Intersection starting method and device of vehicle, vehicle and storage medium

Abstract

The invention relates to an intersection starting method and device for a vehicle, the vehicle and a storage medium, and the method comprises the steps: obtaining the traffic data of a current intersection under the condition that the vehicle meets a preset intersection starting condition; the traffic data and the current state data of the vehicle are input into a pre-constructed intersection starting model, an intersection starting strategy of the vehicle is obtained, and the intersection starting model is composed of a vehicle starting behavior model, a vehicle energy consumption model and a vehicle safe passing model; and controlling the vehicle to pass the current intersection based on the intersection starting strategy. Therefore, the technical problems that in the related technology, a vehicle follow-up model is difficult to accurately describe the unsteady-state driving behavior, so that the model is difficult to obtain an accurate vehicle starting control strategy according to an actual scene, the intersection passing efficiency cannot be improved, and large potential safety hazards exist are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of vehicle control, in particular to a method and device for starting at an intersection, a vehicle and a storage medium. BACKGROUND

[0002] Current ecological driving research mostly focuses on high-speed cruising and simple car-following scenarios, which are relatively stable and the prediction of driving behavior is relatively easy.

[0003] However, in intersection scenarios, especially in the red-green light starting stage, the influencing factors are complex, for example, the randomness of the starting behavior of the preceding vehicle. The traditional car-following model is difficult to accurately describe this non-steady-state driving behavior, which makes it difficult for the model to obtain an accurate vehicle starting control strategy according to the actual scene, which is not conducive to improving the intersection passing efficiency, and there is a great safety hazard, which needs to be improved. SUMMARY

[0004] The present application provides a method and device for starting at an intersection, a vehicle and a storage medium to solve the technical problem that the car-following model in the related art is difficult to accurately describe this non-steady-state driving behavior, which makes it difficult for the model to obtain an accurate vehicle starting control strategy according to the actual scene, which is not conducive to improving the intersection passing efficiency, and there is a great safety hazard.

[0005] The first aspect of the present application provides a method for starting at an intersection of a vehicle, comprising the following steps: acquiring traffic data of a current intersection under the condition that a vehicle meets a preset intersection starting condition; inputting the traffic data and current state data of the vehicle into a pre-constructed intersection starting model to obtain an intersection starting strategy of the vehicle, wherein the intersection starting model is composed of a vehicle starting behavior model for predicting the starting behavior of a preceding vehicle, a vehicle energy consumption model, and a vehicle safety passing model for processing the uncertainty of the starting of the preceding vehicle; and controlling the vehicle to pass the current intersection based on the intersection starting strategy.

[0006] Optionally, in one embodiment of the present application, before the traffic data and the current state data of the vehicle are input into the pre-constructed intersection starting model, the method further comprises: acquiring historical traffic data of the current intersection; based on the historical traffic data, constructing a normal distribution model for describing the reaction time of the driver of the preceding vehicle and a cumulative probability distribution model for describing the cumulative probability of the driver of the preceding vehicle starting at a target time, to combine the normal distribution model and the cumulative probability distribution model to constitute the vehicle starting behavior model.

[0007] Optionally, in an embodiment of the present application, before the traffic data and the current state data of the vehicle are input into the pre-constructed intersection start model, the method further comprises: defining that the acceleration of the preceding vehicle at the starting behavior moment is subject to a normal distribution, so as to model the acceleration process of the preceding vehicle as a Gaussian process, and obtain a Gaussian process framework; performing chi-square test on the Gaussian process framework based on the acceleration data in the historical traffic data, and obtaining a chi-square test result; constructing an initial Gaussian process model by using the chi-square test result; using a radial basis function kernel as a covariance function, so as to combine the covariance function and the initial Gaussian process model to obtain an actual Gaussian process model; and combining the normal distribution model, the cumulative probability distribution model and the actual Gaussian process model to constitute the vehicle starting behavior model.

[0008] Optionally, in an embodiment of the present application, before the traffic data and the current state data of the vehicle are input into the pre-constructed intersection start model, the method further comprises: obtaining the type of the power source of the vehicle; and matching the vehicle energy consumption model based on the type of the power source.

[0009] Optionally, in an embodiment of the present application, before the traffic data and the current state data of the vehicle are input into the pre-constructed intersection start model, the method further comprises: creating a virtual preceding vehicle satisfying a preset characteristic condition; constructing a stochastic model predictive control framework with the virtual preceding vehicle as a tracking target; constructing a chance constraint by combining the uncertainty constraint and the historical speed sequence of the preceding vehicle in the historical traffic data; and constructing the vehicle safe passing model by combining the stochastic model predictive control framework and the chance constraint.

[0010] Optionally, in an embodiment of the present application, the method further comprises: obtaining periodic driving data of the vehicle every preset time length; determining whether the vehicle satisfies a preset update condition based on the periodic driving data, the historical intersection traffic data and historical driving data of a driver of the vehicle; and updating the intersection start model by using the periodic driving data in a case where the preset update condition is satisfied.

[0011] The second aspect embodiment of the present application provides an intersection start device of a vehicle, comprising: a first obtaining module configured to obtain traffic data of a current intersection in a case where a vehicle satisfies a preset intersection start condition; a calculating module configured to input the traffic data and current state data of the vehicle into a pre-constructed intersection start model, and obtain an intersection start strategy of the vehicle, wherein the intersection start model is constituted by a vehicle starting behavior model for predicting starting behavior of a preceding vehicle, a vehicle energy consumption model and a vehicle safe passing model for processing uncertainty of starting of the preceding vehicle; and a control module configured to control the vehicle to pass the current intersection based on the intersection start strategy.

[0012] Optionally, in an embodiment of the present application, further comprising: a second obtaining module, configured to obtain historical traffic data of the current intersection; a first constructing module, configured to construct, based on the historical traffic data, a normal distribution model for describing reaction time of a driver of a preceding vehicle and a cumulative probability distribution model for describing cumulative probability of the driver of the preceding vehicle starting at a target time, so as to combine the normal distribution model and the cumulative probability distribution model to form the vehicle starting behavior model.

[0013] Optionally, in an embodiment of the present application, further comprising: a second constructing module, configured to define that acceleration of the preceding vehicle at a starting behavior time is subject to a normal distribution, so as to model acceleration process of the preceding vehicle as a Gaussian process to obtain a Gaussian process framework; a testing module, configured to perform chi-square test on the Gaussian process framework based on acceleration data in the historical traffic data to obtain a chi-square test result; a third constructing module, configured to construct an initial Gaussian process model by using the chi-square test result; a fourth constructing module, configured to use a radial basis function kernel as a covariance function to combine the covariance function and the initial Gaussian process model to obtain an actual Gaussian process model; and a fifth constructing module, configured to combine the normal distribution model, the cumulative probability distribution model and the actual Gaussian process model to form the vehicle starting behavior model.

