Processing method of automatic driving longitudinal control calibration table

文档序号：180995 发布日期：2021-11-02 浏览：45次中文

阅读说明：本技术 一种自动驾驶纵向控制标定表的处理方法 (Processing method of automatic driving longitudinal control calibration table ) 是由李世军刘志超张雨于 2021-08-31 设计创作，主要内容包括：本发明实施例涉及一种自动驾驶纵向控制标定表的处理方法,所述方法包括：获取第一原始数据组集合；进行数据预处理生成第一训练数据组集合；训练第一油门踏板开度控制模型并标定第一纵向控制标定表；训练第一机械制动压力控制模型并标定第二纵向控制标定表；训练第一制动踏板开度控制模型并标定第三纵向控制标定表；获取纵向实时车速、纵向期望加速度、控制模式；控制模式为驱动模式时参考纵向控制标定表进行油门踏板开度估算；并基于估算数据进行纵向驱动控制；控制模式为制动模式时参考纵向控制标定表进行机械制动压力估算；参考纵向控制标定表进行制动踏板开度估算；并基于估算数据进行纵向制动控制。通过本发明提高了自动驾驶纵向控制精准度。(The embodiment of the invention relates to a processing method of an automatic driving longitudinal control calibration table, which comprises the following steps: acquiring a first original data set; carrying out data preprocessing to generate a first training data set; training a first accelerator pedal opening control model and calibrating a first longitudinal control calibration table; training a first mechanical brake pressure control model and calibrating a second longitudinal control calibration table; training a first brake pedal opening control model and calibrating a third longitudinal control calibration table; acquiring a longitudinal real-time vehicle speed, a longitudinal expected acceleration and a control mode; when the control mode is the driving mode, estimating the opening degree of the accelerator pedal by referring to a longitudinal control calibration table; and performing longitudinal drive control based on the estimation data; when the control mode is the braking mode, estimating the mechanical braking pressure by referring to a longitudinal control calibration table; estimating the opening degree of a brake pedal by referring to a longitudinal control calibration table; and performs longitudinal braking control based on the estimated data. The invention improves the longitudinal control accuracy of automatic driving.)

1. A processing method of an automatic driving longitudinal control calibration table is characterized by comprising the following steps:

acquiring a first original data set;

performing data preprocessing on the first original data group set to generate a first training data group set;

training a first accelerator pedal opening control model according to the first training data group set, and calibrating a first longitudinal control calibration table by using the first accelerator pedal opening control model which is well trained;

training a first mechanical brake pressure control model according to the first training data set, and calibrating a second longitudinal control calibration table by using the first mechanical brake pressure control model which is well trained;

training a first brake pedal opening control model according to the first training data set, and calibrating a third longitudinal control calibration table by using the first brake pedal opening control model which is well trained;

acquiring a first longitudinal real-time vehicle speed, a first longitudinal expected acceleration and a first control mode;

when the first control mode is a driving mode, estimating the opening degree of an accelerator pedal by referring to the first longitudinal control calibration table according to the first longitudinal real-time vehicle speed and the first longitudinal expected acceleration to generate first real-time estimation data; and performing automatic driving longitudinal driving control based on the first real-time estimation data;

when the first control mode is a braking mode, estimating the mechanical braking pressure by referring to the second longitudinal control calibration table according to the first longitudinal real-time vehicle speed and the first longitudinal expected acceleration to generate second real-time estimated data; according to the first longitudinal real-time vehicle speed and the first longitudinal expected acceleration, the opening degree of a brake pedal is estimated by referring to the third longitudinal control calibration table, and third real-time estimation data are generated; and performing automatic driving longitudinal braking control based on the second real-time estimated data and the third real-time estimated data.

2. The method of processing an autopilot longitudinal control calibration chart according to claim 1,

the first set of raw data sets comprises a plurality of first raw data sets; the first raw data set comprises a first time stamp t_iFirst vehicle speed v_iFirst accelerator pedal opening th_iFirst brake pedal opening eb_iFirst brake pressure bp_iFirst steering wheel angle theta_iAnd a first pitch angle pitch_i，i>0；

The first set of training data sets comprises a plurality of first training data sets; the first training data set comprising a second time stamp t_jSecond vehicle speed v_jA second acceleration a_jSecond accelerator pedal opening th_jSecond brake pedal opening eb_jSecond brake pressure bp_jSecond steering wheel angle theta_jAnd a second pitch angle pitch_j，j>0；

The network structures of the first accelerator pedal opening control model, the first mechanical braking pressure control model and the first braking pedal opening control model are all three-layer feedforward neural networks; the neuron excitation function of the three-layer feedforward neural network is a Sigmoid function, the loss function of the three-layer feedforward neural network is a root mean square error function, and the loss function of the three-layer feedforward neural network is optimized and calculated by using an Adam optimizer;

the first longitudinal control calibration table consists of a plurality of first accelerator pedal opening data elements; the column coordinate of the first accelerator pedal opening data element is vehicle speed, and the row coordinate is acceleration;

the second longitudinal control calibration table is composed of a plurality of first brake pressure data elements; the column coordinate of the first brake pressure data element is vehicle speed, and the row coordinate is acceleration;

the third longitudinal control calibration table consists of a plurality of first brake pedal opening data elements; and the column coordinate of the first brake pedal opening degree data element is the vehicle speed, and the row coordinate is the acceleration.

3. The method for processing the calibration table of the automatic driving longitudinal control according to claim 2, wherein the step of performing data preprocessing on the first raw data group set to generate a first training data group set specifically comprises:

according to a_i＝(v_i-v_i-1)/△t_i，△t_i＝t_i-t_i-1Calculating a first acceleration a corresponding to each of the first raw data sets_iAnd applying the first acceleration a_iAdding the data to the corresponding first original data group;

the first vehicle speed v for the first set of raw data sets_iFiltering the period C according to a predetermined speed₁Push buttonSpeed mean filteringWave, k ranges from 1 to C₁；

The first acceleration a for the first set of raw data sets_iAccording to a predetermined acceleration filtering period C₂Push buttonCarrying out acceleration mean filtering, wherein the value range of l is from 1 to C₂；

Calculating the first vehicle speed v_iDeleting the first original data group which is lower than a preset speed minimum threshold value;

all the first pitch angles pitch according to the first raw data set_iPush-buttonCalculating a first pitch angle meanAnd a first pitch angle standard difference pitch_stdM is the total number of the first raw data group set; and according to the mean value of the first pitch angleAnd a first pitch angle standard difference pitch_stdPush-buttonCalculating a first gradient p corresponding to each first original data set_i(ii) a And the first gradient p_iDeleting the first original data group which is larger than a preset maximum gradient threshold value;

rotating the first steering wheel by an angle theta_iDeleting the first original data group beyond a preset rotation angle range;

and taking the first original data group set with the data preprocessing completed as the first training data group set.

