Method, device and vehicle end control equipment for predicting speed of preceding vehicle in platoon

Abstract

The application relates to a method and device for predicting the speed of a preceding vehicle in a car-following vehicle group and a vehicle end control device, and belongs to the technical field of vehicles. The method comprises the following steps: in the case that the ego vehicle and at least two preceding vehicles which follow and drive in the same direction establish communication respectively, acquiring a real-time vehicle speed set of each preceding vehicle for a period of time; inputting the real-time vehicle speed set into a vehicle speed prediction model to obtain an output neighboring vehicle speed first prediction result; the neighboring vehicle speed first prediction result comprises a plurality of preliminary prediction speeds; the neighboring vehicle is the closest preceding vehicle in front of the ego vehicle in the driving direction; a preset compensation value is acquired and used to process the neighboring vehicle speed first prediction result to obtain a neighboring vehicle speed second prediction result; the preset compensation value is determined based on a historical acceleration error value, a historical impact error value and a historical feature point error value of a historical car-following vehicle group; the real-time vehicle speed set is processed through the vehicle speed prediction model, and the processing result is corrected based on the preset compensation value, so that accurate long-time prediction of the future speed of the preceding vehicle in the target lane is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of vehicles, in particular to a method and device for predicting the speed of a preceding vehicle in a car-following vehicle group and a vehicle-end control device. BACKGROUND

[0002] With the development of technology, people have higher and higher requirements for the economy, comfort and safety of automobiles. However, in actual roads, especially in congested urban roads, due to the great difference in driving ability of each person, and limited by the vision range of human drivers, the driver often cannot accurately predict the motion behavior of the front vehicle, which easily leads to difficulties in the process of following the vehicle, such as the rear vehicle accelerating and decelerating rapidly due to the non-standard driving habits of the leading vehicle, thereby causing the phenomenon of vehicle shock wave. It is very unfavorable for driving safety and greatly reduces the fuel economy and driving comfort of the automobile. Therefore, it is very necessary to predict the speed of the preceding vehicle. SUMMARY

[0003] Therefore, it is necessary to provide a method and device for predicting the speed of a preceding vehicle in a car-following vehicle group, a vehicle-end control device and a computer readable storage medium, which can accurately predict the speed of the preceding vehicle.

[0004] In a first aspect, the present application provides a method for predicting the speed of a preceding vehicle in a car-following vehicle group, comprising:

[0005] In the case that the ego vehicle and at least two preceding vehicles driving in a car-following manner in front of the ego vehicle in the driving direction of the ego vehicle establish communication respectively, obtaining real-time vehicle speeds of each of the at least two preceding vehicles in a period of time, to obtain a real-time vehicle speed set;

[0006] inputting the real-time vehicle speed set into a pre-trained vehicle speed prediction model to obtain a first prediction result of a neighboring vehicle speed output by the vehicle speed prediction model; wherein the first prediction result of the neighboring vehicle speed includes a plurality of preliminary prediction vehicle speeds of the neighboring vehicle in a future period; the neighboring vehicle is the preceding vehicle closest to the ego vehicle in front of the ego vehicle in the driving direction of the ego vehicle;

[0007] obtaining a preset compensation value, processing the first prediction result of the neighboring vehicle speed based on the preset compensation value to obtain a second prediction result of the neighboring vehicle speed; wherein the preset compensation value is determined based on historical acceleration error values, historical impact error values and historical feature point error values corresponding to historical car-following vehicle groups.

[0008] In one embodiment, the method further comprises:

[0009] obtaining a neighbor vehicle speed history prediction result corresponding to a historical vehicle speed set; the historical vehicle speed set comprises historical vehicle speeds of each of the at least two preceding vehicles in communication with the ego vehicle within a historical time period; the neighbor vehicle speed history prediction result is obtained by taking the historical vehicle speed set as model input and through output of the pre-trained vehicle speed prediction model;

[0010] obtaining a neighbor vehicle history actual speed associated with each of the plurality of historical initial prediction speeds in the neighbor vehicle speed history prediction result, and calculating an error value between each of the neighbor vehicle history actual speeds and the corresponding historical initial prediction speed;

[0011] calling a target function to process each of the historical acceleration error values, each of the historical jerk error values, and the historical feature point error value, to obtain a minimum value of a matching error in the matching errors;

[0012] processing the minimum value of the matching error, and the maximum value and the minimum value in the error values, through an optimal solution algorithm, to determine the preset compensation value.

[0013] In one of the embodiments, the method further comprises:

[0014] differentially processing each of the neighbor vehicle history actual speeds to obtain corresponding first acceleration features;

[0015] differentially processing the historical initial prediction speed corresponding to each of the neighbor vehicle history actual speeds to obtain corresponding second acceleration features;

[0016] determining a historical acceleration error value corresponding to each of the neighbor vehicle history actual speeds based on each of the first acceleration features and the second acceleration features.

[0017] In one of the embodiments, after obtaining each of the first acceleration features and each of the second acceleration features, the method further comprises:

[0018] differentially processing each of the first acceleration features and each of the second acceleration features to obtain corresponding first jerk features and second jerk features;

[0019] determining a historical jerk error value corresponding to each of the neighbor vehicle history actual speeds based on each of the first jerk features and each of the second jerk features.

[0020] In one of the embodiments, the obtaining of the neighbor vehicle history actual speed associated with each of the plurality of historical initial prediction speeds in the neighbor vehicle speed history prediction result comprises: obtaining a neighbor vehicle history actual speed associated with K historical initial prediction speeds in the neighbor vehicle speed history prediction result; wherein K≥3 and K is an integer.

[0021] After the first acceleration features and the second acceleration features are obtained, the method further comprises:

[0022] The product of the first acceleration feature corresponding to the second acceleration feature and the first acceleration feature corresponding to the next time point is obtained, and the first number of positive numbers, the zero and the second number of negative numbers in the product are obtained;

[0023] The historical turning point feature is determined based on the first number and the second number, and the corresponding historical feature point error value is obtained by taking the opposite of the historical turning point feature.

[0024] In one embodiment, the method further comprises:

[0025] The preliminary prediction speed in the first prediction result of the adjacent vehicle speed is added to the preset compensation value to obtain the second prediction result of the adjacent vehicle speed.

[0026] In one embodiment, the method further comprises:

[0027] The GA-BP neural network model is trained based on the historical speed data of the historical car-following vehicle group to obtain a pre-trained speed prediction model.

[0028] In a second aspect, the application also provides a device for predicting the speed of a preceding vehicle in a car-following vehicle group, the device comprising:

[0029] A data acquisition module is configured to acquire real-time speeds of each of at least two preceding vehicles in a time period when the ego vehicle and the at least two preceding vehicles are in communication, to obtain a real-time speed set;

[0030] A data processing module is configured to input the real-time speed set into a pre-trained speed prediction model to obtain a first prediction result of adjacent vehicle speed output by the speed prediction model; wherein the first prediction result of adjacent vehicle speed includes a plurality of preliminary prediction speeds of the adjacent vehicle in a future time period; the adjacent vehicle is the closest preceding vehicle in front of the ego vehicle in the ego vehicle's driving direction;

[0031] A result correction module is configured to obtain a preset compensation value, process the first prediction result of adjacent vehicle speed based on the preset compensation value, and obtain a second prediction result of adjacent vehicle speed; wherein the preset compensation value is determined based on historical acceleration error values, historical impact error values and historical feature point error values corresponding to a historical car-following vehicle group.