[0014] Optionally, in an embodiment of the present application, further comprising: a third obtaining module, configured to obtain a power source type of the vehicle; and a matching module, configured to match the vehicle energy consumption model based on the power source type.

[0015] Optionally, in an embodiment of the present application, further comprising: a creating module, configured to create a virtual preceding vehicle satisfying a preset characteristic condition; a sixth constructing module, configured to construct a stochastic model predictive control framework by taking the virtual preceding vehicle as a tracking target; a seventh constructing module, configured to construct a chance constraint by combining an uncertainty constraint of a starting probability of the preceding vehicle and a historical speed sequence of the preceding vehicle in the historical traffic data; and an eighth constructing module, configured to construct the vehicle safe passing model by combining the stochastic model predictive control framework and the chance constraint.

[0016] Optionally, in an embodiment of the present application, further comprising: a fourth obtaining module, configured to obtain periodic driving data of the vehicle every preset time length; a judging module, configured to judge whether the vehicle satisfies a preset updating condition based on the periodic driving data, historical intersection traffic data and historical driving data of a driver of the vehicle; and an updating module, configured to update the intersection starting model by using the periodic driving data in a case where the preset updating condition is satisfied.

[0017] The third aspect of the present application provides a vehicle, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the intersection starting method of the vehicle as described in the above embodiments.

[0018] The fourth aspect of the present application provides a computer readable storage medium, which stores computer instructions for causing the computer to execute the intersection starting method of the vehicle as described in the above embodiments.

[0019] The fifth aspect of the present application provides a computer program product comprising a computer program, which, when executed, implements the intersection starting method of the vehicle as described above.

[0020] The embodiments of the present application can model the intersection starting behavior to more comprehensively consider the influence of the changes in the starting stage on the intersection passing, introduce the vehicle safety passing model to process the uncertainty in the starting process, and finally combine the energy consumption model to achieve more efficient energy-saving driving, thereby achieving full use of traffic information, significantly improving the intersection passing efficiency of the vehicle and reducing energy consumption, and improving the driving experience of the driver. Thus, the technical problem that the car-following model in the related art is difficult to accurately describe such non-steady-state driving behavior, which leads to the difficulty of the model to obtain accurate vehicle starting control strategies according to the actual scene, and the existence of great safety hazards while being not conducive to improving the intersection passing efficiency is solved.

[0021] Additional aspects and advantages of the present application will be in part apparent and in part pointed out hereinafter. BRIEF DESCRIPTION OF DRAWINGS

[0022] The above and / or additional aspects and advantages of the present application will become apparent and be readily appreciated from the following description, taken in conjunction with the accompanying drawings, in which: Figure 1 A flowchart of the intersection starting method of the vehicle according to an embodiment of the present application is provided. Figure 2 A flowchart of the intersection starting method of the vehicle according to an embodiment of the present application is provided. Figure 3 A structural schematic diagram of the intersection starting device of the vehicle according to an embodiment of the present application is provided. Figure 4 A structural schematic diagram of the vehicle according to an embodiment of the present application is provided. DETAILED DESCRIPTION

[0023] Embodiments of the present application are described below in detail, examples of which are shown in the drawings, wherein the same or similar notations represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by reference to the drawings are exemplary and are intended to explain the present application, and cannot be understood as a limitation of the present application.

[0024] The intersection starting method, device, vehicle and storage medium of the vehicle of the embodiments of the present application are described below with reference to the drawings. For the related technologies mentioned in the above background, the vehicle following model is difficult to accurately describe such non-steady state driving behavior, which leads to the difficulty of the model to obtain accurate vehicle starting control strategy according to the actual scene, which is not conducive to improving the intersection passing efficiency, and there is a great safety hazard. The present application provides an intersection starting method of a vehicle, in which the intersection starting behavior can be modeled to more comprehensively consider the influence of the change of the starting stage on the intersection passing, and then the vehicle safety passing model is introduced to process the uncertainty in the starting process, and finally the energy consumption model is combined to realize more efficient energy-saving driving, thereby realizing the full use of traffic information, significantly improving the intersection passing efficiency of the vehicle and reducing the energy consumption, and improving the driving experience of the driver. Thus, the technical problem of the related art that the vehicle following model is difficult to accurately describe such non-steady state driving behavior, which leads to the difficulty of the model to obtain accurate vehicle starting control strategy according to the actual scene, which is not conducive to improving the intersection passing efficiency, and there is a great safety hazard is solved.

[0025] Specifically, Figure 1 A flowchart of an intersection starting method of a vehicle provided by an embodiment of the present application is shown.

[0026] As Figure 1 shown, the intersection starting method of the vehicle includes the following steps: In step S101, traffic data of the current intersection is obtained when the vehicle meets the preset intersection starting condition.

[0027] In actual execution process, the embodiments of the present application can first determine whether the vehicle meets the intersection starting condition, for example, the embodiments of the present application can confirm whether it is in the intersection range according to the vehicle navigation, can determine whether it is in the intersection range according to the behavior of other vehicles in the surrounding, and can confirm whether it is in the intersection range according to the positioning information. In the case of being in the intersection range, it can be determined whether it needs to start at the intersection according to the behavior of the front vehicle or the traffic signal.

[0028] In the case of determining that the vehicle is in the intersection starting condition, the embodiments of the present application can obtain the traffic data of the current intersection, including the front vehicle driving data, traffic signal data, etc.

[0029] In step S102, the traffic data and the current state data of the vehicle are input into a pre-constructed intersection start model to obtain an intersection start strategy of the vehicle, wherein the intersection start model is composed of a vehicle start behavior model for predicting the start behavior of the preceding vehicle, a vehicle energy consumption model, and a vehicle safe passing model for processing the uncertainty of the start of the preceding vehicle.

[0030] It can be understood that, in the start process at the intersection, the vehicle needs to start quickly from the static state, which significantly increases the acceleration demand, which often leads to a significant increase in instantaneous energy consumption. The existing energy-saving driving model is difficult to provide the best energy efficiency planning for the following vehicle due to the inability to accurately predict the start acceleration and reaction time of the preceding vehicle. For example, the traditional car-following model fails to fully consider the random acceleration factors in the intersection environment, resulting in a conservative acceleration strategy for the vehicle, which increases unnecessary energy consumption. In addition, the failure to effectively utilize traffic signal information further limits the energy-saving potential.