4. The method for processing the calibration table of the automatic driving longitudinal control according to claim 2, wherein the training the first accelerator pedal opening control model according to the first training data set and the calibrating the first longitudinal control calibration table by using the first accelerator pedal opening control model which is well trained specifically comprises:

establishing a first model three-dimensional coordinate by taking the opening degree of an accelerator pedal as an x axis, the speed as a y axis and the acceleration as a z axis;

using any one of the second accelerator pedal opening th in the first training data set_jFor x-axis coordinates, corresponding to the second vehicle speed v_jMaking scatter marks on an x-y two-dimensional plane of the three-dimensional coordinate of the first model for a y-axis coordinate to generate corresponding first scatter points

Dividing an x-y two-dimensional plane of the three-dimensional coordinate of the first model into a plurality of first grids according to a preset first accelerator pedal opening interval and a preset first vehicle speed interval;

in each first grid, calculating all first scatter points in the current gridCorresponding second acceleration a_jTo generate a corresponding first grid average accelerationComputing all of the first scatter points in the current gridCorresponding second acceleration a_jTo generate a corresponding first grid acceleration standard deviationAverage acceleration according to the first gridAnd the first grid acceleration standard deviationCalculating each of the first scatter points in the current gridCorresponding second acceleration a_jTo generate a corresponding first degree of deviation

Extracting the first degree of deviation from the first set of training data setsThe second acceleration a is not higher than a preset first deviation threshold_jThe corresponding first training data groups form a first training data group set;

sequentially extracting the second vehicle speed v from the first training data set_jAnd the corresponding second acceleration a_jExtracting the corresponding second accelerator pedal opening th as a model input_jAs model output supervision, training the first accelerator pedal opening control model;

after the model training is mature, polling each first accelerator pedal opening data element of the first longitudinal control calibration table, and taking the currently polled first accelerator pedal opening data element as a first current data element; extracting column coordinates and row coordinates of the first current data element as corresponding first input vehicle speed and first input acceleration; inputting the first input vehicle speed and the first input acceleration into the first accelerator pedal opening control model for operation to generate a corresponding first output accelerator pedal opening; and calibrating the content of the first current data element using the first output accelerator pedal opening.

5. The method for processing the calibration table of the automatic driving longitudinal control according to claim 2, wherein the training of the first mechanical brake pressure control model according to the first training data set and the calibration of the second longitudinal control calibration table using the first mechanical brake pressure control model which is well trained specifically comprise:

establishing a second model three-dimensional coordinate by taking the brake pressure as an x axis, the vehicle speed as a y axis and the acceleration as a z axis;

using any one of the second brake pressure bp in the first training data set_jFor x-axis coordinates, corresponding to the second vehicle speed v_jMaking scatter marks on an x-y two-dimensional plane of the three-dimensional coordinate of the second model for a y-axis coordinate to generate corresponding second scatter points

Dividing an x-y two-dimensional plane of the three-dimensional coordinates of the second model into a plurality of second grids according to a preset first pressure interval and a preset second vehicle speed interval;

in each of the second grids, all the second scatter points in the current grid are calculatedCorresponding second acceleration a_jTo generate a corresponding second grid average accelerationComputing all of the second scatter points in the current gridCorresponding second acceleration a_jStandard deviation of (2)To a corresponding second grid acceleration standard deviationAverage acceleration according to the second gridAnd the second grid acceleration standard deviationCalculating each of the second scatter points in the current gridCorresponding second acceleration a_jTo generate a corresponding second degree of deviation

Extracting the second degree of deviation from the first set of training data setsThe second acceleration a is not higher than a preset second deviation threshold_jThe corresponding first training data set forms a first and second training data set;

sequentially extracting the second vehicle speed v from the first and second training data sets_jAnd the corresponding second acceleration a_jExtracting the corresponding second brake pressure bp as a model input_jAs model output supervision, training the first mechanical brake pressure control model;

after the model training is mature, polling each first brake pressure data element of the second longitudinal control calibration table, and taking the currently polled first brake pressure data element as a second current data element; extracting column coordinates and row coordinates of the second current data element as corresponding second input vehicle speed and second input acceleration; inputting the second input vehicle speed and the second input acceleration into the first mechanical brake pressure control model for operation to generate corresponding first output brake pressure; and scaling the content of the second current data element using the first output brake pressure.

6. The method for processing the calibration table of the automatic driving longitudinal control according to claim 2, wherein the training of the first brake pedal opening control model according to the first training data set and the calibration of the third longitudinal control calibration table using the first brake pedal opening control model which is well trained specifically comprise:

establishing a third model three-dimensional coordinate by taking the opening of a brake pedal as an x axis, the vehicle speed as a y axis and the acceleration as a z axis;

according to any one of the second brake pedal opening eb in the first training data set_jFor x-axis coordinates, corresponding to the second vehicle speed v_jMaking scatter marks on an x-y two-dimensional plane of the three-dimensional coordinate of the third model for a y-axis coordinate to generate a corresponding third scatter

Dividing an x-y two-dimensional plane of the three-dimensional coordinate of the third model into a plurality of third grids according to a preset first brake pedal opening interval and a preset third vehicle speed interval;

in each of the third grids, all the third scatter points in the current grid are calculatedCorresponding second acceleration a_jTo generate a corresponding third grid average accelerationComputing all of the third scatter points in the current gridCorresponding second acceleration a_jTo generate a corresponding third grid acceleration standard deviationAverage acceleration according to the third gridAnd the third grid acceleration standard deviationComputing each of the third scatter points in the current gridCorresponding second acceleration a_jTo generate a corresponding third degree of deviation

Extracting the third degree of deviation from the first set of training data setsThe second acceleration a is not higher than a preset third deviation threshold_jThe corresponding first training data set forms a first third training data set;

sequentially extracting the second vehicle speed v from the first set of third training data sets_jAnd the corresponding second acceleration a_jExtracting the corresponding second brake pedal opening eb as a model input_jAs model output supervision, training the first brake pedal opening control model;

after the model training is mature, polling each first brake pedal opening data element of the third longitudinal control calibration table, and taking the currently polled first brake pedal opening data element as a third current data element; extracting column coordinates and row coordinates of the third current data element as corresponding third input vehicle speed and third input acceleration; inputting the third input vehicle speed and the third input acceleration into the first brake pedal opening control model for operation to generate a corresponding first output brake pedal opening; and scaling the content of the third current data element using the first output brake pedal opening.

7. The method of processing an autopilot longitudinal control calibration chart according to claim 1,

and realizing the accelerator pedal opening estimation, the mechanical brake pressure estimation and the brake pedal opening estimation by adopting a two-dimensional linear interpolation calculation method.

8. The processing method of the automatic driving longitudinal control calibration table according to claim 2, wherein the step of performing accelerator pedal opening estimation by referring to the first longitudinal control calibration table according to the first longitudinal real-time vehicle speed and the first longitudinal desired acceleration to generate first real-time estimation data specifically comprises:

81, judging whether a first accelerator pedal opening data element with a column coordinate matched with the first longitudinal real-time vehicle speed and a row coordinate matched with the first longitudinal expected acceleration exists in the first longitudinal control calibration table, and if so, turning to 82; if not, go to step 83;

step 82, taking the content of the matched first accelerator pedal opening data element as the first real-time estimation data; go to step 87;

step 83, in the first longitudinal control calibration table, recording the coordinates of the front row and the rear row which are closest to the first longitudinal real-time vehicle speed as a first row coordinate v_minAnd second column coordinates v_max；v_max>First longitudinal real-time vehicle speed>v_min；

Step 84, in the first longitudinal control calibration tableRecording the two line coordinates before and after the first desired longitudinal acceleration which are the closest as a first line coordinate a_minAnd a second row coordinate a_max；a_max>First desired longitudinal acceleration>a_min；

Step 85, extracting the coordinate (v) from the first longitudinal control calibration table_min,a_min)、(v_max,a_min)、(v_min,a_max) And (v)_max,a_max) Generates corresponding first parameters according to the content of the four first accelerator pedal opening data elementsSecond parameterThird parameterAnd a fourth parameter

86, according to the first parameterThe second parameterThe third parameterAnd the fourth parameterCalculating and generating the first real-time estimation data:

and step 87, outputting the first real-time estimation data as an estimation result of the accelerator pedal opening.

