into the multi-trip historical operation data of the intelligent network-connected vehicle in a single day by using a two-dimensional mean shift clustering algorithm also includes: Compare the timestamps of adjacent frames of historical operation data of the intelligent connected vehicle on a single day in sequence. If the timestamp difference between the current frame historical operation data and the next frame historical operation data of the intelligent connected vehicle on a single day is greater than 300 seconds for the first time, the current frame historical operation data is considered to be the last frame historical operation data of the first shift historical operation data of the intelligent connected vehicle on a single day. Then, the first shift operation time of the intelligent connected vehicle on a single day is calculated according to the timestamp of the last frame historical operation data in the first shift historical operation data and the timestamp of the first frame historical operation data. Half of the first shift operation time is used as the radius parameter of the two-dimensional mean shift clustering algorithm. The two-dimensional mean shift clustering algorithm is used to determine the geometric center of the historical operation data of subsequent shifts of the intelligent connected vehicle in a single day. Based on the timestamps corresponding to the geometric centers, the first frame of historical operation data and the last frame of historical operation data in the subsequent shifts are determined by the timestamp comparison method mentioned above, thereby dividing the historical operation data of multiple shifts of the intelligent connected vehicle in a single day.

4. The method for early warning of safety risks of operation of intelligent networked vehicles for fixed route driving according to claim 3 is characterized in that: The two-dimensional mean shift clustering algorithm comprises the following steps: (1) Initialize all frame historical running data as the class center point; (2) traverse each cluster center point, calculate the time dimension mean of all points within the radius parameter range, and replace the timestamp of the original cluster center point; (3) De-duplicate the cluster center points; (4) Repeat steps (2) and (3) until the number of cluster centers no longer changes.

5. The method for early warning of safety risks of operation of intelligent networked vehicles for fixed route driving according to claim 1, characterized in that: The method of determining the vehicle driving state corresponding to each frame of historical operation data of a single trip of a smart connected vehicle on a single day by using a three-dimensional Gaussian mixture model clustering algorithm also includes: By using the three-dimensional Gaussian mixture model clustering algorithm, according to the corresponding relationship of "driving speed-driving acceleration-angular velocity" of each frame of historical running data in the historical running data of a single trip of a smart connected vehicle on a single day, the vehicle driving state corresponding to each frame of historical running data in the historical running data of a single trip of a smart connected vehicle on a single day is determined; Among them, the calculation formula of the angular velocity ω in each frame of historical running data is: Among them, ω represents the angular velocity, v represents the center speed of the vehicle's rear axle, L represents the vehicle's wheelbase, and δ represents the front wheel deviation angle.

6. The method for early warning of safety risks of operation of intelligent networked vehicles traveling on fixed routes according to claim 5, characterized in that: The three-dimensional Gaussian mixture model clustering algorithm comprises the following steps: (1) Initialize the algorithm parameters: set the category to 3, the covariance matrix type to share the same general covariance matrix, the maximum number of iterations to 500, and the convergence threshold to 0.001; (2) For each data point, the posterior probability of the data point belonging to the three vehicle driving states of "straight driving, turning, and stopping" is calculated based on the current parameter estimation value; (3) Update the center point, covariance matrix and weight coefficient of various vehicle driving states according to the posterior probability; (4) Repeat steps (2) and (3) until the log-likelihood function reaches the convergence threshold or the algorithm reaches the maximum number of iterations.

7. The method for early warning of safety risks of operation of intelligent networked vehicles for fixed route driving according to claim 1, characterized in that: The method of determining the route scene category corresponding to the same cumulative mileage according to the vehicle driving state corresponding to the same cumulative mileage in the historical operation data of multiple shifts of the intelligent connected vehicle in a single day, that is, determining the scene category of the same position point on the fixed route, also includes: When the driving states of all vehicles corresponding to the same cumulative driving mileage are all "parked", the route scene category corresponding to the same cumulative driving mileage is determined to be "platform", that is, the scene category of the same position point on the fixed route is determined to be "platform"; wherein the first "platform" on the fixed route is classified as the "starting point"; When the driving states of all vehicles corresponding to the same cumulative driving mileage are "straight driving", the route scene category corresponding to the same cumulative driving mileage is determined to be "straight road", that is, the scene category of the same position point on the fixed route is determined to be "straight road"; When the driving states of all vehicles corresponding to the same cumulative driving mileage are "turning", it is determined that the route scene category corresponding to the same cumulative driving mileage is "curve", that is, the scene category of the same position point on the fixed route is determined to be "curve"; When all vehicle driving states corresponding to the same cumulative mileage include "parking" and "going straight", the route scene category corresponding to the same cumulative mileage is determined to be "intersection", that is, the scene category of the same position point on the fixed route is determined to be "intersection".

8. The method for early warning of safety risks of operation of intelligent networked vehicles traveling on fixed routes according to claim 1 is characterized in that: The method of updating the scene category of each location point on the fixed route by using a voting mechanism to obtain a final scene category of each location point on the fixed route also includes: The route range covered by a 30-meter length on a fixed route is taken as the interval window. When the scene categories of multiple location points in the same interval window are different, a voting mechanism is used to confirm the scene category of the central location point in the same interval window. If a tie occurs, the scene category of the central location point is updated to the scene category of the next location point in the same interval window.

9. The method for early warning of safety risks of operation of intelligent networked vehicles for fixed route driving according to claim 1, characterized in that: When the intelligent network-connected vehicle travels to each position point on the fixed route, the operation safety risk of the intelligent network-connected vehicle is warned according to the final scene category of each position point on the fixed route, and further includes: When the intelligent network-connected vehicle drives to the "intersection" scene on a fixed route, the automatic driving control strategy of the intelligent network-connected vehicle is analyzed by comprehensively analyzing the braking, throttle opening and vehicle speed change trend 30 meters before the stop line of the intersection, and an early warning is issued when the automatic driving control strategy has an operational safety risk; When the smart connected car drives to a "straight road" scenario on a fixed route, the average speed of the road section, the parking time and the lane selection are comprehensively analyzed to determine whether the smart connected car has entered a minimum risk state, and an early warning is issued when the smart connected car has not entered a minimum risk state; When the smart connected vehicle drives to a "platform" scene on a fixed route, the longitude and latitude information of the platform where the smart connected vehicle stops is integrated, and the variance value of the longitude and latitude information of the platform where the smart connected vehicle stops is analyzed. When the variance value is greater than a threshold, it is considered that the smart connected vehicle's ability to stop at the platform is weak and an early warning is issued; The abnormal driving behavior of the intelligent networked vehicle is analyzed by combining the acceleration and longitude and latitude information of the intelligent networked vehicle on the entire fixed route; the abnormal driving behavior of the intelligent networked vehicle with an acceleration greater than 3m / s 2 Or less than -5m / s 2 It can also issue warnings for sudden acceleration / deceleration situations, and for situations where the position of the smart connected vehicle deviates from the fixed route by more than 10.5 meters.

Images