[0031] In actual driving, the start of the preceding vehicle at the intersection has uncertainty, which is affected by various factors such as driver reaction time and intersection. The traditional model predictive control method is relatively conservative when dealing with these uncertainties, and usually sets a too large safety distance, so that the autonomous vehicle reduces the speed or delays the start, affecting the traffic efficiency. Although some studies have tried to introduce robust model predictive control to deal with uncertainty, in complex and random traffic behavior, robust model predictive control is still too cautious, and cannot achieve efficient energy-saving driving.

[0032] At the same time, the traditional intelligent driving model fails to effectively capture the dynamic behavior of the preceding vehicle in these two stages, making it difficult to support the start decision optimization of the intelligent driving vehicle.

[0033] Further, the embodiments of the present application can input traffic data and current state data of the vehicle, such as energy consumption state, into the intersection start model to predict the behavior of the preceding vehicle through the vehicle start behavior model, plan the intersection passing path of the vehicle and the vehicle control scheme, optimize the vehicle control scheme through the vehicle energy consumption model, and constrain the intersection passing path and the vehicle control scheme through the vehicle safe passing model, thereby fully utilizing the traffic information, significantly improving the intersection passing efficiency of the vehicle and reducing the energy consumption, and improving the driving experience of the driver.

[0034] Optionally, in an embodiment of the present application, before the traffic data and the current state data of the vehicle are input into the pre-constructed intersection start model, it further includes: obtaining historical traffic data of the current intersection; based on the historical traffic data, constructing a normal distribution model for describing the reaction time of the driver of the preceding vehicle and a cumulative probability distribution model for describing the cumulative probability of the driver of the preceding vehicle starting at the target time, to combine the normal distribution model and the cumulative probability distribution model to constitute the vehicle start behavior model.

[0035] First, the embodiments of the present application can collect sample data, and when collecting time and place, it can be collected at multiple typical intersections, covering different traffic flow scales (such as congested intersections in urban core areas and low-flow intersections in suburban areas) and layouts (cross-shaped, T-shaped, etc.), and the collection period covers weekdays and weekends, day and night periods, and the sampling amount is balanced in each period. Collect data in different weather (sunny, rainy, snowy, foggy, hazy) and seasons, and the sample size of special weather is not less than 10%, covering road wetness, visibility obstruction and other working conditions, and strengthening the universality of the model.

[0036] The equipment deployment during data collection, such as drones equipped with cameras, is used to obtain an aerial view of the intersection, observe traffic flow, vehicle position, and traffic signal status. The drone can record video in real time and be equipped with a high-resolution camera to capture clear images; mobile camera: installed at fixed positions on the intersection, used to capture traffic conditions from different angles; sensors: various sensors (such as radar, laser range finder, ground inductor) are installed at the intersection to collect vehicle speed, acceleration, position, and direction information, and considering the difficulty of deployment, a mobile camera can be deployed on the sidewalk to capture traffic light countdown and other detailed vehicle information such as vehicle type, power type, etc. Store the collected data in the hardware device.

[0037] Data processing is performed on the collected data, data cleaning: eliminate invalid data (such as missing, damaged records) and outliers to ensure the quality of the data set. Denoising: use filters to remove noise in video and sensor data to ensure data accuracy.

[0038] Data labeling: vehicle type needs to be labeled, identifying its power type such as electric vehicles, fuel vehicles, in order to further analyze vehicle behavior, using YOLOv5 model to fuse multi-scale feature maps and attention mechanism, enhancing small target detection and tracking accuracy, combining KF and particle filtering to dynamically estimate vehicle state. Identify the time period of the start-up behavior.

[0039] Feature extraction: feature selection: extract features related to the preceding vehicle start-up behavior from labeled data, such as: vehicle state before starting (speed, acceleration), traffic signal status, surrounding environment (such as traffic flow, pedestrian situation), relative position of the preceding vehicle and the vehicle.

[0040] On this basis, further data processing is performed, and the extracted features are standardized to eliminate the influence of different feature scales on modeling; use sliding window technology to extract time series features to capture the dynamic changes of the preceding vehicle start-up behavior.

[0041] After obtaining the sample data, the embodiment of the application can determine the historical traffic data of the corresponding intersection according to the current intersection, or directly use the sample data.

[0042] The starting behavior of the vehicle is divided into two stages: an initial preparation stage and an acceleration stage. The initial preparation stage refers to the transient process from the driver having a starting intention, stepping on the accelerator pedal to the vehicle actually starting to move. The acceleration stage refers to the stage from when the vehicle starts to move until it reaches a stable speed and passes through the intersection. If the vehicle starts to move at a red light, it means that the driver has predicted the change of the red and green light.

[0043] In the initial preparation stage, the embodiment of the application can use a normal distribution model to describe the reaction time (the time stamp at which the preparation stage ends) of the driver of the preceding vehicle, and use a chi-square test ( ) to test the fitting result. The cumulative probability of the preceding vehicle starting at a certain time can be described by the cumulative distribution in the time domain.

[0044] The driver reaction time conforms to the normal distribution model, and in the preparation stage, the starting probability at a certain time is described by the cumulative probability in the T0 to T1 time domain, and the cumulative probability distribution is: (1) wherein, is the mean of the reaction time, is the variance thereof, the earliest starting time can be set to -3s, and as the phase time changes, the red light changes to green light, and the probability of the vehicle starting gradually increases, from the preparation stage to the acceleration stage.

[0045] Optionally, in one embodiment of the application, before inputting the traffic data and the current state data of the vehicle into the pre-constructed intersection starting model, it further includes: defining that the acceleration of the preceding vehicle at the starting behavior time conforms to a normal distribution, modeling the acceleration process of the preceding vehicle as a Gaussian process to obtain a Gaussian process framework; performing a chi-square test on the Gaussian process framework based on the acceleration data in the historical traffic data to obtain a chi-square test result; constructing an initial Gaussian process model by using the chi-square test result; using a radial basis function kernel as a covariance function to combine the covariance function and the initial Gaussian process model to obtain an actual Gaussian process model; and combining the normal distribution model, the cumulative probability distribution model and the actual Gaussian process model to form a vehicle starting behavior model.

[0046] Acceleration phase modeling: In the acceleration phase, the acceleration of the preceding vehicle is assumed to follow a normal distribution at each starting behavior time, and a Gaussian process (GP) is used to describe this behavior. The mean and variance of the Gaussian process are written as time series, where the mean series represents the mean function of the GP, and the normal distribution at each time is combined with the original data to perform a chi-square test. The GP model is fitted using the moments derived from the chi-square test, and a radial basis function (RBF) kernel is used to describe the covariance function of the acceleration Gaussian process.