9. An electronic device, comprising: a memory, a processor, and a transceiver;

the processor is used for being coupled with the memory, reading and executing the instructions in the memory to realize the method steps of any one of claims 1 to 8;

the transceiver is coupled to the processor, and the processor controls the transceiver to transmit and receive messages.

10. A computer-readable storage medium having stored thereon computer instructions which, when executed by a computer, cause the computer to perform the method of any of claims 1-8.

Technical Field

The invention relates to the technical field of data processing, in particular to a processing method of an automatic driving longitudinal control calibration table.

Background

Some of the longitudinal control interfaces of an autonomous vehicle include accelerator pedal opening and brake pedal opening, some include accelerator pedal opening, brake pedal opening, and mechanical brake pressure. How to allocate the brake pedal opening and the mechanical brake pressure here belongs to the vehicle design technology and is a black box for the automatic driving developer. This therefore also presents a significant problem for the longitudinal calibration of the autopilot.

Disclosure of Invention

The invention aims to provide a processing method of an automatic driving longitudinal control calibration table, electronic equipment and a computer readable storage medium, which aim to overcome the defects of the prior art, train three longitudinal control models based on a feedforward neural network, calibrate the longitudinal control calibration table for controlling the opening degree of an accelerator pedal, the brake pressure and the opening degree of the brake pedal based on the models, and provide three real-time longitudinal control parameters in the automatic driving process based on the calibration table. According to the invention, even though a developer does not know the design details of electric braking and mechanical braking of the vehicle, the driving or braking strategy given during automatic driving can be ensured to be consistent with the effect of manual driving, so that the longitudinal control accuracy of automatic driving can be improved, and the body feeling comfort can be improved.

In order to achieve the above object, a first aspect of the embodiments of the present invention provides a method for processing an automatic driving longitudinal control calibration table, where the method includes:

acquiring a first original data set;

performing data preprocessing on the first original data group set to generate a first training data group set;

acquiring a first longitudinal real-time vehicle speed, a first longitudinal expected acceleration and a first control mode;

Preferably, the first raw data group set includes a plurality of first raw data groups; the first raw data set comprises a first time stamp t_iFirst vehicle speed v_iFirst accelerator pedal opening th_iFirst brake pedal opening eb_iFirst brake pressure bp_iFirst steering wheel angle theta_iAnd a first pitch angle pitch_i，i>0；

The first set of training data sets comprises a plurality of first training data sets; the first training data set comprising a second time stamp t_jSecond vehicle speed v_jA second acceleration a_jSecond accelerator pedal opening th_jSecond brake pedal opening eb_jSecond brakePressure bp_jSecond steering wheel angle theta_jAnd a second pitch angle pitch_j，j>0；

Preferably, the performing data preprocessing on the first raw data group set to generate a first training data group set specifically includes:

the first vehicle speed v for the first set of raw data sets_iFiltering the period C according to a predetermined speed₁Push buttonCarrying out speed average value filtering, wherein the value range of k is from 1 to C₁；

Calculating the first vehicle speed v_iDeleting the first original data group which is lower than a preset speed minimum threshold value;

rotating the first steering wheel by an angle theta_iDeleting the first original data group beyond a preset rotation angle range;

and taking the first original data group set with the data preprocessing completed as the first training data group set.

Preferably, the training of the first accelerator pedal opening control model according to the first training data set and the calibration of the first longitudinal control calibration table by using the first accelerator pedal opening control model with mature training specifically include:

establishing a first model three-dimensional coordinate by taking the opening degree of an accelerator pedal as an x axis, the speed as a y axis and the acceleration as a z axis;

Preferably, the training a first mechanical brake pressure control model according to the first training data set, and calibrating a second longitudinal control calibration table by using the first mechanical brake pressure control model with mature training specifically include:

establishing a second model three-dimensional coordinate by taking the brake pressure as an x axis, the vehicle speed as a y axis and the acceleration as a z axis;

in each of the second grids, all the second scatter points in the current grid are calculatedCorresponding second acceleration a_jTo generate a corresponding second grid average accelerationComputing all of the second scatter points in the current gridCorresponding second acceleration a_jTo generate a corresponding second grid acceleration standard deviationAverage acceleration according to the second gridAnd the second grid acceleration standard deviationCalculating each of the second scatter points in the current gridCorresponding second acceleration a_jTo generate a corresponding second degree of deviation

Preferably, the training of the first brake pedal opening control model according to the first training data set and the calibration of the third longitudinal control calibration table by using the first brake pedal opening control model with mature training specifically include:

establishing a third model three-dimensional coordinate by taking the opening of a brake pedal as an x axis, the vehicle speed as a y axis and the acceleration as a z axis;

Preferably, the accelerator pedal opening estimation, the mechanical brake pressure estimation and the brake pedal opening estimation are realized by a two-dimensional linear interpolation calculation method.

Preferably, the estimating the opening degree of the accelerator pedal by referring to the first longitudinal control calibration table according to the first longitudinal real-time vehicle speed and the first longitudinal expected acceleration to generate first real-time estimated data specifically includes:

step 82, taking the content of the matched first accelerator pedal opening data element as the first real-time estimation data; go to step 87;

Step 84, in the first longitudinal control calibration table, recording the coordinates of the two rows before and after the first longitudinal expected acceleration is the closest as a first row coordinate a_minAnd a second row coordinate a_max；a_max>First desired longitudinal acceleration>a_min；

86, according to the first parameterThe second parameterThe third parameterAnd the fourth parameterCalculating and generating the first real-time estimation data:

and step 87, outputting the first real-time estimation data as an estimation result of the accelerator pedal opening.

A second aspect of an embodiment of the present invention provides an electronic device, including: a memory, a processor, and a transceiver;

the processor is configured to be coupled to the memory, read and execute instructions in the memory, so as to implement the method steps of the first aspect;

the transceiver is coupled to the processor, and the processor controls the transceiver to transmit and receive messages.

A third aspect of embodiments of the present invention provides a computer-readable storage medium storing computer instructions that, when executed by a computer, cause the computer to perform the method of the first aspect.

The embodiment of the invention provides a processing method of an automatic driving longitudinal control calibration table, electronic equipment and a computer readable storage medium, which train three longitudinal control models based on a feedforward neural network, calibrate the longitudinal control calibration table for controlling the opening degree of an accelerator pedal, the brake pressure and the opening degree of a brake pedal based on the models, and give out three real-time longitudinal control parameters in the automatic driving process based on the calibration table. According to the invention, even though the developer does not know the design details of the electric braking and the mechanical braking of the vehicle, the driving or braking strategy given during automatic driving is consistent with the effect of manual driving, so that the longitudinal control accuracy of automatic driving is improved, and the body feeling comfort is also improved.