[0032] In a third aspect, the present application also provides a vehicle-side control device, comprising a memory and a processor, the memory storing a computer program, and the processor implementing the following steps when executing the computer program:

[0033] In a case where the ego vehicle and at least two preceding vehicles in front thereof in a driving direction thereof are respectively in communication, real-time vehicle speeds of each of the at least two preceding vehicles in a time period are obtained, to obtain a real-time vehicle speed set;

[0034] The real-time vehicle speed set is input into a pre-trained vehicle speed prediction model, to obtain a first prediction result of a neighboring vehicle speed output by the vehicle speed prediction model; wherein the first prediction result of the neighboring vehicle speed comprises a plurality of preliminary prediction vehicle speeds of the neighboring vehicle in a future time period; the neighboring vehicle is the preceding vehicle closest to the ego vehicle in the driving direction thereof;

[0035] A preset compensation value is obtained, and the first prediction result of the neighboring vehicle speed is processed based on the preset compensation value, to obtain a second prediction result of the neighboring vehicle speed; wherein the preset compensation value is determined based on historical acceleration error values, historical impact error values and historical feature point error values corresponding to a historical platoon.

[0036] In a fourth aspect, the present application also provides a computer-readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the following steps:

[0037] In a case where the ego vehicle and at least two preceding vehicles in front thereof in a driving direction thereof are respectively in communication, real-time vehicle speeds of each of the at least two preceding vehicles in a time period are obtained, to obtain a real-time vehicle speed set;

[0038] The real-time vehicle speed set is input into a pre-trained vehicle speed prediction model, to obtain a first prediction result of a neighboring vehicle speed output by the vehicle speed prediction model; wherein the first prediction result of the neighboring vehicle speed comprises a plurality of preliminary prediction vehicle speeds of the neighboring vehicle in a future time period; the neighboring vehicle is the preceding vehicle closest to the ego vehicle in the driving direction thereof;

[0039] A preset compensation value is obtained, and the first prediction result of the neighboring vehicle speed is processed based on the preset compensation value, to obtain a second prediction result of the neighboring vehicle speed; wherein the preset compensation value is determined based on historical acceleration error values, historical impact error values and historical feature point error values corresponding to a historical platoon.

[0040] The method, device, vehicle terminal control equipment and computer readable storage medium for predicting the speed of a preceding vehicle in a platoon of vehicles provide a method for predicting the speed of a preceding vehicle in a platoon of vehicles, which comprises: obtaining real-time speeds of each of at least two preceding vehicles in a time period, to obtain a set of real-time speeds, in the case that the at least two preceding vehicles are in communication with the ego vehicle and are platooning in front of the ego vehicle in the direction of travel of the ego vehicle; inputting the set of real-time speeds into a pre-trained speed prediction model to obtain a first prediction result of the speed of a neighboring vehicle output by the speed prediction model; wherein the first prediction result of the speed of the neighboring vehicle comprises a plurality of preliminary predicted speeds of the neighboring vehicle in a future time period; the neighboring vehicle is the preceding vehicle closest to the ego vehicle in front of the ego vehicle in the direction of travel of the ego vehicle; obtaining a preset compensation value, and processing the first prediction result of the speed of the neighboring vehicle based on the preset compensation value to obtain a second prediction result of the speed of the neighboring vehicle; wherein the preset compensation value is determined based on a historical acceleration error value, a historical impact error value and a historical feature point error value corresponding to a historical platoon of vehicles. The set of real-time speeds of the platoon of vehicles in front of the ego vehicle is processed using the pre-trained speed prediction model to obtain the first prediction result of the speed of the neighboring vehicle for the neighboring vehicle platooning in front of the ego vehicle, and then the plurality of preliminary predicted speeds in the first prediction result of the speed of the neighboring vehicle are respectively corrected based on the preset compensation value determined based on the historical acceleration error value, the historical impact error value and the historical feature point error value corresponding to the historical platoon of vehicles, to obtain the second prediction result of the speed of the neighboring vehicle for the neighboring vehicle in front, which is more accurate. In this way, long-term accurate prediction of the future speed of the preceding vehicle in the target lane is realized, which is beneficial to adjusting the speed of the ego vehicle in advance based on the prediction of the future speed of the neighboring vehicle in front, and provides a basis for improving the energy efficiency, safety and comfort performance of vehicles on a congested road section. BRIEF DESCRIPTION OF DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related art, the drawings needed to be used in the description of the embodiments of the present application or the related art will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other related drawings can be obtained by those skilled in the art without any creative effort.

[0042] Figure 1 A flowchart of a method for predicting the speed of a preceding vehicle in a platoon of vehicles in an embodiment;

[0043] Figure 2 A flowchart of a method for determining a preset compensation value in an embodiment;

[0044] Figure 3 A flowchart of a method for predicting the speed of a preceding vehicle in a platoon of vehicles in another embodiment;

[0045] Figure 4 A schematic diagram of a model fitness decline curve in an embodiment;

[0046] Figure 5 A schematic diagram of a long-time prediction effect diagram of a test set of a training model in an embodiment;

[0047] Figure 6 A structural schematic diagram of a prediction device for a front vehicle speed in a car-following vehicle platoon in an embodiment;

[0048] Figure 7 An internal structural diagram of a vehicle end control device in an embodiment. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical scheme and advantages of the present application clearer, further detailed description will be made to the present application in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not to limit the present application.

[0050] In recent years, with the development of intelligent network connection technology, many scholars have also proposed corresponding methods for predicting the speed of the front vehicle, but most of them focus on how to predict the short-term speed of the vehicle. For example, a short-term speed working condition real-time prediction method based on the interaction between the front vehicle (a front vehicle adjacent to the front of the vehicle) and the vehicle itself is provided. By obtaining the historical speed and distance information of the vehicle and the front vehicle, a model is constructed based on an artificial neural network to realize short-term prediction of the vehicle speed. It can be seen that most of the existing vehicle speed prediction methods focus on short-term prediction of vehicle speed, but the problem of long-term prediction has not been well solved, and the prediction accuracy needs to be improved.

[0051] In an exemplary embodiment, as shown in Figure 1 A method for predicting the speed of a front vehicle in a car-following vehicle platoon is provided, which can be applied to a vehicle, a server corresponding to the vehicle, and a system including the vehicle and the corresponding server, and is realized through the interaction of the vehicle and the server. The present application takes the application of the method to the vehicle as an example for illustration, including the following steps 101 to 103. Among them:

[0052] Step 101, in the case that the self-vehicle and at least two front vehicles following in the front direction of the self-vehicle establish communication respectively, obtaining the real-time speed of each front vehicle in the at least two front vehicles within a period of time, to obtain a real-time speed set.

[0053] It should be noted that the self-vehicle can be the current vehicle or the present vehicle, and can also be a data collection vehicle in a car-following vehicle platoon. Car-following means that several vehicles are arranged to follow one another. The vehicle platoon formed by the car-following vehicles does not include other vehicles. It should also be noted that the vehicle platoon formed by the car-following vehicles and the self-vehicle is located on the same lane.

[0054] In the vehicle platoon formed by the vehicles following the vehicle, the leading vehicle in the driving direction of the vehicle platoon can be the head vehicle, and at least one vehicle can be further included between the ego vehicle and the head vehicle; that is, at least two vehicles following the vehicle are included in front of the ego vehicle in the driving direction of the vehicle platoon. Each vehicle in the vehicle platoon can be configured to have a communication function with the ego vehicle, so that the ego vehicle can obtain some driving information of other following vehicles in the vehicle platoon based on the communication function. The driving information herein includes, for example, the distance between the other following vehicles and the ego vehicle, the driving speed of the other following vehicles, whether the other following vehicles are in the braking state, etc. The specific content of the driving information is not limited in the present application, and the specific information included in the driving information can be set according to the needs.

[0055] In addition, on the basis that the ego vehicle and other following vehicles in the vehicle platoon have established communication connections, the other following vehicles in the vehicle platoon can also be selected to further establish communication connections with each other, which is only an optional embodiment provided by the present application, but the present application does not make specific limitations thereon.