[0047] The acceleration phase uses a Gaussian process defined by a mean function and a covariance function: (2) Mean function : (3) where tanh is the hyperbolic tangent function, sin is the sine function, , , , and , are the parameters to be optimized. can accurately reflect the rapid rise and saturation characteristics of acceleration, can depict the acceleration fluctuations caused by engine vibration or road bumps, describes the slow rise process of the initial acceleration.

[0048] The covariance function of the acceleration Gaussian process is described by the radial basis function kernel (4) where is the covariance function, , , , and , The hyperparameters are determined by fitting the least squares or maximum likelihood estimation based on the collected multi-scenario vehicle acceleration data, and a more optimal GP model is constructed. l is the length parameter, which controls the horizontal length scale of the function change; is the equation of the process noise.

[0049] The hyperparameters are determined by maximum likelihood estimation: (5) where denotes the set of hyperparameters, denotes the probability of observing the acceleration at a given time and hyperparameters accurately captures the acceleration trend and uncertainty.

[0050] So far, the embodiment of the present application can obtain an actual Gaussian process model to combine the model in the preparation stage to constitute a vehicle starting behavior model.

[0051] Optionally, in an embodiment of the present application, before the traffic data and the current state data of the vehicle are input into the pre-constructed intersection starting model, the method further comprises: obtaining the power source type of the vehicle; and matching the vehicle energy consumption model based on the power source type.

[0052] Further, the embodiment of the present application can establish a vehicle energy consumption model.

[0053] The embodiment of the present application can be applied to fuel vehicles, hybrid vehicles and the like, and according to the type of power source, the embodiment of the present application can match the corresponding energy consumption model.

[0054] For example, for a hybrid vehicle, the instantaneous energy consumption calculation formula is as follows (6) Engine energy consumption: (7) Wherein, Affected by the engine speed torque T and the fuel injection amount, etc., it can be obtained from the engine model, is the engine efficiency. is a constant value, such as gasoline about 32 MJ / L.

[0055] Motor energy consumption: = (8) Wherein It needs to be calculated according to its type, DC motor P=UI, AC motor needs to use the corresponding electromagnetic equation to calculate; Related to the motor speed and load.

[0056] Weight coefficient determination: And It needs to be dynamically allocated according to the working mode, pure electric drive Engine direct drive , the hybrid mode needs to be determined according to the power distribution strategy of the engine and the motor.

[0057] Optionally, in an embodiment of the present application, before the traffic data and the current state data of the vehicle are input into the pre-constructed intersection start model, the following steps are further included: creating a virtual preceding vehicle satisfying a preset characteristic condition; constructing a random model prediction control framework with the virtual preceding vehicle as a tracking target; constructing an opportunity constraint by combining the start probability of the preceding vehicle with the uncertainty constraint and the historical speed sequence of the preceding vehicle in the historical traffic data; and constructing a vehicle safe passing model by combining the random model prediction control framework and the opportunity constraint.

[0058] As a possible implementation manner, the embodiment of the present application can construct a vehicle safe passing model to perform ecological driving longitudinal planning by further constructing a random model.

[0059] When the preceding vehicle is in the preparation stage, only the start probability can be obtained, and the behavior in the acceleration stage cannot be predicted. In order to simulate the driver's prediction of the fast reaction of the preceding vehicle to the green light, a virtual preceding vehicle that does not actually exist in the actual traffic scene is introduced to form a feasible region of earlier and more stable start of the electric vehicle, thereby improving the time efficiency and reducing the energy consumption. Therefore, the virtual preceding vehicle in the intersection ecological driving problem can be regarded as a target vehicle in the equivalent following problem.

[0060] First, a random model prediction control (SMPC) problem is constructed: (9) (10) (11) (12) (13) (14) In formula (9), the objective function is minimized to solve the optimal control input of the vehicle state , which comprehensively considers the vehicle state , the vehicle control , the preceding vehicle state , and the virtual preceding vehicle state to optimize the overall performance of the vehicle.

[0061] Formula (10) updates the vehicle state according to the vehicle dynamics principle, and calculates the vehicle state at the next time point according to the current vehicle state and the control input .

[0062] Formula (11) defines the constraint acceleration and the acceleration change rate to prevent exceeding the output range of the power system and affecting the vehicle comfort.

[0063] Equation (12) deals with the uncertainty of the front vehicle state, which is quantified by the probability form. The state of the virtual front vehicle is estimated by the statistical model based on the historical driving data. Probability, β To pre-set the confidence level, adjust the risk tolerance and the dangerous degree of traffic scene, and balance the risk and performance optimization requirements.

[0064] The embodiment of the application only considers longitudinal planning, and the vehicle longitudinal model is constructed as: (15) (16) The environmental state covers the phase and time information of the signal light.

[0065] Objective function setting: consider safety, traffic efficiency, traffic rules and energy consumption design objective function.

[0066] (17) The first two terms represent the time efficiency of the vehicle, The expected following distance is represented by the first term, which focuses on the behavior of the front vehicle in the preparation stage. The position difference between the vehicle and the virtual leading vehicle and the expected following distance are used to construct this stage. In this stage, accurate control of the relationship between the virtual front vehicle and the position is critical to the subsequent driving rhythm. For example, when waiting at an intersection, adjusting according to this can reduce the start-up delay and optimize the overall travel time. The second term focuses on the behavior of the front vehicle in the acceleration stage after starting. The speed difference square measures the speed coordination between the two vehicles, so that the vehicle can respond to the acceleration of the front vehicle in time, maintain a reasonable distance and speed adaptation, and prevent the vehicle from being too close or too loose, which will affect the efficiency. For example, in the city expressway following scene, a reasonable speed difference can keep the traffic smooth and improve the traffic efficiency; the third and fourth terms represent comfort, which constrain the acceleration change, and the fifth term considers energy consumption Expected following distance: (18) Where, The minimum safe following distance is The driver reaction time is L, and the length of the front vehicle is further described to ensure safety constraints.

[0067] (19) Virtual front vehicle modeling:

[0068] The matrices A and B are related to the characteristics of the system, and the state of the virtual front vehicle is updated.

[0069] Acceleration fusion start probability Gaussian distribution mean ​, the noise w(k) follows a normal distribution, defined by equation (4).

[0070] A prediction error recursion is introduced between the virtual leading vehicle and the real leading vehicle to evaluate the confidence level of the virtual leading vehicle model. The prediction error is: (21) Specifically, the prediction error model can be divided into deterministic and random parts.