Drawings

Fig. 1 is a schematic diagram of a processing method of an automatic driving longitudinal control calibration table according to an embodiment of the present invention;

fig. 2 is a schematic structural diagram of an electronic device according to a second embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention clearer, the present invention will be described in further detail with reference to the accompanying drawings, and it is apparent that the described embodiments are only a part of the embodiments of the present invention, not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

An embodiment of the present invention provides a method for processing an autopilot longitudinal control calibration table, as shown in fig. 1, which is a schematic diagram of a method for processing an autopilot longitudinal control calibration table provided in an embodiment of the present invention, the method mainly includes the following steps:

step 1, acquiring a first original data set;

wherein the first set of raw data sets comprises a plurality of first raw data sets; the first raw data set comprises a first time stamp t_iFirst vehicle speed v_iFirst accelerator pedal opening th_iFirst brake pedal opening eb_iFirst brake pressure bp_iFirst steering wheel angle theta_iAnd a first pitch angle pitch_i，i>0。

The first original data group set is a data set obtained by collecting various outputs of the manual driving, and corresponding data are obtained from relevant sensors of the vehicle according to a set collection frequency during collection to form a first original data group; in each first original data group, a first time stamp t_iFor real-time acquisition time, the data may be acquired from a timing device of the vehicle; first vehicle speed v_iFor the real-time acquisition of longitudinal speed information, first accelerator pedal opening th_iFor the information of the opening degree of the accelerator pedal, the opening degree eb of the first brake pedal, which is collected in real time_iThe first brake pressure bp is the opening information of the electric brake pedal acquired in real time_iFirst steering wheel angle θ for mechanical brake pressure information collected in real time_iSteering wheel angle information is collected in real time, and the information can be obtained from a device or a sensor capable of obtaining a vehicle chassis signal, a brake device signal and a steering turn signal; first pitch angle pitch_iThis data may be obtained from the Inertial Measurement Unit (IMU) of the vehicle for real-time collected vehicle Inertial information.

The request for manual driving when the data collection is performed includes: the method needs to carry out violent acceleration driving and violent deceleration driving, so that a real-time data set capable of reflecting positive and negative large acceleration can be generated; normal acceleration driving and normal deceleration driving are required, so that a real-time data set capable of reflecting positive and negative moderate acceleration can be generated; the method needs to carry out slow acceleration driving and slow deceleration driving so as to generate a real-time data set capable of reflecting positive and negative small acceleration; driving at a constant speed is required, so that a real-time data set capable of reflecting zero acceleration can be generated; the driving time is generally not less than 30 minutes, that is, the time embodied by the first original data set is not less than 30 minutes; the data acquisition is required to be carried out on a relatively flat road, and the acquired road cannot have an obvious gradient.

Step 2, carrying out data preprocessing on the first original data group set to generate a first training data group set;

here, before training the model using the first set of raw data sets, it needs to be filtered and denoised;

the method specifically comprises the following steps: step 21, according to a_i＝(v_i-v_i-1)/△t_i，△t_i＝t_i-t_i-1Calculating first accelerations a corresponding to the respective first raw data sets_iAnd applying the first acceleration a_iAdding the data to the corresponding first original data group;

here, each of the first raw data sets in the first raw data set is raw collected data, which does not relate to acceleration information of longitudinal control, so that it is necessary to estimate corresponding acceleration according to a speed difference and a time difference between adjacent first raw data sets; after the acceleration estimation, a corresponding acceleration information, i.e. a first acceleration a, is added to the data structure of each first raw data set of the first raw data set_i；

Note that, when i is 1, since no previous sample data is involved in the calculation of the acceleration, the first acceleration a is calculated_i＝1Will be set to null;

after the data item addition in the above steps, the embodiment of the present invention will perform filter shaping on the speed and acceleration information in the first original data group set after item addition through the subsequent steps 22-23;

step 22, a first vehicle speed v for the first set of raw data sets_iFiltering the period C according to a predetermined speed₁Push buttonCarrying out speed average value filtering, wherein the value range of k is from 1 to C₁；

Here, in order to reduce an error caused by abnormal fluctuation of the speed information, that is, a speed baseline drift phenomenon, speed mean filtering for the speed information needs to be performed on the first original data group set; when speed average value filtering is carried out, the speed filtering period C is set according to the preset system parameter₁Filtering is carried out;

for example, the velocity filter period C₁5, the first raw data set comprises 10 first raw data sets, i.e. i takes a value from 1 to 10, and the first vehicle speed v of the first raw data set_iThe data sequence is formed as { v₁,v₂…v_i…v₁₀}；

Then, when performing velocity mean filtering of velocity, the front v₁To v₄The probability of baseline drift is extremely small due to short acquisition time, so that the baseline drift can be kept unchanged; from v₅The filtering is started to be carried out and,

novel v₅Is set to v'₅，

Novel v₆Is set to v'₆，

By the way of analogy, the method can be used,

novel v₁₀Is set to v'₁₀，

Step 23, a first acceleration a of the first set of raw data sets_iAccording to a predetermined acceleration filtering period C₂Push buttonCarrying out acceleration mean filtering, wherein the value range of l is from 1 to C₂；

Here, the error caused by abnormal fluctuation of acceleration information, that is, the acceleration baseline in commonA drift phenomenon, wherein acceleration mean filtering aiming at acceleration information is required to be carried out on the first original data group set; when the acceleration mean value of the acceleration is filtered, the acceleration filtering period C is set according to the preset system parameter₂Filtering is carried out;

for example, the acceleration filter period C₂For 20, the first raw data set includes 30 first raw data sets, i.e. i takes a value from 1 to 30, and the first acceleration a of the first raw data set_iThe data sequence is formed as { a₁,a₂…a_i…a₃₀}；

Then, when the acceleration average value filtering of the acceleration is performed, front a₁To a₁₉The probability of baseline drift is extremely small due to short acquisition time, so that the baseline drift can be kept unchanged; from a₂₀The filtering is started to be carried out and,

new a₂₀Is a'₂₀，

New a₂₁Is a'₂₁，

By the way of analogy, the method can be used,

new a₃₀Is a'₃₀，

After the filtering shaping in steps 22-23, the embodiment of the present invention will continue to perform noise reduction on the first original data group set in subsequent steps 24-26;

step 24, the first vehicle speed v_iDeleting the first original data group which is lower than a preset speed minimum threshold;

here, from the perspective of vehicle speed, noise in the first raw data set is filtered; the speed minimum threshold being a predetermined threshold parameter for identifying a noisy vehicle speed, e.g.Setting a speed minimum threshold value as 0; a first vehicle speed v below the speed minimum threshold_iThe vehicle speed is regarded as the noise vehicle speed, therefore, the corresponding first original data group is regarded as the noise first original data group to be filtered;

step 25, all first pitch angles pitch according to the first raw data set_iPush-buttonCalculating a first pitch angle meanAnd a first pitch angle standard difference pitch_stdM is the total number of the first original data groups of the first original data group set; and according to the first pitch angle mean valueAnd a first pitch angle standard difference pitch_stdPush-buttonCalculating a first gradient p corresponding to each first original data set_i(ii) a And the first gradient p_iDeleting the first original data group larger than a preset maximum gradient threshold;

here, from a slope perspective, noise in the first raw data set is filtered out;

the gradient information corresponding to each first raw data set, i.e., the first gradient p_iIs calculated in a manner thatWhereinh is the pitch angle mean of the first set of raw data sets,pitch_stdstandard deviation of pitch angle for first set of raw data set，

The maximum gradient threshold is a preset threshold parameter for identifying the gradient of the noise, for example, the maximum gradient threshold is set to be 1; a first gradient p greater than the maximum gradient threshold_iThe corresponding vehicle speed is regarded as the noise vehicle speed, therefore, the corresponding first original data group is regarded as the noise first original data group to be filtered;

step 26, turning the first steering wheel by angle theta_iDeleting the first original data group beyond the preset rotation angle range;

here, the noise first raw data group in the first raw data group set is filtered from the angle of the steering wheel angle; the steering angle range is a preset threshold range for identifying the steering angle of the noisy steering wheel, such as +/-delta, wherein delta is a preset maximum value of the steering angle; first steering wheel angle theta lower than the threshold range_iThe corresponding vehicle speed is regarded as the noise vehicle speed, therefore, the corresponding first original data group is regarded as the noise first original data group to be filtered;

step 27, using the first original data group set after data preprocessing as a first training data group set;

wherein the first set of training data sets comprises a plurality of first training data sets; the first training data set comprises a second time stamp t_jSecond vehicle speed v_jA second acceleration a_jSecond accelerator pedal opening th_jSecond brake pedal opening eb_jSecond brake pressure bp_jSecond steering wheel angle theta_jAnd a second pitch angle pitch_j，j>0。

Here, the first original data set after the data enhancement, filtering and noise reduction processing of the previous steps 21-26 will be used as model training data of the subsequent steps, that is, the first training data set; the data structure of the first set of training data sets is consistent with the data structure of the first set of raw data sets with acceleration data augmentations completed.