[0056] It should be further supplemented that in a vehicle platoon including the ego vehicle, if the head vehicle drives away from the lane where the ego vehicle is located, the front vehicle in front of the original head vehicle will be updated as the new head vehicle in the driving direction of the ego vehicle, so that the number of vehicles in the vehicle platoon including the ego vehicle remains unchanged. Correspondingly, if any other vehicle in front of the ego vehicle drives away from the lane where the ego vehicle is located, the front vehicle in front of the original head vehicle will be updated as the new head vehicle, and the original head vehicle will be updated as the front vehicle between the new head vehicle and the ego vehicle, so that the number of vehicles in the vehicle platoon including the ego vehicle remains unchanged.

[0057] In the case that the ego vehicle and a plurality of platoon vehicles following and driving in front of the ego vehicle in the driving direction have established communication connections, step 101 can be performed to obtain the real-time vehicle speeds of the corresponding plurality of front vehicles in a period of time based on the ego vehicle, so as to obtain a real-time vehicle speed set including the vehicle speeds of the plurality of front vehicles at a plurality of time points. The real-time vehicle speed set is, for example, a real-time vehicle speed matrix, which includes a plurality of rows and a plurality of columns. Each row of data in the real-time vehicle speed matrix is the real-time vehicle speed of a corresponding front vehicle at each time point, and each column of data in the real-time vehicle speed matrix includes the real-time vehicle speeds of each front vehicle at the same time point. Optionally, the plurality of real-time vehicle speeds in a row are sequentially arranged from left to right in the order of corresponding time, and the plurality of real-time vehicle speeds in a column are sequentially arranged from top to bottom in the order of distance between the front vehicle and the ego vehicle; that is, the first row of the real-time vehicle speed matrix is the real-time vehicle speeds of the head vehicle at a plurality of time points, and the last row of the real-time vehicle speed matrix is the real-time vehicle speeds of the front vehicle adjacent to the ego vehicle at a plurality of time points.

[0058] It also needs to be added that for the collection of the speed of each vehicle in the team vehicle, the corresponding time is the same, and as for the interval length between two adjacent times, the present application does not make specific limitation; for example, the interval length between any two adjacent times can be selected to be the same.

[0059] It also needs to be added that for the communication connection between the ego vehicle and the front vehicle, the present application does not make specific limitation, for example, V2V (Vehicle to Vehicle, vehicle team communication technology) communication can be selected to realize the communication connection between the ego vehicle and the front vehicle, and of course other communication modes can also be selected.

[0060] Step 102, input the real-time vehicle speed set into the pre-trained vehicle speed prediction model to obtain the first prediction result of the adjacent vehicle speed output by the vehicle speed prediction model; wherein the first prediction result of the adjacent vehicle speed includes multiple preliminary prediction speeds of the adjacent vehicle in the future period; the adjacent vehicle is the front vehicle closest to the ego vehicle in the forward direction of the ego vehicle.

[0061] Among them, the above-mentioned adjacent vehicle is the front vehicle closest to the ego vehicle in the forward direction of the ego vehicle, which can also be explained as the adjacent vehicle is the front vehicle adjacent to the ego vehicle in the forward direction of the ego vehicle.

[0062] In the case of obtaining the real-time vehicle speed set based on step 101, step 102 can be further executed, and the pre-trained vehicle speed prediction model is used to process the collected real-time vehicle speed set to obtain the first prediction result of the future speed of the adjacent vehicle output by the vehicle speed prediction model (the first prediction result of the adjacent vehicle speed). Using the pre-trained vehicle speed prediction model to process the real-time vehicle speed set of the platoon in front of the ego vehicle is beneficial to guarantee the real-time and efficiency of data processing, and the pre-trained vehicle speed prediction model can also be continuously trained under demand, so that the first prediction result of the speed of the adjacent vehicle (front vehicle) in front of the ego vehicle output by the vehicle speed prediction model is more accurate.

[0063] In the embodiments provided by the present application, the first prediction result of the speed of the adjacent vehicle includes preliminary prediction speeds of the adjacent vehicle at multiple time points in a future period of time, and does not include prediction speeds of the following vehicle with respect to the remaining following vehicles, but this is only one optional implementation provided by the present application; the present application is not limited thereto, and in the case of demand, the result output by the pre-trained speed prediction model after processing can also be selected to include multiple preliminary prediction speeds of multiple vehicles in the following vehicle team in a future period of time. The present application does not limit the number of preliminary prediction speeds included in the first prediction result of the speed of the adjacent vehicle, and the number of data output by the speed prediction model can be set according to demand. In addition, the interval time of the data output by the speed prediction model can also be set according to demand.

[0064] In step 103, a preset compensation value is obtained, and the first prediction result of the speed of the adjacent vehicle is processed based on the preset compensation value to obtain a second prediction result of the speed of the adjacent vehicle; wherein the preset compensation value is determined based on the historical acceleration error value, the historical impact error value and the historical feature point error value corresponding to the historical following vehicle team.

[0065] In order to further improve the prediction accuracy of the future speed of the adjacent vehicle (front vehicle) in front of the ego vehicle, on the basis of obtaining the first prediction result of the speed of the adjacent vehicle in step 102, step 103 can be further performed to correct the first prediction result of the speed of the adjacent vehicle by using the preset compensation value determined based on the historical acceleration error value, the historical impact error value and the historical feature point error value corresponding to the historical following vehicle team, so that the accuracy of the second prediction result of the speed of the adjacent vehicle obtained after correction is higher. The preset compensation value is used to correct each preliminary prediction speed in the first prediction result of the speed of the adjacent vehicle, and the target prediction speed obtained by correcting the preliminary prediction speed constitutes the second prediction result of the speed of the adjacent vehicle.

[0066] It can be seen that, by using the pre-trained vehicle speed prediction model to process the real-time vehicle speed set of the platooning vehicle in front of the ego vehicle, the first prediction result of the adjacent vehicle speed of the adjacent vehicle in front of the ego vehicle is obtained, and then the preset compensation value determined based on the historical acceleration error value, the historical impact error value and the historical feature point error value corresponding to the historical platooning vehicle is used to correct the multiple preliminary prediction speeds in the first prediction result of the adjacent vehicle speed, so that the second prediction result of the adjacent vehicle speed more accurate to the adjacent vehicle in front is obtained. In this way, the long-term accurate prediction of the future speed of the vehicle in front of the target lane is realized, which is beneficial to adjusting the speed of the ego vehicle in advance according to the prediction of the future speed of the adjacent vehicle in front, and provides a basis for improving the energy efficiency, safety and comfort performance of vehicles on crowded road sections. That is, it needs to be emphasized that the present application is not based on the real-time speed information of the first adjacent vehicle in front of the ego vehicle to predict the future speed of the first adjacent vehicle in front of the ego vehicle, but based on the real-time speed information of multiple vehicles in front of the ego vehicle to predict the future speed of the first adjacent vehicle in front of the ego vehicle.

[0067] Please combine Figure 1 Refer to Figure 2 In one exemplary embodiment, the present application also provides a method for determining a preset compensation value, which can include steps 201-204, wherein:

[0068] Step 201, obtaining the historical prediction result of the adjacent vehicle speed corresponding to the historical vehicle speed set; the historical vehicle speed set includes the historical vehicle speed of each vehicle in front of the ego vehicle in communication with the ego vehicle within a historical time period; the historical prediction result of the adjacent vehicle speed is obtained by inputting the historical vehicle speed set into the output of the pre-trained vehicle speed prediction model;

[0069] Step 202, obtaining the adjacent vehicle historical actual speed associated with the multiple historical preliminary prediction speeds in the historical prediction result of the adjacent vehicle speed, and calculating the error value between each adjacent vehicle historical actual speed and the corresponding historical preliminary prediction speed;

[0070] Step 203, calling a target function to process each historical acceleration error value, each historical impact error value and the historical feature point error value to obtain the minimum value of the matching error in the matching error;

[0071] Step 204, processing the minimum value of the matching error, and the maximum value and the minimum value in the error value by an optimal solution algorithm to determine the preset compensation value.