[0071] (22) Substitute equation (21) into equation (22).

[0072] (23) Chance constraint transformation: The uncertainty of the leading vehicle is converted into an opportunity constraint in the stochastic model predictive control (SMPC) to ensure that the ego vehicle meets the safety distance and avoids collision risks. In order to more accurately quantify and control risks while considering the uncertainty of the leading vehicle, ensure that the decision of the ego vehicle can be optimized under the guarantee of a certain safety probability, and thus effectively deal with the unpredictability of the leading vehicle behavior.

[0073] GPR (Global Positioning and Recognition) based leading vehicle speed prediction: Radial basis function and Matern kernel function combination are selected, radial basis function is used to capture the global trend of data, and Matern kernel performs well in processing local changes, and adapts to the complex transformation characteristics of full vehicle acceleration. Bayesian optimization algorithm is used to optimize hyperparameters (such as kernel function length scale, signal variance, etc.), by constructing a probability model of hyperparameters, evaluating the performance of the model (such as log marginal likelihood) under different hyperparameter combinations according to a small amount of training data, efficiently searching the optimal hyperparameter space, reducing the workload of manual parameter adjustment, and improving the model fitting accuracy and generalization ability.

[0074] Further, the embodiments of the present application can perform model training and verification: k-fold cross-validation (k = 5 or 10) is used to divide the data set, most of the data is used for training, and a small part is used for verification. The training and verification process is repeated multiple times to fully evaluate the performance of the model. During the training process, based on the negative log-likelihood loss function, optimization algorithms such as conjugate gradient method are used to iteratively solve the model parameters, so that the model prediction distribution can fit the true data distribution as much as possible. At the same time, monitor the root mean square error (RMSE), mean absolute error (MAE) and other indicators on the validation set to prevent overfitting and ensure the prediction accuracy and reliability of the model on unseen data, such as controlling the root mean square error on the validation set to 0.2-0.5m / s 2within the range (adjusted according to actual vehicle motion characteristics and data quality).

[0075] After obtaining the complete vehicle intersection passing model, the embodiment of the present application can also perform multi-scenario testing through traffic simulation software, record and analyze the performance of the ego vehicle, and evaluate the effectiveness of the strategy.

[0076] In step S103, the vehicle is controlled to pass the current intersection based on the intersection start strategy.

[0077] The vehicle intersection passing model of the embodiment of the present application can transmit the longitudinal speed and acceleration curve planned after energy consumption consideration to the lower energy management strategy according to the current intersection traffic data and the current state data of the vehicle, replace the predicted speed trajectory in the original MPC algorithm, and perform torque distribution of the engine and the motor.

[0078] Optionally, in one embodiment of the present application, it further includes: acquiring periodic driving data of the vehicle every preset time length; judging whether the vehicle meets a preset update condition based on the periodic driving data, historical intersection traffic data and historical driving data of the driver of the vehicle; and updating the intersection start model using the periodic driving data in the case where the preset update condition is met.

[0079] The embodiment of the present application can periodically collect data and update data. Or when the amount of newly collected data reaches a certain threshold, the model update mechanism is started.

[0080] An incremental learning method is adopted to include new data into the training set, and a stochastic gradient descent variant algorithm (such as Adagrad and Adadelta) is used to fine-tune the model parameters, so that the model can quickly adapt to the recent behavior changes of the preceding vehicle (such as driving style change and driving characteristics change caused by intersection change) on the basis of maintaining the original knowledge of the model, and continuously improve the prediction accuracy of the model in the entire driving process.

[0081] In combination with Figure 2 As shown in FIG. 1, the working principle of the intersection start method of the vehicle of the embodiment of the present application is described in detail.

[0082] As shown in FIG. 1, the working principle of the intersection start method of the vehicle of the embodiment of the present application is described in detail. Figure 2 Step S201, determine the data collection time and place.

[0083] ​In the traffic data collection of a certain city, the collection time and place are first determined. Selecting multiple types of typical intersections including congested intersections in the city core (such as cross-shaped intersections in the central business district) and low-flow intersections in the suburbs (such as T-shaped intersections in the suburbs), etc. The collection period is set to cover weekdays and weekends, and the sampling frequency is increased during the morning and evening peak hours to ensure balanced sampling amount in each period. At the same time, different weather conditions such as sunny, rainy, snowy, and foggy are fully considered, and the collection frequency is dynamically adjusted through real-time weather monitoring system to ensure that the sample size of special weather conditions accounts for no less than 10%, so as to comprehensively cover the working conditions such as slippery road and obstructed vision. In terms of equipment deployment, unmanned aerial vehicles are equipped with cameras at intersections, and the aerial view of the intersection is obtained in real time by using unmanned aerial vehicles, and clear traffic condition images are captured by cameras. At the same time, mobile cameras are installed at fixed positions on the intersection to capture traffic conditions from different angles. In addition, various sensors including radar, laser range finder, ground sensor, infrared sensor, etc. are installed at the intersection, which can collect vehicle speed, acceleration, direction and position information. Considering the deployment difficulty, mobile cameras are also deployed on the sidewalk to capture detailed information such as traffic light countdown.

[0084] Step S202, data processing.

[0085] The embodiments of the present application can analyze the massive video data collected by using professional image recognition and processing technology. Advanced deep learning algorithms (such as target detection algorithms based on convolutional neural networks) are used to recognize vehicles in videos, accurately locate the position of each vehicle in the image, and integrate vehicle speed, acceleration and position information from radar, laser range finder and other devices in time synchronization mechanism by using multi-sensor fusion technology, to construct complete vehicle trajectory data. Data cleaning algorithms are used to remove abnormal values caused by device failure or environmental interference, such as setting speed and acceleration thresholds to exclude obviously unreasonable data points. Data smoothing techniques (such as moving average method and Kalman filter) are used to process noise data and improve data quality. For missing data or abnormal data, interpolation algorithms (such as linear interpolation and spline interpolation) are used to fill in according to the physical law of vehicle driving and the surrounding vehicle data, to ensure data continuity and integrity, and provide a solid and reliable data foundation for subsequent model construction.

[0086] Step S203, constructing a vehicle starting behavior model.

[0087] The starting behavior of a vehicle is divided into two stages: the initial preparation stage and the acceleration stage. The initial preparation stage refers to the short process from the driver's intention to start (such as seeing a green light) to actually stepping on the accelerator pedal and preparing the vehicle to start. In this stage, reaction time is the key factor, usually affected by traffic signals, driver attention, and psychological preparation. The acceleration stage refers to the process after the vehicle starts moving, during which the vehicle accelerates to reach the target speed and smoothly pass through the intersection. In this stage, the dynamic behavior of the vehicle is complex and affected by road conditions and environmental factors.