Step 3, training the first accelerator pedal opening control model according to the first training data set, and calibrating the first longitudinal control calibration table by using the first accelerator pedal opening control model which is well trained;

the network structure of the first accelerator pedal opening control model is a three-layer feedforward neural network; the neuron excitation function of the three layers of feedforward neural networks is a Sigmoid function, the loss function of the three layers of feedforward neural networks is a root mean square error function, and the loss function of the three layers of feedforward neural networks is optimized and calculated by using an Adam optimizer;

here, the first accelerator pedal opening control model is an artificial intelligence model for performing an accelerator pedal opening estimation based on an input vehicle speed and acceleration; in practical application, the first accelerator pedal opening control model can be directly used for estimating the opening of the automatic driving real-time accelerator pedal; however, the operation periods of different models are different, the operation resources of different vehicle types are also different, and sometimes the condition that the operation duration of the models cannot meet the low-delay requirement of automatic driving may occur; in order to solve the problem, after a first accelerator pedal opening control model is trained, a first longitudinal control calibration table for reflecting the corresponding relation between the vehicle speed, the acceleration and the accelerator pedal opening is calibrated in advance based on the model, and then the real-time accelerator pedal opening can be quickly estimated by referring to the first longitudinal control calibration table in the subsequent steps; here, as shown below, the training process for the first accelerator pedal opening control model is steps 31 to 36, and the calibration process for the first longitudinal control calibration table is step 37, which specifically includes:

step 31, establishing a first model three-dimensional coordinate by taking the opening degree of an accelerator pedal as an x axis, the speed as a y axis and the acceleration as a z axis;

step 32, any second accelerator pedal opening th in the first training data set_jAs x-axis coordinates, corresponding toSecond vehicle speed v_jMaking scatter marks on an x-y two-dimensional plane of the three-dimensional coordinate of the first model for the y-axis coordinate to generate corresponding first scatter points

Step 33, dividing an x-y two-dimensional plane of the three-dimensional coordinate of the first model into a plurality of first grids according to a preset first accelerator pedal opening interval and a first vehicle speed interval;

here, in the three-dimensional coordinate space of the first model, each of the first training data sets corresponds to one set (second accelerator pedal opening th)_jSecond vehicle speed v_jSecond acceleration a_j) (ii) a First scatter pointActually, the corresponding projection point of the first training data set j on the designated plane; according to preset system parameters: the purpose of dividing the x-y two-dimensional plane of the three-dimensional coordinates of the first model into a plurality of first grids for the first accelerator pedal opening interval and the first vehicle speed interval is to divide all first scatter pointsClustering is carried out;

step 34, in each first grid, calculating all first scatter points in the current gridCorresponding second acceleration a_jTo generate a corresponding first grid average accelerationComputing all first scatter points in the current gridCorresponding second acceleration a_jTo generate a corresponding first grid acceleration standard deviationAverage acceleration according to the first gridAnd first grid acceleration standard deviationCalculating each first scatter point in the current gridCorresponding second acceleration a_jTo generate a corresponding first degree of deviation

Step 35, extracting a first deviation from the first training data setSecond acceleration a not higher than a preset first deviation threshold_jThe corresponding first training data set forms a first training data set;

in the embodiment of the invention, after clustering scattered points, the first training data set is subjected to data screening once again according to the calculated acceleration deviation degree to obtain the first training data set, so that the accuracy of the training data set is further improved, and the training efficiency of the model is favorably improved;

step 36, sequentially extracting the second vehicle speed v from the first training data set_jAnd a corresponding second acceleration a_jExtracting a corresponding second accelerator pedal opening th as a model input_jAs model output supervision, training a first accelerator pedal opening control model;

here, the model is being performedDuring training, the second vehicle speed v_jSecond accelerator pedal opening th_jAs an independent variable, the second acceleration a_jTraining as a dependent variable; during training, the root mean square error function is adoptedWherein cost is a root mean square error function, N is a total number of first training data groups in the first set of training data groups,for a second acceleration a in the first set of training data sets_jMean value;

step 37, after the model training is mature, polling each first accelerator pedal opening data element of the first longitudinal control calibration table, and taking the currently polled first accelerator pedal opening data element as a first current data element; extracting column coordinates and row coordinates of the first current data element as corresponding first input vehicle speed and first input acceleration; inputting the first input vehicle speed and the first input acceleration into a first accelerator pedal opening control model for operation to generate a corresponding first output accelerator pedal opening; and scaling the content of the first current data element using the first output accelerator pedal opening.

Here, the data structure of the first longitudinal control calibration table is a two-dimensional data table entry or matrix with columns × rows as described above, each column of the calibration table corresponds to a specific vehicle speed value, each row corresponds to a specific acceleration value, and each data unit, that is, the first accelerator pedal opening data element corresponds to a specific accelerator pedal opening value;

column parameters and row parameters of the first longitudinal control calibration table are preset according to actual application requirements, and each first accelerator pedal opening data element is an estimated value estimated by using a first accelerator pedal opening control model; when the first longitudinal control calibration table is calibrated, the column coordinates of the data elements are used as a first input vehicle speed, the row coordinates of the data elements are used as a first input acceleration, then the first input vehicle speed and the first input acceleration are input into a first accelerator pedal opening control model for estimation, and the estimation result is used as the content of the current data elements for calibration.

Step 4, training the first mechanical brake pressure control model according to the first training data set, and calibrating the second longitudinal control calibration table by using the first mechanical brake pressure control model which is well trained;

the network structure of the first mechanical brake pressure control model is similar to that of the first accelerator pedal opening control model and is also a three-layer feedforward neural network; the neuron excitation function is also a Sigmoid function, the loss function is also a root mean square error function, and the Adam optimizer is used for carrying out optimization calculation on the loss function;

here, the first mechanical brake pressure control model is an artificial intelligence model for brake pressure estimation based on the input vehicle speed and acceleration; in practical application, the first mechanical brake pressure control model can be directly used for estimating the real-time brake pressure of automatic driving; however, the operation periods of different models are different, the operation resources of different vehicle types are also different, and sometimes the condition that the operation duration of the models cannot meet the low-delay requirement of automatic driving may occur; in order to solve the problem, in the embodiment of the invention, after the first mechanical brake pressure control model is trained, a second longitudinal control calibration table for reflecting the corresponding relation between the vehicle speed, the acceleration and the brake pressure is calibrated in advance based on the model, and then the real-time brake pressure can be quickly estimated by referring to the second longitudinal control calibration table in the subsequent steps; here, as shown below, the training process for the first mechanical brake pressure control model is steps 41 to 46, and the calibration process for the second longitudinal control calibration table is step 47, which specifically includes:

step 41, establishing a second model three-dimensional coordinate by taking the brake pressure as an x axis, the vehicle speed as a y axis and the acceleration as a z axis;

step 42, using any one second brake pressure bp in the first training data set_jAs x-axis coordinate, corresponding to a second vehicle speed v_jMaking scatter marks on an x-y two-dimensional plane of the three-dimensional coordinate of the second model for the y-axis coordinate to generate corresponding second scatter points