[0072] Specifically, the preset compensation value is determined based on historical vehicle information, that is, before predicting the future vehicle speed of the adjacent vehicle in front of the ego vehicle, some preliminary work is needed to realize the prediction function and to ensure the accuracy of the prediction results. For example, the historical vehicle speed set corresponding to the historical prediction result of the adjacent vehicle speed can be obtained by performing step 201. Specifically, the historical vehicle speed set can be obtained by obtaining the historical vehicle speed of each preceding vehicle in the platoon communicating with the ego vehicle, that is, obtaining the historical vehicle speed of each preceding vehicle in at least one historical time period, thereby obtaining at least one historical vehicle speed set. Then, the historical vehicle speed set is processed based on the pre-trained vehicle speed prediction model to obtain the output historical prediction result of the adjacent vehicle speed. The historical prediction result of the adjacent vehicle speed includes a plurality of historical initial prediction speeds of the future prediction speed of the adjacent vehicle. Then, by performing step 202, the historical actual speed of the adjacent vehicle corresponding to all historical initial prediction speeds in the historical prediction result of the adjacent vehicle speed is obtained. By comparing each historical actual speed of the adjacent vehicle with the corresponding historical initial prediction speed, the accuracy of the historical prediction result of the adjacent vehicle speed can be determined, that is, the difference (error value) between the historical prediction result of the adjacent vehicle speed and the actual situation can be determined. Then, step 203 can be performed to call the target function to process each historical acceleration error value, each historical impact error value, and the historical feature point error value of the historical platoon to obtain the minimum value of the matching error in the corresponding matching error. Finally, step 204 is performed to process the minimum value of the matching error obtained in step 203, and the maximum and minimum values of the error values obtained in step 202 based on the optimal solution algorithm, thereby determining the preset compensation value.

[0073] It should be noted that the historical time period and the corresponding road for which the historical vehicle speed is collected are not specifically limited in the present application. For example, different preset compensation values can be used for different time periods, different roads, and / or different road sections. The steps 201-204 are only one way to determine the preset compensation value provided by the present application, but the present application is not limited thereto.

[0074] Although the predicted result (the first prediction result of the adjacent vehicle speed) can well reflect the traffic flow characteristics, the reaction time has a large uncertainty due to the different vehicle models and drivers of the preceding vehicle (the adjacent vehicle) of the host vehicle, and therefore the time and space misalignment error easily occurs. Therefore, the initial prediction speed in the first prediction result of the adjacent vehicle speed is corrected by the preset compensation value determined in advance, which is beneficial to improve the prediction accuracy of the future vehicle speed of the adjacent vehicle, so that the ego vehicle can know the future vehicle speed of the preceding adjacent vehicle in advance, thereby adjusting the driving speed of the ego vehicle in advance, which is beneficial to reduce the occurrence of traffic accidents between the platoon vehicles, thereby improving the driving safety of the vehicle.

[0075] Further, the present application is based on the driving information of multiple preceding vehicles driving in front of the subject vehicle to predict the future vehicle speed of the neighboring vehicle in front of the subject vehicle, rather than only based on the driving information of the neighboring vehicle in front of the subject vehicle to predict the future vehicle speed of the neighboring vehicle, and further combined with a preset compensation value obtained based on historical driving information in the prediction process. Compared with only based on the driving information of the neighboring vehicle in front of the subject vehicle to predict the future vehicle speed of the neighboring vehicle, the information factor for predicting the future vehicle speed of the neighboring vehicle is increased, thereby being beneficial to improve the prediction accuracy of the future vehicle speed of the neighboring vehicle, and also being able to avoid the vehicle shock wave phenomenon caused by the sudden acceleration and deceleration of the following vehicle due to the non-standard driving habits of the leading vehicle, thereby being beneficial to improve the driving safety of the subject vehicle.

[0076] Please refer to Figure 1 and Figure 2 In one exemplary embodiment, the present application also provides a method for determining a historical acceleration error value, which can specifically include steps 205-207 after obtaining the neighboring vehicle historical actual speeds associated with the multiple historical preliminary predicted vehicle speeds in the neighboring vehicle speed historical prediction result by performing step 202, wherein:

[0077] Step 205, differentiating each neighboring vehicle historical actual speed to obtain corresponding first acceleration characteristics;

[0078] Step 206, differentiating the historical preliminary predicted vehicle speed corresponding to each neighboring vehicle historical actual speed to obtain corresponding second acceleration characteristics;

[0079] Step 207, determining the historical acceleration error value corresponding to each neighboring vehicle historical actual speed based on the first acceleration characteristics and the second acceleration characteristics.

[0080] Specifically, for example, the first acceleration characteristics corresponding to each neighboring vehicle historical actual speed can be obtained by differentiating each neighboring vehicle historical actual speed; the second acceleration characteristics corresponding to each neighboring vehicle historical actual speed can be obtained by differentiating the historical preliminary predicted vehicle speed corresponding to each neighboring vehicle historical actual speed, and then the historical acceleration error value corresponding to each neighboring vehicle historical actual speed can be determined by calculating the difference between each corresponding matched first acceleration characteristics and second acceleration characteristics.

[0081] Herein, a calculation formula of the acceleration characteristics A(i) is provided, for example: wherein represents the neighboring vehicle historical actual speed at a certain (i) moment, or the historical preliminary predicted vehicle speed at a certain (i) moment; A(i) is the first acceleration characteristics corresponding to the neighboring vehicle historical actual speed at a certain (i) moment, or the second acceleration characteristics corresponding to the historical preliminary predicted vehicle speed at a certain (i) moment.

[0082] Further provided is a calculation formula of the historical acceleration error value δ A (i) = A real (i) - A predict (i), wherein A real (i) represents the first acceleration feature corresponding to the historical actual vehicle speed of the adjacent vehicle at the (i) moment, A predict (i) represents the second acceleration feature corresponding to the historical initial prediction vehicle speed at the (i) moment.

[0083] Please refer to Figure 1 and Figure 2 In an exemplary embodiment, after obtaining the first acceleration features and the second acceleration features, the method further comprises:

[0084] differentially processing the first acceleration features and the second acceleration features to obtain corresponding first jerk features and second jerk features;

[0085] determining the historical jerk error values corresponding to the historical actual vehicle speeds of the adjacent vehicles based on the first jerk features and the second jerk features.

[0086] Specifically, for example, the first jerk features corresponding to the historical actual vehicle speeds of the adjacent vehicles can be obtained by differentially processing the first acceleration features, and the second jerk features corresponding to the historical initial prediction vehicle speeds can be obtained by differentially processing the second acceleration features, and then the historical jerk error values corresponding to the historical actual vehicle speeds of the adjacent vehicles can be determined by calculating the difference between each corresponding matched first jerk feature and second jerk feature.

[0087] Herein, a calculation formula of the jerk feature G(i) is provided, for example: G(i) is the first jerk feature corresponding to the historical actual vehicle speed of the adjacent vehicle at the (i) moment, or the second jerk feature corresponding to the historical initial prediction vehicle speed at the (i) moment.

[0088] Further provided is a calculation formula of the historical jerk error value δ G (i) of the historical jerk error value δ G (i) = G real (i) - G predict (i), wherein G real (i) represents the first jerk feature corresponding to the historical actual vehicle speed of the adjacent vehicle at the (i) moment, G predict (i) represents the second jerk feature corresponding to the historical initial prediction vehicle speed at the (i) moment.

[0089] Please refer to Figure 1 and Figure 2In an example embodiment, the step 202 performs the step of obtaining the historical actual vehicle speed of the neighboring vehicle associated with the plurality of historical preliminary predicted vehicle speeds in the historical prediction result of the vehicle speed of the neighboring vehicle, for example, obtaining the historical actual vehicle speed of the neighboring vehicle associated with K historical preliminary predicted vehicle speeds in the historical prediction result of the vehicle speed of the neighboring vehicle; wherein K≥3 and K is an integer.