[0088] Secondly, a probability model of the leading vehicle's starting behavior is established. In the initial preparation stage, the reaction time characteristics of the front vehicle driver are focused on. Through rigorous statistical analysis and distribution fitting test, it is assumed that it follows a normal distribution and the fitting result is tested by chi-square test. The mean and variance are calculated using a large amount of sample data, and the mean reaction time μ and variance σ are determined in a rich sample. According to the normal distribution principle, the cumulative probability function is constructed:

[0089] (where t is the earliest starting time The earliest starting time) accurately describes the starting probability of the front vehicle, and accurately models the preparation stage behavior Finally, the acceleration stage model is established: the starting acceleration stage of the front vehicle is regarded as a complex Gaussian process. Based on deep mining of the rules contained in the data, the mean function is carefully determined as a cubic polynomial:

[0090] The least squares method is used to fit the polynomial coefficients to a large number of historical data points. The radial basis function kernel is used to describe the covariance function of the acceleration Gaussian process:

[0091] The maximum likelihood estimation (MLE) algorithm is used to solve the hyperparameters, and thus the Gaussian process model is completely constructed, accurately capturing the uncertainty and dynamic characteristics of the front vehicle's acceleration in the acceleration stage.

[0092] Step S204, construct a vehicle energy consumption model.

[0093] According to the physical characteristics and energy conversion principles of the vehicle, an instantaneous energy consumption model is constructed. The key parameters of the vehicle, including vehicle mass, transmission efficiency, air resistance coefficient, rolling resistance coefficient, vehicle frontal area, rotational mass conversion factor, and slope resistance coefficient, are closely related to the construction of an accurate energy consumption calculation model, which provides a key basis for vehicle energy consumption prediction and optimization.

[0094] Step S205, construct a vehicle safety passing model, and use a stochastic model predictive control ecological driving longitudinal planning.

[0095] When the current vehicle is in the preparation stage, only the starting probability can be obtained, and the behavior in the acceleration stage cannot be predicted. To simulate the driver's prediction of the front vehicle's quick response to the green light, a virtual front vehicle that does not exist in the actual traffic scene is introduced to form a feasible region for the electric vehicle to start earlier and more smoothly, thereby improving time efficiency and reducing energy consumption. Therefore, in the intersection ecological driving problem, the virtual front vehicle can be regarded as the target vehicle in the equivalent following problem.

[0096] The design formula of the objective function is considered in terms of safety, traffic efficiency, traffic rules and energy consumption. The multi-objective optimization is achieved by scientifically setting the weight parameters, and the longitudinal kinematic model and strict state constraints and opportunity constraints are defined.

[0097] The virtual front vehicle motion trajectory is recursively generated by referring to the vehicle kinematic model and the front vehicle behavior model, the starting feasible region of the ego vehicle is expanded, and the starting decision is optimized.

[0098] Opportunity constraint conversion: The probability of opportunity constraint equation is difficult to solve in SMPC. The uncertainty of the front vehicle is converted into an opportunity constraint in SMPC to ensure that the ego vehicle meets the safety distance and avoids collision risk. In order to more accurately quantify and control the risk while considering the uncertainty of the front vehicle, ensure that the decision of the ego vehicle can be optimized under the guarantee of a certain safety probability, so as to effectively deal with the unpredictability of the front vehicle behavior.

[0099] Front vehicle speed prediction based on GPR: In the preparation stage, make full use of the Gaussian process model to predict the behavior trend of the front vehicle in combination with the virtual front vehicle starting acceleration; in the acceleration stage, deeply mine the collected vehicle historical data, and use the powerful prediction ability of Gaussian process regression (GPR) to strictly deduce the posterior distribution of the front vehicle acceleration according to the Bayesian formula, accurately calculate the posterior mean and variance, and verify the prediction accuracy by using strict evaluation indexes (MAE and RMSE) to provide accurate decision basis for the speed planning of the ego vehicle, effectively improve the planning accuracy and foresight, and realize efficient ecological driving.

[0100] Step S206: The speed planned by the upper layer is transmitted to the lower layer energy management strategy based on MPC to realize more accurate engine and motor torque distribution.

[0101] Step S207: Simulation and experimental verification: The effectiveness of the proposed strategy is verified through traffic simulation software and real vehicle experiments.

[0102] Step S208: The actual collected intersection data and vehicle state data are input into the vehicle intersection passing model that has been verified to obtain the actual vehicle intersection passing strategy.

[0103] In summary, the embodiment of the present application can describe the behavior of the preceding vehicle during the starting process at the signalized intersection through a two-stage model based on data driving. This model is divided into a preparation stage and an acceleration stage, which can more accurately capture the reaction time and acceleration change of the preceding vehicle, making up for the shortcomings of traditional models in describing the complex dynamic behavior at the intersection.

[0104] By introducing the SMPC strategy, uncertainty is included in the optimization framework, which can better cope with the randomness of the starting behavior of the preceding vehicle in the dynamic traffic environment. This method effectively expands the feasible range of the safety distance, improves the energy efficiency and traffic efficiency compared with the traditional model.

[0105] By using the Gaussian process regression method, the acceleration of the preceding vehicle is predicted in real time. This enables more accurate adjustment of the acceleration strategy of the following vehicle during the starting stage, reduces the conservative decision-making caused by uncertainty, and thus improves the overall energy efficiency.

[0106] At the same time, in the algorithm design, the energy efficiency and traffic efficiency are considered at the same time, and a multi-performance optimization framework is proposed to ensure that the energy consumption and the passing time are balanced during the starting process at the intersection. This comprehensive consideration strategy is more in line with the actual needs of urban traffic.

[0107] By fully utilizing the information of the traffic signal, the starting strategy of the vehicle is optimized based on the phase and timing information of the traffic signal, the response speed of the vehicle to the traffic signal change is improved, and the energy consumption during the starting process is further reduced.

[0108] The intersection starting method of the vehicle according to the embodiment of the present application can model the intersection starting behavior to more comprehensively consider the influence of the changes in the starting stage on the intersection passing, introduce a vehicle safety passing model to process the uncertainty in the starting process, and finally combine an energy consumption model to realize more efficient energy-saving driving, thereby realizing the full utilization of traffic information, significantly improving the intersection passing efficiency of the vehicle and reducing energy consumption, and improving the driving experience of the driver. Thus, the technical problem that in the related art, the car-following model is difficult to accurately describe such non-steady-state driving behavior, resulting in that the model is difficult to obtain accurate vehicle starting control strategy according to the actual scene, which is not conducive to improving the intersection passing efficiency and has a great safety hazard is solved.