Step 43, dividing the x-y two-dimensional plane of the three-dimensional coordinates of the second model into a plurality of second grids according to a preset first pressure interval and a preset second vehicle speed interval;

here, in the three-dimensional coordinate space of the second model, each of the first training data sets in the first training data set corresponds to one set (second brake pressure bp)_jSecond vehicle speed v_jSecond acceleration a_j) (ii) a Second scatter pointActually, the corresponding projection point of the first training data set j on the designated plane; according to preset system parameters: the purpose of the first pressure interval and the second vehicle speed interval to divide the x-y two-dimensional plane of the three-dimensional coordinates of the second model into a plurality of second grids is to divide all second scatter pointsClustering is carried out;

step 44, in each second grid, calculating all second scatter points in the current gridCorresponding second acceleration a_jTo generate a corresponding second grid average accelerationComputing all second scatter points in the current gridTo a corresponding secondTwo accelerations a_jTo generate a corresponding second grid acceleration standard deviationAverage acceleration according to the second gridAnd second grid acceleration standard deviationCalculating each second scatter point in the current gridCorresponding second acceleration a_jTo generate a corresponding second degree of deviation

Step 45, extracting a second deviation degree from the first training data group setA second acceleration a not higher than a preset second deviation threshold_jThe corresponding first training data set forms a first and second training data set;

in the embodiment of the invention, after clustering scattered points, the first training data group set is subjected to data screening once again according to the calculated acceleration deviation degree to obtain the first and second training data group sets, so that the accuracy of the training data sets is further improved, and the training efficiency of the model is favorably improved;

step 46, sequentially extracting the second vehicle speed v from the first and second training data sets_jAnd a corresponding second acceleration a_jAs model input, the corresponding second brake pressure bp is extracted_jAs model output supervision, for the first machineTraining a mechanical brake pressure control model;

here, in performing the model training, the second vehicle speed v is set, similarly to step 36_jSecond brake pressure bp_jAs an independent variable, the second acceleration a_jTraining as a dependent variable; during training, the root mean square error function is adoptedWherein cost is a root mean square error function, N is a total number of first training data groups in the first set of second training data groups,for a second acceleration a in the first set of two training data sets_jMean value;

step 47, after the model training is mature, polling each first brake pressure data element of the second longitudinal control calibration table, and taking the currently polled first brake pressure data element as a second current data element; extracting column coordinates and row coordinates of a second current data element as a corresponding second input vehicle speed and a second input acceleration; inputting a second input vehicle speed and a second input acceleration into the first mechanical brake pressure control model for operation to generate corresponding first output brake pressure; and scaling the content of the second current data element using the first output brake pressure.

Here, the data structure of the second longitudinal control calibration table is a two-dimensional data table entry or matrix with columns × rows as described above, each column of the calibration table corresponds to a specific vehicle speed value, each row corresponds to a specific acceleration value, and each data unit, that is, the first brake pressure data element corresponds to a specific brake pressure value;

column parameters and row parameters of the second longitudinal control calibration table are preset according to actual application requirements, and each first brake pressure data element is an estimated value estimated by using the first mechanical brake pressure control model; when the second longitudinal control calibration table is calibrated, the column coordinates of the data elements are used as a second input vehicle speed, the row coordinates of the data elements are used as a second input acceleration, then the second input vehicle speed and the second input acceleration are input into the first mechanical brake pressure control model for estimation, and the estimation result is used as the content of the current data elements for calibration.

Step 5, training the first brake pedal opening control model according to the first training data set, and calibrating the third longitudinal control calibration table by using the first brake pedal opening control model which is well trained;

the network structure of the first brake pedal opening control model is similar to that of the first accelerator pedal opening control model and is also a three-layer feedforward neural network; the neuron excitation function is also a Sigmoid function, the loss function is also a root mean square error function, and the Adam optimizer is used for carrying out optimization calculation on the loss function;

the third longitudinal control calibration table consists of a plurality of first brake pedal opening data elements; the column coordinate of the first brake pedal opening degree data element is vehicle speed, and the row coordinate is acceleration;

here, the first brake pedal opening control model is an artificial intelligence model for performing brake pedal opening estimation based on the input vehicle speed and acceleration; in practical application, the first brake pedal opening control model can be directly used for estimating the automatic driving real-time brake pedal opening; however, the operation periods of different models are different, the operation resources of different vehicle types are also different, and sometimes the condition that the operation duration of the models cannot meet the low-delay requirement of automatic driving may occur; in order to solve the problem, after the first brake pedal opening control model is trained, a third longitudinal control calibration table for reflecting the corresponding relation between the vehicle speed, the acceleration and the brake pedal opening is calibrated in advance based on the model, and then the real-time brake pedal opening can be quickly estimated by referring to the third longitudinal control calibration table in the subsequent steps; here, as shown below, the training process for the first brake pedal opening control model is steps 51 to 45, and the calibration process for the third longitudinal control calibration table is step 57, which specifically includes:

step 51, establishing a third model three-dimensional coordinate by taking the opening degree of a brake pedal as an x axis, the speed as a y axis and the acceleration as a z axis;

step 52, any second brake pedal opening eb in the first training data set_jAs x-axis coordinate, corresponding to a second vehicle speed v_jMaking scatter marks on an x-y two-dimensional plane of the three-dimensional coordinate of the third model for the y-axis coordinate to generate a corresponding third scatter

Step 53, dividing an x-y two-dimensional plane of a three-dimensional coordinate of the third model into a plurality of third grids according to a preset first brake pedal opening interval and a preset third vehicle speed interval;

here, in the third model three-dimensional coordinate space, each of the first training data sets corresponds to one set (second brake pedal opening eb_jSecond vehicle speed v_jSecond acceleration a_j) (ii) a The third scatter pointActually, the corresponding projection point of the first training data set j on the designated plane; according to preset system parameters: the purpose of dividing the x-y two-dimensional plane of the three-dimensional coordinate of the third model into a plurality of third grids for the first brake pedal opening interval and the third vehicle speed interval is to divide all third scattered pointsClustering is carried out;

step 54, in each third grid, calculating all third scatter points in the current gridCorresponding second acceleration a_jTo generate a corresponding third grid average accelerationComputing all third nodes in the current gridScattered pointCorresponding second acceleration a_jTo generate a corresponding third grid acceleration standard deviationAverage acceleration according to the third gridAnd third grid acceleration standard deviationCalculating each third scatter point in the current gridCorresponding second acceleration a_jTo generate a corresponding third degree of deviation

Step 55, extracting a third deviation from the first set of training data setsA second acceleration a not higher than a preset third deviation threshold_jA corresponding first training data set, forming a first third training data set;

in the embodiment of the invention, after clustering scattered points, the first training data group set is subjected to data screening once again according to the calculated acceleration deviation degree to obtain a first third training data group set, so that the accuracy of the training data set is further improved, and the training efficiency of the model is favorably improved;

step 56, sequentially extracting the second vehicle speed v from the first set of training data sets_jAnd a corresponding second acceleration a_jExtracting corresponding second brake pedal opening eb as model input_jAs model output supervision, training a first brake pedal opening control model;

here, in performing the model training, the second vehicle speed v is set, similarly to step 36_jSecond brake pedal opening eb_jAs an independent variable, the second acceleration a_jTraining as a dependent variable; during training, the root mean square error function is adoptedWherein cost is a root mean square error function, N is a total number of the first training data groups in the first set of three training data groups,for the second acceleration a in the first set of three training data sets_jMean value;

step 57, after the model training is mature, polling each first brake pedal opening data element of the third longitudinal control calibration table, and taking the currently polled first brake pedal opening data element as a third current data element; extracting column coordinates and row coordinates of a third current data element as a corresponding third input vehicle speed and a third input acceleration; inputting a third input vehicle speed and a third input acceleration into the first brake pedal opening control model for operation to generate a corresponding first output brake pedal opening; and scaling the content of the third current data element using the first output brake pedal opening.