[0090] After obtaining the first acceleration features and the second acceleration features, further comprising:

[0091] obtaining the product of the first acceleration feature of the previous time and the first acceleration feature of the next time corresponding to each of the second to K-1th first acceleration features, and obtaining the first number of positive numbers, the second number of zeros and negative numbers in the product;

[0092] determining the historical turning point feature based on the first number and the second number, and obtaining the corresponding historical feature point error value by taking the opposite of the historical turning point feature.

[0093] Specifically, for example, the product of the first acceleration feature of the previous time and the first acceleration feature of the next time corresponding to each of the second to K-1th first acceleration features can be extracted, and then the product of the extracted first acceleration feature of the previous time and the first acceleration feature of the next time corresponding to each of the second to K-1th first acceleration features can be calculated, and the number of positive numbers, zeros and negative numbers in the product can be recorded, and the total number of positive numbers is taken as the first number and the total number of negative numbers and zeros is taken as the second number; wherein the first number and the second number obtained at this time are the historical turning point feature corresponding to the historical actual vehicle speed of the neighboring vehicle, and the corresponding historical feature point error value can be obtained by taking the opposite of the historical turning point feature.

[0094] Here, a calculation formula of the historical turning point feature P(i) is provided as follows: wherein the turning point feature condition 1 is that the product of the accelerations at the i-1th and i+1th time is greater than zero at the i th time, and the total number of all times satisfying the condition 1 is taken as the first number m1; the turning point feature condition 2 is that the product of the accelerations at the i-1th and i+1th time is less than or equal to zero at the i th time, and the total number of all times satisfying the condition 2 is taken as the second number m2. Further, a calculation formula of the historical feature point error value δ P (i) is provided as follows: P (i)=-P(i).

[0095] The objective function provided in the present application can be, for example, Wherein, i = 0 indicates that the first historical initial predicted vehicle speed in the plurality of historical initial predicted vehicle speeds corresponds to a time, or in other words, the adjacent vehicle historical actual vehicle speed corresponding to the plurality of historical initial predicted vehicle speeds corresponds to a time, T is the entire data period corresponding to the plurality of historical initial predicted vehicle speeds.

[0096] In an exemplary embodiment, the step 103 performs the processing of the first prediction result of the adjacent vehicle speed based on the preset compensation value to obtain the second prediction result of the adjacent vehicle speed, which can be executed as: adding each initial predicted vehicle speed in the first prediction result of the adjacent vehicle speed and the preset compensation value to obtain the second prediction result of the adjacent vehicle speed.

[0097] That is, the plurality of future predicted vehicle speeds of the adjacent vehicle included in the second prediction result of the adjacent vehicle speed are obtained based on the addition of each initial predicted vehicle speed in the first prediction result and the preset compensation value, so as to realize the correction of the first prediction result of the adjacent vehicle speed based on the preset compensation value, thereby improving the prediction accuracy of the future vehicle speed of the adjacent vehicle.

[0098] In an exemplary embodiment, it further includes: training the GA-BP neural network model based on the historical vehicle speed data of the historical car-following platoon to obtain a pre-trained vehicle speed prediction model. Wherein, GA is the abbreviation of Genetic Algorithm, and the Chinese translation is genetic algorithm; BP is the abbreviation of Back Propagation, and the Chinese translation is neural network.

[0099] That is, the pre-trained vehicle speed prediction model used in the present application can be obtained based on the training of the GA-BP neural network model, and the sample data used for training includes the historical vehicle speed data of the historical car-following platoon.

[0100] For the problem of how to predict the future vehicle speed of the adjacent vehicle in the front vehicle in the car-following platoon proposed in the application, the present application further provides an alternative embodiment, which specifically provides a long-time vehicle speed prediction method based on V2V as shown in Figure 3 First, the driving data set collected on urban roads is obtained and processed through V2V, then the long-time vehicle speed prediction model based on GA optimized BP neural network is established, and the model is trained. Then online prediction is performed, and finally feature matching parameters are used to eliminate the space-time misplacement error.

[0101] The long-time vehicle speed prediction method based on the V2V environment comprises the following steps S21-S24. In step S21, the V2V obtained vehicle driving data set is acquired and processed to analyze the relationship between the speed characteristics of the front and rear vehicles in the following vehicle platoon. In step S22, a GA-BP neural network long-time vehicle speed prediction model is established, and the neural network model is trained. In step S23, the trained BP neural network model is used to perform online prediction on the real-time collected test data set. In step S24, the preliminary misplacement error is evaluated, and the feature matching parameters are used to eliminate the misplacement error to finally realize the relatively accurate prediction of the future long-time vehicle speed sequence.

[0102] Further, based on the steps S21-S24 provided above, the application also provides an alternative embodiment as follows:

[0103] In step S21, the following vehicle speed information on the urban road is acquired. The vehicle platoon information with a quantity of N in each time period per day is intercepted as the training target.

[0104] In step S22, the historical data set is processed. Here, the early morning peak time period data set is taken as an example. Within the distance range supported by the V2V communication, the data collection vehicle (the ego vehicle) is defined as Veh_host, and the data collection vehicle front vehicle (the adjacent vehicle) is defined as Veh_N. If the initially defined front vehicle of the host vehicle drives away from the current lane, the front vehicle of the host vehicle is updated forward to control the total number of vehicles in front of the ego vehicle to be N. That is, the head vehicle is also updated forward at this time. The head vehicle is defined as Veh_1, and if the initially defined head vehicle drives away, the head vehicle is updated forward to control the total number of vehicles in front of the ego vehicle to be N. It can be seen that the total number of vehicles in the lane of the data collection vehicle Veh_host is N, and since each vehicle is driven by a human driver, there is a reaction time. The average reaction time of each driver of the N vehicles in front of the data collection vehicle is t i . Therefore, the total time lag from the start of Veh_1 to the start of the front vehicle Veh_N of the host vehicle can be represented as: Here, T delay is the time length between the start of the head vehicle and the start of the front vehicle (adjacent vehicle) of the host vehicle.

[0105] Here, a specific embodiment is provided, which defines the moment when the data collection vehicle and the head vehicle establish V2V communication as the starting moment t0, and records the speed sequence of the N vehicles in front of the data collection vehicle after T delay seconds from the starting moment t0. The speed information of the N vehicles in front of the data collection vehicle is represented by the matrix (historical speed set) as follows:

[0106] The first m·T delayThe characteristic value vector of the sample with m∈(0, 1) is taken as a training vector S i : The remaining time length is (1-m)·T delay The sample data with m∈(0, 1) is taken as a test set. Wherein, at the time when the data collection vehicle and the lead vehicle establish V2V communication, the data collection vehicle also establishes V2V communication with other vehicles traveling with it between it and the lead vehicle. Wherein, the first m data refers to the first m columns of data extracted in time sequence, not m columns of data extracted at random.

[0107] Step S23, training GA-BP neural network model. Let the population size be GA_NUM, let the input data be X, the true output be y, and the predicted output after the model be ω1 is the weight matrix connecting the input layer and the hidden layer, e1 is the bias connecting the input layer and the hidden layer, ω2 is the weight matrix connecting the output layer and the hidden layer, and e2 is the bias connecting the output layer and the hidden layer. The optimization strategy adopted is: first, optimize the forward propagation of network parameters, find the optimal initialization parameters, and finally optimize the back propagation, which has very fast network prediction speed and can be applied to real-time prediction. The specific content is shown in the following steps:

[0108] S1. Set the population size and randomly generate the initial population, the dimension of the population individual represents the number of parameters to be optimized of the BP neural network, set the population size to N p , the iteration target number is N m times, and the initial value of the population is randomly generated;

[0109] S2. Set the crossover probability to P c ;

[0110] S3. Input the training set S i to the BP neural network with randomly generated parameters, and calculate the output value of each individual according to the forward calculation of different BP neural networks

[0111] S4. Let the error be equal to the fitness, calculate the fitness value of each individual, and sort the fitness values of each individual in descending order.