[0109] Secondly, the intersection starting device of the vehicle according to the embodiment of the present application is described with reference to the accompanying drawings.

[0110] Figure 3 is a block schematic diagram of the intersection starting device of the vehicle according to the embodiment of the present application.

[0111] As Figure 3 shown, the intersection starting device 10 of the vehicle includes a first acquisition module 100, a calculation module 200, and a control module 300.

[0112] Specifically, the first obtaining module 100 is configured to obtain traffic data of a current intersection when the vehicle meets preset intersection start conditions.

[0113] The calculation module 200 is configured to input the traffic data and current state data of the vehicle into a pre-constructed intersection start model to obtain an intersection start strategy of the vehicle, wherein the intersection start model is composed of a vehicle start behavior model for predicting start behavior of a preceding vehicle, a vehicle energy consumption model, and a vehicle safe passing model for processing start uncertainty of the preceding vehicle.

[0114] The control module 300 is configured to control the vehicle to pass the current intersection based on the intersection start strategy.

[0115] Optionally, in an embodiment of the present application, the intersection start device 10 of the vehicle further comprises a second obtaining module and a first constructing module.

[0116] The second obtaining module is configured to obtain historical traffic data of the current intersection.

[0117] The first constructing module is configured to construct, based on the historical traffic data, a normal distribution model for describing reaction time of a driver of the preceding vehicle and a cumulative probability distribution model for describing cumulative probability of the driver of the preceding vehicle starting at a target time, so as to combine the normal distribution model and the cumulative probability distribution model to constitute the vehicle start behavior model.

[0118] Optionally, in an embodiment of the present application, the intersection start device 10 of the vehicle further comprises a second constructing module, a verification module, a third constructing module, a fourth constructing module, and a fifth constructing module.

[0119] The second constructing module is configured to define that acceleration of the preceding vehicle at a start behavior time is subject to a normal distribution, so as to model an acceleration process of the preceding vehicle as a Gaussian process to obtain a Gaussian process framework.

[0120] The verification module is configured to perform chi-square test on the Gaussian process framework based on acceleration data in the historical traffic data to obtain a chi-square test result.

[0121] The third constructing module is configured to construct an initial Gaussian process model by using the chi-square test result.

[0122] The fourth constructing module is configured to use a radial basis function kernel as a covariance function, so as to combine the covariance function and the initial Gaussian process model to obtain an actual Gaussian process model.

[0123] The fifth constructing module is configured to combine the normal distribution model, the cumulative probability distribution model, and the actual Gaussian process model to constitute the vehicle start behavior model.

[0124] Optionally, in an embodiment of the present application, the intersection starting device 10 of the vehicle further comprises a third obtaining module and a matching module.

[0125] The third obtaining module is configured to obtain a power source type of the vehicle.

[0126] The matching module is configured to match the vehicle energy consumption model based on the power source type.

[0127] Optionally, in an embodiment of the present application, the intersection starting device 10 of the vehicle further comprises a creating module, a sixth constructing module, a seventh constructing module and an eighth constructing module.

[0128] The creating module is configured to create a virtual preceding vehicle satisfying a preset characteristic condition.

[0129] The sixth constructing module is configured to construct a stochastic model predictive control framework with the virtual preceding vehicle as a tracking target.

[0130] The seventh constructing module is configured to construct a chance constraint by combining a start probability of the preceding vehicle as an uncertainty constraint and a historical speed sequence of the preceding vehicle in the historical traffic data.

[0131] The eighth constructing module is configured to construct a vehicle safe passing model by combining the stochastic model predictive control framework and the chance constraint.

[0132] Optionally, in an embodiment of the present application, the intersection starting device 10 of the vehicle further comprises a fourth obtaining module, a judging module and an updating module.

[0133] The fourth obtaining module is configured to obtain periodic driving data of the vehicle every preset time length.

[0134] The judging module is configured to judge whether the vehicle satisfies a preset updating condition based on the periodic driving data, historical intersection traffic data and historical driving data of a driver of the vehicle.

[0135] The updating module is configured to update the intersection starting model by using the periodic driving data in a case where the preset updating condition is satisfied.

[0136] It should be noted that the above description of the intersection starting method of the vehicle is also applicable to the intersection starting device of the vehicle, which will not be described here.

[0137] The intersection starting device of the vehicle provided by the embodiment of the present application can model the intersection starting behavior, more comprehensively consider the influence of the change of the starting stage on the intersection passing, introduce the vehicle safety passing model to process the uncertainty in the starting process, and finally combine the energy consumption model to realize more efficient energy-saving driving, so as to fully utilize the traffic information, significantly improve the intersection passing efficiency of the vehicle and reduce the energy consumption, and improve the driving experience of the driver. Thus, the technical problem in the prior art that the car-following model is difficult to accurately describe the non-steady-state driving behavior, so that the model is difficult to obtain an accurate vehicle starting control strategy according to the actual scene, is solved, which is not conducive to improving the intersection passing efficiency and has a great safety risk.

[0138] Figure 4 The vehicle provided by the embodiment of the present application is shown in the structural schematic diagram. The vehicle can include: The memory 401, the processor 402 and the computer program stored in the memory 401 and executable on the processor 402.

[0139] The processor 402 implements the intersection starting method of the vehicle provided in the above embodiment when executing the program.

[0140] Further, the vehicle further includes: The communication interface 403 is used for communication between the memory 401 and the processor 402.

[0141] The memory 401 is used for storing the computer program executable on the processor 402.

[0142] The memory 401 can include a high-speed RAM memory, and can also include a non-volatile memory, for example, at least one disk memory.

[0143] If the memory 401, the processor 402 and the communication interface 403 are independently implemented, the communication interface 403, the memory 401 and the processor 402 can be connected with each other through a bus and complete the communication between each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, Figure 4 Only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0144] Optionally, in a specific implementation, if the memory 401, the processor 402 and the communication interface 403 are integrated on a chip, the memory 401, the processor 402 and the communication interface 403 can complete the communication with each other through an internal interface.

[0145] The processor 402 can be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement one or more embodiments of the present application.

[0146] The embodiment further provides a computer readable storage medium, having stored thereon a computer program, which, when executed by a processor, implements the intersection starting method of the vehicle as above.

[0147] The embodiment further provides a computer program product, comprising a computer program, which, when executed by a processor, implements the intersection starting method of the vehicle as provided by the embodiment of the present application.