Here, the data structure of the third longitudinal control calibration table is a two-dimensional data table entry or matrix with columns × rows as described above, each column of the calibration table corresponds to a specific vehicle speed value, each row corresponds to a specific acceleration value, and each data unit, that is, the first brake pedal opening data element corresponds to a specific brake pedal opening value;

column parameters and row parameters of the third longitudinal control calibration table are preset according to actual application requirements, and each first brake pedal opening data element is an estimated value estimated by using the first brake pedal opening control model; and when the third longitudinal control calibration table is calibrated, taking the column coordinate of the data element as a third input vehicle speed and the row coordinate as a third input acceleration, then inputting the third input vehicle speed and the third input acceleration into the first brake pedal opening control model for estimation, and calibrating by taking the estimation result as the content of the current data element.

After the calibration of the model training and the longitudinal control calibration in the above steps 3-5, the first, second and third longitudinal calibration tables after the calibration can be used in the following steps 6-10, and the corresponding accelerator pedal opening, brake pressure and brake pedal opening are estimated in real time according to the obtained real-time vehicle speed and acceleration, and the vehicle is longitudinally controlled in real time according to the estimation result.

And 6, acquiring a first longitudinal real-time vehicle speed, a first longitudinal expected acceleration and a first control mode.

Here, the first longitudinal real-time vehicle speed is a real-time longitudinal vehicle speed of the autonomous vehicle; the first desired longitudinal acceleration is a desired longitudinal acceleration value given by a control system of the autonomous vehicle; the first control mode includes both a driving mode, i.e., an acceleration mode, and a braking mode, i.e., a deceleration mode.

Step 7, when the first control mode is a driving mode, estimating the opening degree of an accelerator pedal by referring to a first longitudinal control calibration table according to a first longitudinal real-time vehicle speed and a first longitudinal expected acceleration to generate first real-time estimation data; and performing automatic driving longitudinal driving control based on the first real-time estimation data;

the method specifically comprises the following steps: step 71, when the first control mode is the driving mode, estimating the opening degree of an accelerator pedal by referring to a first longitudinal control calibration table according to a first longitudinal real-time vehicle speed and a first longitudinal expected acceleration to generate first real-time estimation data;

the method specifically comprises the following steps: step 711, judging whether a first accelerator pedal opening data element with a column coordinate matched with the first longitudinal real-time vehicle speed and a row coordinate matched with the first longitudinal expected acceleration exists in the first longitudinal control calibration table, and if so, turning to step 712; if not, go to step 713;

step 712, using the content of the matched first accelerator pedal opening data element as first real-time estimation data; go to step 717;

here, in the first longitudinal control calibration table, the first longitudinal real-time vehicle speed corresponds to a specific column parameter, and the first longitudinal expected acceleration corresponds to a specific row parameter, and then the content of the first accelerator pedal opening data element corresponding to the column parameter + the row parameter can be directly extracted as the estimation result, that is, the first real-time estimation data; next, go to step 77 to output the estimation result;

713, recording the coordinates of the front row and the rear row which are closest to the first longitudinal real-time vehicle speed in the first longitudinal control calibration table as first one-row coordinatesAnd first and second coordinates

Here, in the first longitudinal control calibration table, the first longitudinal real-time vehicle speed does not correspond to a specific column parameter or the first longitudinal expected acceleration does not correspond to a specific row parameter, and then the contents of four nearest first accelerator pedal opening data elements need to be obtained in the table for estimation; the first one-column coordinateAnd first and second coordinatesI.e. the four most recent dataMarking the column of the element;

step 714, in the first longitudinal control calibration table, the coordinates of the two rows before and after the first longitudinal expected acceleration is the closest are recorded as the first one-row coordinatesAnd first parallel coordinate

Here, the first one-line coordinatesAnd first parallel coordinateI.e. the row that is the four most recent data elements is marked;

step 715, extracting coordinate 1 from the first longitudinal control calibration tableCoordinate 2 Coordinate 3And coordinates 4Generates corresponding first parameters according to the content of the four first accelerator pedal opening data elementsFirst and second parametersFirst three parametersAnd a first four parameter

Here, the first parameter mentioned aboveFirst and second parametersFirst three parametersAnd a first four parameterI.e. the content of the four most recent data elements;

step 716, according to the first parameterFirst and second parametersFirst three parametersAnd a first four parameterCalculating to generate first real-time estimation data:

S₁for the first real-time estimation of the data,

here, the embodiment of the present invention estimates according to the two-dimensional linear interpolation algorithm based on the four nearest data elements, so as to obtain the estimated value of the accelerator pedal opening, that is, the first real-time estimated data; next, the estimation result is outputted in step 717;

step 717, outputting the first real-time estimation data as an estimation result of the accelerator pedal opening degree;

and 72, performing automatic driving longitudinal driving control based on the first real-time estimation data.

Here, the first real-time estimated data is the estimated accelerator pedal opening, and when the automatic driving longitudinal driving control is performed based on the first real-time estimated data, the accelerator pedal opening of the vehicle is controlled using the first real-time estimated data for acceleration.

Step 8, when the first control mode is a braking mode, estimating the mechanical braking pressure by referring to a second longitudinal control calibration table according to the first longitudinal real-time vehicle speed and the first longitudinal expected acceleration to generate second real-time estimated data; according to the first longitudinal real-time vehicle speed and the first longitudinal expected acceleration, estimating the opening degree of the brake pedal by referring to a third longitudinal control calibration table to generate third real-time estimation data; and performing automatic driving longitudinal braking control based on the second real-time estimation data and the third real-time estimation data;

step 81, when the first control mode is a braking mode, estimating the mechanical braking pressure by referring to a second longitudinal control calibration table according to the first longitudinal real-time vehicle speed and the first longitudinal expected acceleration to generate second real-time estimated data;

the method specifically comprises the following steps: step 811, judging whether a first brake pressure data element with a column coordinate matched with the first longitudinal real-time vehicle speed and a row coordinate matched with the first longitudinal expected acceleration exists in the second longitudinal control calibration table, and if so, turning to step 812; if not, go to step 813;

step 812, taking the content of the matched first brake pressure data element as second real-time estimation data; go to step 817;

here, in the second longitudinal control calibration table, the first longitudinal real-time vehicle speed corresponds to a specific column parameter, and the first longitudinal expected acceleration corresponds to a specific row parameter, so that the content of the first brake pressure data element corresponding to the column parameter + the row parameter can be directly extracted as the estimation result, that is, the second real-time estimation data; next, go to step 87 to output the estimation result;

step 813, in the second longitudinal control calibration table, the two front and back row coordinates closest to the first longitudinal real-time vehicle speed are recorded as the second row coordinateAnd second column coordinates