[0112] S5. Calculate the vector distance and fitness value difference between the optimal solutions of adjacent populations in all populations, and when the vector distance and fitness error meet the corresponding conditions, perform migration and communication between sub-populations, so that the local worst solution and the local optimal solution between adjacent populations can learn from each other.

[0113] S6. Using the roulette wheel method to select new individuals from the parent chromosomes, the greater the fitness, the greater the probability of being selected, thus as the next generation of population.

[0114] S7. The selected individuals are crossed and mutated according to the given cross and mutation probabilities.

[0115] S8. When the iteration target number N m is not reached, then return to S4, and finally when the iteration number reaches the maximum, compare the local optimal solutions of the remaining two populations, and the one with the lowest fitness value is taken as the global optimal solution.

[0116] S9. Obtain the global optimal solution, including the optimal solution of the number of neurons, the number of network layers, the initial weight and the initial threshold value, etc. The obtained optimal solution parameters are used as the setting parameters of the BP neural network, and the training set is used as the input of the model.

[0117] S10. Perform output calculation of the hidden layer.

[0118] S11. Perform output calculation of the output layer.

[0119] S12. Error calculation.

[0120] S13. Record the number of learned samples, and update the weight and threshold value.

[0121] S14. Determine whether the iteration is complete, if not, return to S13 to continue iteration, until the iteration number is completed and the chromosome with the highest fitness is selected as the final result, thus obtaining the trained genetic algorithm optimized BP neural network long-time vehicle speed prediction model.

[0122] The fitness value in S4 can be expressed as: The output layer calculation in S11 can be: Where A1=tansig(ω1*X+e1).

[0123] The parameters of the neural network prediction model example of the embodiment are set as follows: the specification of the original training data set is 2278x9, corresponding to taking the first 70% as the training set and the last 30% as the test set. As shown in Figure 4 the fitness value change curve of the example is shown; as shown in Figure 5 the comparison chart of the test set and the predicted vehicle speed in the example is shown.

[0124] Step S24, online prediction and optimization. The distance length between the host vehicle and the lead vehicle is defined as the standard, the host vehicle is numbered as Veh_host, the lead vehicle is numbered as Veh_1, and the front vehicle of the host vehicle is numbered as Veh_N. If the front vehicle of the host vehicle defined initially drives away from the current lane, the front vehicle of the host vehicle is updated forwardly, and the total number of vehicles in front of the host vehicle is still controlled to be N. Similarly, if the lead vehicle defined initially drives away, the lead vehicle is updated forwardly to control the total number of vehicles in front of the host vehicle to be N. It can be seen that the total number of vehicles in the lane where the data collection vehicle Veh_host is located is N vehicles. Since each vehicle is driven by a human driver and the vehicle type is different, the reaction time t i There is uncertainty, for example, the average reaction time of ordinary drivers is temporarily taken as 2s, so the total time delay from the start of Veh_1 to the start of the front vehicle Veh_N of the host vehicle can be represented as:

[0125] Based on this, the start time when the host vehicle and the lead vehicle establish V2V communication can be defined as And record the speed sequence of the N vehicles in front of the host vehicle for the next seconds from this time, denoted as As a training set, input into the trained neural network model for prediction, through the prediction model, the future speed information with a dimension of can be obtained At the same time, the time when the prediction starts is recorded as the starting time T0, and the real-time speed sequence of the front vehicle of the host vehicle for the next time τ is recorded from this time: R N = [v N (T0), v N (T1), v N (T2), …, v N (T τ-1 )] T , wherein

[0126] In the above process, the BP neural network based on GA optimization is used to obtain the future predicted speed. Although the predicted result well reflects the traffic flow characteristics, due to the different types of vehicles and drivers in front of the host vehicle, there is a large uncertainty in the reaction time, so there is a time and space misplacement error, which leads to a large difference between the predicted speed and the actual speed in time and space. The specific steps to eliminate the time and space error are as follows:

[0127] Firstly, the following three parameter characteristics are defined:

[0128] Speed matching positioning feature: acceleration feature: ​

[0129] Impact degree feature:

[0130] Turning point feature:

[0131]

[0132] Turning point feature condition 1: the product of the accelerations at the i-1th and i+1th moments is greater than zero at the ith moment, and the total number of all moments satisfying condition 1 is denoted as m1; turning point feature condition 2: the product of the accelerations at the i-1th and i+1th moments is less than or equal to zero at the ith moment, and the total number of all moments satisfying condition 2 is denoted as m2.

[0133] Secondly, an optimization problem is constructed: in order to obtain the optimal time offset error compensation value, a simple optimization algorithm is established, including but not limited to, for example, particle swarm optimization, genetic algorithm, ant colony algorithm, etc., and optionally, an optimization algorithm such as genetic algorithm, particle swarm optimization, etc. is established.

[0134] The three positioning matching errors defined are as follows:

[0135] Acceleration error: δ A (i) = A real (i) - A predict (i);

[0136] Impact degree error: δ G (i) = G real (i) - G predict (i);

[0137] Feature point error: δ P (i) = -P(i).

[0138] Then, the maximum and minimum value principle can be used to normalize the data.

[0139] The objective function is:

[0140] It should be noted that the above objective function can also be expressed as: ρ1, ρ2, ρ3 are the weights corresponding to the acceleration error (value), impact degree error (value), and feature point error (value), respectively, wherein the weight values can be adaptively allocated according to the current road congestion degree, for example, if the road is very congested, the acceleration and impact degree weights can be appropriately increased.

[0141] Finally, the optimization problem is solved: the optimization problem described above can be described as follows, the goal is to obtain the appropriate time offset error compensation value input value u, so as to minimize the matching positioning error J. Here, the time and space offset error value range between the predicted vehicle speed and the actual vehicle speed is defined as u∈[u min ,u max ], wherein u min and u max are respectively the minimum and maximum time and space offset errors compared between the test set and the prediction result in the model training process, that is, the minimum and maximum values of the error values described in step 204.

[0142] Thus, the optimal time offset error compensation value (preset compensation value) u best is obtained by the optimization algorithm, and the time coordinate of the predicted vehicle speed is added to the time offset error compensation value u best , which can ensure the time and space consistency of the predicted vehicle speed and the actual vehicle speed, thereby reducing the time and space errors caused by the uncertainty of the front vehicle driver and the vehicle type, improving the prediction accuracy of the long-time vehicle speed, and realizing the long-time prediction of the future vehicle speed of the front vehicle.

[0143] As can be seen, the V2V technology is used to obtain the driving data information of a fixed period, the BP neural network is trained to establish a long-time vehicle speed prediction model based on GA optimization, and then the feature matching parameters are used to match and position the predicted vehicle speed to eliminate the time and space offset error, and finally the long-time prediction of the future vehicle speed of the target lane front vehicle is realized. The vehicle control system is provided with relatively accurate future long-time vehicle speed information, which provides a basis for improving the energy efficiency, safety and comfort performance of vehicles on congested road sections. In addition, the feature matching parameters are used to optimize the predicted vehicle speed, reduce the time and space offset error caused by the uncertainty of the front vehicle driver and the vehicle type, and improve the accuracy of the long-time prediction of the front vehicle speed.

[0144] It should be understood that although each step in the flowchart involved in each embodiment as described above is displayed in sequence according to the arrow, these steps are not necessarily executed in sequence according to the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other orders. Moreover, at least part of the steps in the flowchart involved in each embodiment as described above can include multiple steps or stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution sequence of these steps or stages is not necessarily sequential, but can be executed in rotation or alternation with at least part of other steps or steps or stages in other steps.

[0145] Based on the same inventive concept, the application further provides a device for predicting the speed of the preceding vehicle in a car-following vehicle group, which is used to implement the method for predicting the speed of the preceding vehicle in a car-following vehicle group as described above. The device provides a solution similar to the implementation described in the method above, and thus the specific limitations in one or more embodiments of the device for predicting the speed of the preceding vehicle in a car-following vehicle group provided below can refer to the limitations of the method for predicting the speed of the preceding vehicle in a car-following vehicle group described above, which will not be repeated here.