[0148] In the description of the present specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or N embodiments or examples. In addition, the person skilled in the art can combine and combine the different embodiments or examples described in the present specification and the features of the different embodiments or examples, without contradiction.

[0149] In addition, the terms "first", "second" are only for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include at least one of the features. In the description of the present application, the meaning of "N" is at least two, for example, two, three, etc., unless otherwise specifically limited.

[0150] Any processes or methods described in the flowcharts or otherwise described herein can be understood as representing a module, segment, or portion of code that includes one or N steps for implementing the specified logical functions or processes. The scope of preferred embodiments of the present application encompasses alternatives that implement the functions in different orders, or use different functions, or combine functions, or use the functions with the opposite inputs or outputs. These alternatives will be apparent to those skilled in the art based on the teachings of the present application.

[0151] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a list of executable instructions for implementing the logic function, and can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor- containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. For purposes of this specification, a "computer-readable medium" can be any apparatus that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. The computer-readable medium can be a product of the manufacturing and / or processing, and can be a machine-readable storage medium (alternatively, the medium can be a machine-readable signal medium). The computer-readable medium can be, for example, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device. More specific examples (a non-exhaustive list) of the computer-readable medium include the following: an electrical connection (electronic) having one or more wires, a portable computer diskette (magnetic), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber (optical), and a portable compact disc read-only memory (CDROM). Note that the computer-readable medium can even be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, via, for instance, optical scanning of the paper or other medium, then compiled, interpreted, or otherwise processed in a suitable manner, if necessary, and then stored in a computer memory.

[0152] It should be understood that aspects of the present application can be implemented in hardware, software, firmware, or combinations thereof. In the above embodiments, the N steps or methods can be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. If implemented in hardware, as in another embodiment, any of the following technologies, or combinations thereof, can be used: discrete logic circuitry having logic gates for implementing logic functions on data signals, application specific integrated circuits having appropriate combinational logic gates, programmable gate arrays (PGA), field programmable gate arrays (FPGA), and the like.

[0153] Those skilled in the art of the present technology can understand that all or part of the steps carried out by the above-mentioned embodiment method can be completed by programs instructing related hardware, and the programs can be stored in a computer readable storage medium. When the program is executed, it includes one of the steps of the method embodiment or a combination thereof.

[0154] In addition, each functional unit in each embodiment of the present application can be integrated into one processing module, or each unit can exist physically alone, or two or more units can be integrated into one module. The integrated module can be realized in the form of hardware or in the form of a software functional module. When the integrated module is realized in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer readable storage medium.

[0155] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it should be understood that the above-mentioned embodiments are exemplary and cannot be understood as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above-mentioned embodiments within the scope of the present application.

Claims

1. A method of starting at an intersection for a vehicle, characterized by, The method comprises the following steps: In the case that the vehicle meets the preset intersection starting condition, traffic data of the current intersection is acquired; The traffic data and the current state data of the vehicle are input into a pre-constructed intersection starting model to obtain an intersection starting strategy of the vehicle, wherein the intersection starting model is composed of a vehicle starting behavior model for predicting a starting behavior of a preceding vehicle, a vehicle energy consumption model, and a vehicle safe passing model for processing starting uncertainty of the preceding vehicle; The intersection starting strategy is used to control the vehicle to pass through the current intersection.

2. The method of claim 1, wherein, Before the traffic data and the current state data of the vehicle are input into the pre-constructed intersection starting model, the following steps are further included: Historical traffic data of the current intersection is acquired; Based on the historical traffic data, a normal distribution model for describing a reaction time of a driver of the preceding vehicle and a cumulative probability distribution model for describing a cumulative probability of the driver of the preceding vehicle starting at a target time are constructed to combine the normal distribution model and the cumulative probability distribution model to constitute the vehicle starting behavior model.

3. The method of claim 2, wherein, Before the traffic data and the current state data of the vehicle are input into the pre-constructed intersection starting model, the following steps are further included: A normal distribution of an acceleration of the preceding vehicle at a starting behavior time is defined to model an acceleration process of the preceding vehicle as a Gaussian process to obtain a Gaussian process framework; Chi-square test is performed on the Gaussian process framework based on acceleration data in the historical traffic data to obtain a chi-square test result; An initial Gaussian process model is constructed by using the chi-square test result; A radial basis function kernel is used as a covariance function to combine the covariance function and the initial Gaussian process model to obtain an actual Gaussian process model; The normal distribution model, the cumulative probability distribution model, and the actual Gaussian process model are combined to constitute the vehicle starting behavior model.

4. The method of claim 1, wherein, Before the traffic data and the current state data of the vehicle are input into the pre-constructed intersection starting model, the following steps are further included: A power source type of the vehicle is acquired; The vehicle energy consumption model is matched based on the power source type.

5. The method of claim 2, wherein, Before the traffic data and the current state data of the vehicle are input into the pre-constructed intersection starting model, the following steps are further included: A virtual preceding vehicle meeting a preset characteristic condition is created; A stochastic model predictive control framework is constructed by taking the virtual preceding vehicle as a tracking target; An opportunity constraint is constructed by combining a starting probability of the preceding vehicle as an uncertainty constraint and a historical speed sequence of the preceding vehicle in the historical traffic data; The vehicle safe passing model is constructed by combining the stochastic model predictive control framework and the opportunity constraint.

6. The method of claim 2, wherein, Further included are: Periodic driving data of the vehicle is acquired every preset time length; Whether the vehicle meets a preset update condition is judged based on the periodic driving data, historical intersection traffic data, and historical driving data of a driver of the vehicle; In the case that the preset update condition is met, the intersection starting model is updated by using the periodic driving data.

7. An intersection start-up device for a vehicle, characterized by comprising: Included are: An acquisition module is configured to acquire traffic data of a current intersection in the case that a vehicle meets a preset intersection starting condition; The computing module is configured to input the traffic data and the current state data of the vehicle into a pre-constructed intersection start model to obtain an intersection start strategy of the vehicle, wherein the intersection start model is composed of a vehicle start behavior model for predicting a start behavior of a preceding vehicle, a vehicle energy consumption model, and a vehicle safe passing model for processing start uncertainty of the preceding vehicle. The control module is configured to control the vehicle to pass through the current intersection based on the intersection start strategy.

8. A vehicle characterized by comprising: The intersection start method of the vehicle according to any one of claims 1-6 is implemented by a processor executing a computer program stored in a memory. The program is executed by a processor to implement the intersection start method of the vehicle according to any one of claims 1-6.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed to implement the intersection start method of the vehicle according to any one of claims 1-6.

10. A computer program product comprising a computer program, characterized in that, ​

Images