Here, in the second longitudinal control calibration table, the first longitudinal real-time vehicle speed does not correspond to a specific column parameter or the first longitudinal expected acceleration does not correspond to a specific row parameter, and then the contents of four nearest first brake pressure data elements need to be obtained in the table for estimation; the second row coordinateAnd second column coordinatesI.e. the columns of the four most recent data elements are marked;

step 814, in the second longitudinal control calibration table, the nearest previous expected acceleration of the first longitudinal is obtainedThe last two row coordinates are noted as the second row coordinatesAnd second two-line coordinates

Here, the second row coordinateAnd second two-line coordinatesI.e. the row that is the four most recent data elements is marked;

step 815, extracting the coordinate 1 from the second longitudinal control calibration tableCoordinate 2 Coordinate 3And coordinates 4Generates corresponding second parametersSecond two parametersSecond three parametersAnd a second four parameter

Here, the second parameterSecond two parametersSecond three parametersAnd a second four parameterI.e. the content of the four most recent data elements;

step 816, according to the second parameterSecond two parametersSecond three parametersAnd a second four parameterAnd calculating to generate second real-time estimation data:

S₂for the second real-time estimation of the data,

here, the embodiment of the present invention performs estimation according to a two-dimensional linear interpolation algorithm based on the four nearest data elements, thereby obtaining an estimated value of the brake pressure, that is, second real-time estimated data; next, the estimation result is output in step 817;

817, outputting the second real-time estimation data as an estimation result of the mechanical brake pressure estimation;

step 82, estimating the opening degree of a brake pedal by referring to a third longitudinal control calibration table according to the first longitudinal real-time vehicle speed and the first longitudinal expected acceleration to generate third real-time estimation data;

the method specifically comprises the following steps: step 821, judging whether a first brake pedal opening data element with a column coordinate matched with the first longitudinal real-time vehicle speed and a row coordinate matched with the first longitudinal expected acceleration exists in the third longitudinal control calibration table, and if so, turning to step 822; if not, go to step 823;

step 822, taking the content of the matched first brake pedal opening data element as third real-time estimation data; go to step 827;

here, in the third longitudinal control calibration table, the first longitudinal real-time vehicle speed corresponds to a specific column parameter, and the first longitudinal expected acceleration corresponds to a specific row parameter, and then the content of the first brake pedal opening data element corresponding to the column parameter + the row parameter can be directly extracted as the estimation result, that is, the third real-time estimation data; next, go to step 97 to output the estimation result;

step 823, in the third longitudinal control calibration table, recording the front and back row coordinates closest to the first longitudinal real-time vehicle speed as a third row coordinateAnd third two-column coordinates

Here, in the third longitudinal control calibration table, the first longitudinal real-time vehicle speed does not correspond to a specific column parameter or the first longitudinal expected acceleration does not correspond to a specific row parameter, and then the contents of four nearest first brake pedal opening data elements need to be obtained in the table for estimation; the third row coordinateAnd third two-column coordinatesI.e. the columns of the four most recent data elements are marked;

step 824, in the third longitudinal control calibration table, recording the coordinates of the two rows before and after the first longitudinal expected acceleration is the closest as the third row coordinateAnd third two line coordinates

Here, the third row coordinateAnd third two line coordinatesI.e. the row that is the four most recent data elements is marked;

step 825, from the third verticalExtracting coordinate 1 from control calibration tableCoordinate 2 Coordinate 3And coordinates 4Generates corresponding third first parameters according to the content of the four first brake pedal opening degree data elementsThird two parametersThird three parametersAnd a third four parameter

Here, the third parameterThird two parametersThird three parametersAnd a third four parameterI.e. the four mostThe content of the recent data element;

826, according to the third parameterThird two parametersThird three parametersAnd a third four parameterCalculating to generate third real-time estimation data:

S₃for the third real-time estimation of the data,

here, the embodiment of the present invention performs estimation according to a two-dimensional linear interpolation algorithm based on the four nearest data elements, thereby obtaining an estimated value of the opening degree of the brake pedal, that is, third real-time estimation data; next, the estimation result is outputted in step 827;

step 827, outputting the third real-time estimation data as an estimation result of the brake pedal opening degree estimation;

and step 83, performing automatic driving longitudinal braking control based on the second real-time estimation data and the third real-time estimation data.

Here, the second real-time estimated data is the estimated brake pressure, and the third real-time estimated data is the estimated brake pedal opening degree; and when the longitudinal braking control of the automatic driving is carried out based on the second real-time estimation data and the third real-time estimation data, the mechanical braking pressure of the vehicle is controlled by using the second real-time estimation data, and the opening degree of an electric brake pedal of the vehicle is controlled by using the third real-time estimation data, so that the purpose of combined braking deceleration is achieved. Because the electric brake has the function of energy recovery, the comfort level can be ensured while the effective speed reduction is taken into consideration.

It should be noted that some autonomous vehicles do not have an electric braking function, and thus, for such vehicles, the autonomous longitudinal braking control can be performed using only the second real-time estimated data.

Fig. 2 is a schematic structural diagram of an electronic device according to a second embodiment of the present invention. The electronic device may be a terminal device or a server for implementing the method of the embodiment of the present invention, or may be a terminal device or a server connected to the terminal device or the server for implementing the method of the embodiment of the present invention. As shown in fig. 2, the electronic device may include: a processor 301 (e.g., a CPU), a memory 302, a transceiver 303; the transceiver 303 is coupled to the processor 301, and the processor 301 controls the transceiving operation of the transceiver 303. Various instructions may be stored in memory 302 for performing various processing functions and implementing the processing steps described in the foregoing method embodiments. Preferably, the electronic device according to an embodiment of the present invention further includes: a power supply 304, a system bus 305, and a communication port 306. The system bus 305 is used to implement communication connections between the elements. The communication port 306 is used for connection communication between the electronic device and other peripherals.

The system bus 305 mentioned in fig. 2 may be a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, or the like. The system bus may be divided into an address bus, a data bus, a control bus, and the like. For ease of illustration, only one thick line is shown, but this does not mean that there is only one bus or one type of bus. The communication interface is used for realizing communication between the database access device and other equipment (such as a client, a read-write library and a read-only library). The Memory may include a Random Access Memory (RAM) and may also include a Non-Volatile Memory (Non-Volatile Memory), such as at least one disk Memory.

The Processor may be a general-purpose Processor, including a Central Processing Unit (CPU), a Network Processor (NP), a Graphics Processing Unit (GPU), and the like; but also Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs) or other Programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components.

It should be noted that the embodiment of the present invention also provides a computer-readable storage medium, which stores instructions that, when executed on a computer, cause the computer to execute the method and the processing procedure provided in the above-mentioned embodiment.

The embodiment of the present invention further provides a chip for executing the instructions, where the chip is configured to execute the processing steps described in the foregoing method embodiment.

Those of skill would further appreciate that the various illustrative components and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative components and steps have been described above generally in terms of their functionality in order to clearly illustrate this interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.

The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied in hardware, a software module executed by a processor, or a combination of the two. A software module may reside in Random Access Memory (RAM), memory, Read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

The above-mentioned embodiments are intended to illustrate the objects, technical solutions and advantages of the present invention in further detail, and it should be understood that the above-mentioned embodiments are merely exemplary embodiments of the present invention, and are not intended to limit the scope of the present invention, and any modifications, equivalent substitutions, improvements and the like made within the spirit and principle of the present invention should be included in the scope of the present invention.

27页详细技术资料下载

上一篇：一种医用注射器针头装配设备

下一篇：一种预测车辆运动轨迹的处理方法

Processing method of automatic driving longitudinal control calibration table

相关技术

网友询问留言