[0146] In an exemplary embodiment, as shown in Figure 6 , a device 300 for predicting the speed of the preceding vehicle in a car-following vehicle group is provided, comprising a data acquisition module 81, a data processing module 82 and a result correction module 83, wherein:

[0147] The data acquisition module 81 is configured to, in the case that the ego vehicle establishes communication with at least two preceding vehicles that travel in front of the ego vehicle in the direction of travel, acquire real-time vehicle speeds of each of the at least two preceding vehicles within a period of time, to obtain a set of real-time vehicle speeds;

[0148] The data processing module 82 is configured to input the set of real-time vehicle speeds into a pre-trained vehicle speed prediction model to obtain a first prediction result of the speed of a neighboring vehicle output by the vehicle speed prediction model; wherein the first prediction result of the speed of the neighboring vehicle includes a plurality of preliminary prediction speeds of the neighboring vehicle within a future period of time; the neighboring vehicle is the closest preceding vehicle in front of the ego vehicle in the direction of travel;

[0149] The result correction module 83 is configured to acquire a preset compensation value, process the first prediction result of the speed of the neighboring vehicle based on the preset compensation value to obtain a second prediction result of the speed of the neighboring vehicle; wherein the preset compensation value is determined based on a historical acceleration error value, a historical impact error value and a historical feature point error value corresponding to a historical car-following vehicle group.

[0150] In an exemplary embodiment, the result correction module 83 further comprises: a module for acquiring historical vehicle speeds of each of the at least two preceding vehicles that communicate with the ego vehicle within a historical period of time to obtain a set of historical vehicle speeds; a module for inputting the set of historical vehicle speeds into the pre-trained vehicle speed prediction model to obtain a historical prediction result of the speed of the neighboring vehicle output by the vehicle speed prediction model; a module for acquiring the historical actual speeds of the neighboring vehicle associated with the plurality of historical preliminary prediction speeds in the historical prediction result of the speed of the neighboring vehicle, and calculating error values between each historical actual speed of the neighboring vehicle and the corresponding historical preliminary prediction speed; a module for calling a target function to process each historical acceleration error value, each historical impact error value and the historical feature point error value to obtain a matching error; and a module for processing the matching error and the error values by an optimal solution algorithm to determine the preset compensation value. For details, please refer to Figure 1 and Figure 2 , and the descriptions of Figure 1 and Figure 2 above.

[0151] In an exemplary embodiment, the result correction module 83 further comprises: differentiating each of the historical actual vehicle speeds of the neighboring vehicles to obtain a corresponding first acceleration feature; differentiating each of the historical preliminary predicted vehicle speeds corresponding to the historical actual vehicle speeds of the neighboring vehicles to obtain a corresponding second acceleration feature; and determining a historical acceleration error value corresponding to each of the historical actual vehicle speeds of the neighboring vehicles based on the first acceleration feature and the second acceleration feature. Details can be further referred to Figure 1 and Figure 2 and the above description about Figure 1 and Figure 2 .

[0152] In an exemplary embodiment, the result correction module 83, after obtaining the first acceleration feature and the second acceleration feature, further comprises: differentiating each of the first acceleration feature and the second acceleration feature to obtain a corresponding first jerk feature and a second jerk feature; and determining a historical jerk error value corresponding to each of the historical actual vehicle speeds of the neighboring vehicles based on the first jerk feature and the second jerk feature. Details can be further referred to Figure 1 and Figure 2 and the above description about Figure 1 and Figure 2 .

[0153] In an exemplary embodiment, the result correction module 83 is configured to obtain the historical actual vehicle speeds of the neighboring vehicles associated with a plurality of historical preliminary predicted vehicle speeds in the historical predicted results of the vehicle speeds of the neighboring vehicles, comprising: obtaining the historical actual vehicle speeds associated with K historical preliminary predicted vehicle speeds in the historical predicted results of the vehicle speeds of the neighboring vehicles; wherein K≥3 and K is an integer; and after obtaining the first acceleration feature and the second acceleration feature, further comprising: obtaining a product of the first acceleration feature at a previous time and the first acceleration feature at a next time corresponding to each of the second to the K-1 first acceleration features, and obtaining a first number of positive numbers, a zero and a second number of negative numbers in the product; determining a historical turning point feature based on the first number and the second number, and obtaining a corresponding historical feature point error value by taking an opposite number of the historical turning point feature. Details can be further referred to Figure 1 and Figure 2 and the above description about Figure 1 and Figure 2 .

[0154] In an exemplary embodiment, the result correction module 83 is configured to process the first predicted results of the vehicle speeds of the neighboring vehicles based on a preset compensation value to obtain second predicted results of the vehicle speeds of the neighboring vehicles, comprising: adding each of the preliminary predicted vehicle speeds in the first predicted results of the vehicle speeds of the neighboring vehicles to the preset compensation value to obtain the second predicted results of the vehicle speeds of the neighboring vehicles. Details can be further referred to Figure 1 and Figure 2 and the above description about Figure 1 andFigure 2 .

[0155] In an example embodiment, the data processing module 82 further comprises: a GA-BP neural network model trained based on historical vehicle speed data of historical car-following vehicle groups to obtain a pre-trained vehicle speed prediction model. Details can be further referred to Figure 1 and Figure 2 , and the above description of Figure 1 and Figure 2 .

[0156] Each module in the above prediction of the vehicle speed of the preceding vehicle in the car-following vehicle group can be implemented wholly or partially by software, hardware, or a combination thereof. Each module can be embedded in or independent of the processor in the vehicle-side control device in hardware form, or stored in the memory in the vehicle-side control device in software form, so as to be called and executed by the processor to perform the operation corresponding to each module.

[0157] Figure 7 FIG. 8 is a schematic diagram of the internal structure of the vehicle-side control device in an example embodiment. In an example embodiment, a vehicle-side control device is provided, and the internal structure of the vehicle-side control device can be as shown in Figure 7 The vehicle-side control device includes a processor and a memory. The processor of the vehicle-side control device is configured to provide computing and control capabilities. The memory of the vehicle-side control device includes a non-volatile storage medium, and the non-volatile storage medium stores a computer program. The computer program is executed by the processor to implement a method for predicting the vehicle speed of the preceding vehicle in a car-following vehicle group. The method for predicting the vehicle speed of the preceding vehicle in a car-following vehicle group can be any of the methods for predicting the vehicle speed of the preceding vehicle in a car-following vehicle group mentioned in the embodiments of the present application, and the related embodiments can be referred to above.

[0158] Those skilled in the art can understand that Figure 7 the structure shown in FIG. 8 is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the vehicle-side control device to which the scheme of the present application is applied. Specifically, the vehicle-side control device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0159] Based on the same inventive concept, the present application further provides a computer-readable storage medium having a computer program stored thereon. The computer program is executed by a processor to implement the above-mentioned method for predicting the vehicle speed of the preceding vehicle in a car-following vehicle group. The method for predicting the vehicle speed of the preceding vehicle in a car-following vehicle group can be any of the methods for predicting the vehicle speed of the preceding vehicle in a car-following vehicle group mentioned in the embodiments of the present application, and the related embodiments can be referred to above.

[0160] Based on the same inventive concept, the application further provides a computer program product comprising a computer program which, when executed by a processor, implements the aforementioned method for predicting a speed of a preceding vehicle in a platoon, which is any of the methods for predicting a speed of a preceding vehicle in a platoon mentioned in the application, for which reference can be made to the above.

[0161] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant regulations.

[0162] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiment methods. Any reference to memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. The non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. The volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration but not limitation, the RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., without being limited thereto.

[0163] The technical features of the above embodiments can be combined arbitrarily. In order to make the description simple, all possible combinations of the technical features in the above embodiments are not described, however, as long as the combinations of the technical features do not exist contradictory, they should be considered as the scope of the present application.

[0164] The above-described embodiments are merely illustrative of several embodiments of the present application, and the description is relatively specific and detailed, but should not be understood as a limitation on the scope of the patent. It should be noted that for those skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are all within the scope of the present application. Therefore, the scope of protection of the present application should be subject to the appended claims.

Claims

1. A method for predicting the speed of the vehicle in front in a convoy, characterized in that, The application comprises: In the case of establishing communication with at least two front vehicles respectively following in front of the ego vehicle in the driving direction of the ego vehicle, obtaining the real-time vehicle speed of each of the at least two front vehicles in a period of time to obtain a real-time vehicle speed set; Inputting the real-time vehicle speed set into a pre-trained vehicle speed prediction model to obtain a first prediction result of the adjacent vehicle speed output by the vehicle speed prediction model; wherein the first prediction result of the adjacent vehicle speed includes multiple preliminary prediction vehicle speeds of the adjacent vehicle in a future period; the adjacent vehicle is the closest front vehicle in front of the ego vehicle in the driving direction of the ego vehicle; Obtaining a preset compensation value, processing the first prediction result of the adjacent vehicle speed based on the preset compensation value to obtain a second prediction result of the adjacent vehicle speed; wherein the preset compensation value is determined based on historical acceleration error values, historical impact error values and historical feature point error values corresponding to a historical car-following vehicle group; The historical acceleration error value is determined based on the difference between the first acceleration feature corresponding to each adjacent vehicle historical actual vehicle speed of the adjacent vehicle in the historical car-following vehicle group corresponding to the ego vehicle and the second acceleration feature corresponding to each historical preliminary prediction vehicle speed corresponding to each adjacent vehicle historical actual vehicle speed; The historical impact error value is determined based on the difference between the first impact feature corresponding to each first acceleration feature and the second impact feature corresponding to each second acceleration feature corresponding to each first acceleration feature; The historical feature point error value is determined by taking the opposite number of the historical turning point feature; the historical turning point feature is determined based on the first number of products of the first acceleration feature at the previous moment and the first acceleration feature at the next moment corresponding to the 2nd to K-1th first acceleration features among K first acceleration features, which are positive numbers, and the second number of zero and negative numbers; wherein K≥3 and K is an integer.

2. The method of claim 1, wherein, Further comprising: Obtaining an adjacent vehicle speed historical prediction result corresponding to a historical vehicle speed set; the historical vehicle speed set includes historical vehicle speeds of each of the at least two front vehicles in a historical period of time in communication with the ego vehicle; the adjacent vehicle speed historical prediction result is obtained by inputting the historical vehicle speed set into the vehicle speed prediction model and outputting the pre-trained vehicle speed prediction model; Obtaining adjacent vehicle historical actual vehicle speeds associated with multiple historical preliminary prediction vehicle speeds in the adjacent vehicle speed historical prediction result, and calculating error values between each adjacent vehicle historical actual vehicle speed and the corresponding historical preliminary prediction vehicle speed; Calling a target function to process each historical acceleration error value, each historical impact error value and the historical feature point error value to obtain a minimum value of the matching error in the matching error; Processing the minimum value of the matching error, and the maximum value and the minimum value in the error value by an optimal solution algorithm to determine the preset compensation value.

3. The method of claim 2, wherein, Further comprising: Differential processing each adjacent vehicle historical actual vehicle speed to obtain corresponding each first acceleration feature; Differential processing the historical preliminary prediction vehicle speed corresponding to each adjacent vehicle historical actual vehicle speed to obtain corresponding each second acceleration feature; Based on each of the first acceleration characteristics and the second acceleration characteristics, a historical acceleration error value corresponding to each of the historical actual vehicle speeds of the adjacent vehicle is determined.

4. The method of claim 3, wherein, after obtaining each of the first acceleration characteristics and each of the second acceleration characteristics, the method further comprises: differential processing each of the first acceleration characteristics and each of the second acceleration characteristics to obtain corresponding each of the first impact degree characteristics and each of the second impact degree characteristics; based on each of the first impact degree characteristics and each of the second impact degree characteristics, a historical impact degree error value corresponding to each of the historical actual vehicle speeds of the adjacent vehicle is determined.

5. The method of claim 3, wherein, the obtaining of the historical actual vehicle speeds of the adjacent vehicle associated with the plurality of historical preliminary predicted vehicle speeds in the historical prediction result of the vehicle speed of the adjacent vehicle comprises: obtaining the historical actual vehicle speeds of the adjacent vehicle associated with K historical preliminary predicted vehicle speeds in the historical prediction result of the vehicle speed of the adjacent vehicle.

6. The method of any one of claims 1-5, wherein, the processing of the first prediction result of the vehicle speed of the adjacent vehicle based on the preset compensation value to obtain the second prediction result of the vehicle speed of the adjacent vehicle comprises: summing each of the preliminary predicted vehicle speeds in the first prediction result of the vehicle speed of the adjacent vehicle with the preset compensation value to obtain the second prediction result of the vehicle speed of the adjacent vehicle.

7. The method according to any one of claims 1-5, characterized in that, further comprising: training a GA-BP neural network model based on historical vehicle speed data of the historical car-following platoon to obtain a pre-trained vehicle speed prediction model.

8. A device for predicting a speed of a preceding vehicle in a platoon of vehicles, characterized in that the device comprises: a data acquisition module configured to, in a case where the ego vehicle and at least two front vehicles traveling in front of the ego vehicle in a forward direction thereof establish communication respectively, acquire real-time vehicle speeds of each of the front vehicles in a time period to obtain a real-time vehicle speed set; a data processing module configured to input the real-time vehicle speed set into a pre-trained vehicle speed prediction model to obtain a first prediction result of a vehicle speed of an adjacent vehicle output by the vehicle speed prediction model; wherein the first prediction result of the vehicle speed of the adjacent vehicle comprises a plurality of preliminary predicted vehicle speeds of the adjacent vehicle in a future time period; the adjacent vehicle is the front vehicle closest to the ego vehicle in the forward direction thereof; a result correction module configured to obtain a preset compensation value, process the first prediction result of the vehicle speed of the adjacent vehicle based on the preset compensation value to obtain a second prediction result of the vehicle speed of the adjacent vehicle; wherein the preset compensation value is determined based on a historical acceleration error value, a historical impact degree error value, and a historical feature point error value corresponding to the historical car-following platoon; wherein the historical acceleration error value is determined based on a difference between each of the first acceleration characteristics corresponding to each of the historical actual vehicle speeds of the adjacent vehicle in the historical car-following platoon corresponding to the ego vehicle and each of the second acceleration characteristics corresponding to each of the historical preliminary predicted vehicle speeds corresponding to each of the historical actual vehicle speeds of the adjacent vehicle; the historical impact degree error value is determined based on a difference between each of the first impact degree characteristics corresponding to each of the first acceleration characteristics and each of the second impact degree characteristics corresponding to each of the second acceleration characteristics; and the historical feature point error value is determined based on a difference between each of the first feature points corresponding to each of the first acceleration characteristics and each of the second feature points corresponding to each of the second acceleration characteristics. The historical feature point error value is determined based on taking the opposite of a historical turning point feature; the historical turning point feature is determined based on a first number of products of the first acceleration feature of a previous moment and the first acceleration feature of a next moment corresponding to the second to the K-1th first acceleration features of the K first acceleration features being positive numbers, and a second number of the products being zero and negative numbers; wherein K is greater than or equal to 3 and K is an integer.

9. A vehicle end control device comprising a memory and a processor, the memory storing a computer program, characterized in that, The computer program, when executed by the processor, implements the steps of the method of any one of claims 1-7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program, when executed by the processor, implements the steps of the method of any one of claims 1-7.

Images