A model-free longitudinal control method for autonomous driving vehicles with multi-mode redundancy

Abstract

A multi-mode redundant model-free longitudinal control method for an autonomous driving vehicle includes: collecting a longitudinal control signal source; establishing an offline driving torque database and an offline braking pressure database; designing a model-free vehicle speed control strategy, calculating the expected driving torque of the vehicle speed and the expected braking pressure of the vehicle speed; designing a model-free acceleration control strategy, calculating the expected driving torque of acceleration; designing a model-free deceleration control strategy, calculating the expected braking pressure of deceleration; designing a model-free longitudinal control mode arbitration strategy according to the expected driving torque of the vehicle speed, the expected braking pressure of the vehicle speed, the expected driving torque of acceleration and the expected braking pressure of deceleration, outputting the requested driving torque and the requested braking pressure, and realizing longitudinal operation. The present invention can solve the stability problem of the longitudinal control of an autonomous driving vehicle when the vehicle model parameters are missing or the model parameters are uncertain, and prevent unexpected driving or braking caused by interference from factors such as program abnormalities and communication blockages.

Description

Technical Field

[0001] The present invention relates to the field of automobile technology, and in particular to a model-free longitudinal control method for a multi-mode redundant automatic driving vehicle. Background Art

[0002] As a new technology that is expected to ease traffic congestion, reduce the accident rate and improve work efficiency, autonomous driving vehicle technology has become a current research hotspot. At present, although autonomous driving vehicle technology has made certain breakthroughs, there are still certain defects and drawbacks in the actual application effect. The most obvious point is that autonomous driving vehicles are far from being as stable, safe and efficient as experienced drivers.

[0003] In particular, the existing stable, safe and efficient realization of autonomous driving longitudinal control is often based on accurate physical models. In the absence of vehicle model parameters or model parameter uncertainty, the stability, safety and efficiency of longitudinal control are difficult to guarantee. If you want to obtain the accurate physical model parameters of each autonomous driving vehicle, such as tire lateral stiffness, drag coefficient, road spectrum, etc., you need to invest tens of millions of dollars to complete the relevant tests and sensor installation, which is obviously not conducive to the rapid implementation and cost control of autonomous driving vehicle products. In addition, autonomous driving control systems are often non-real-time, high-concurrency operating systems, which are prone to program anomalies, communication blockages and other factors, which can easily interfere with the longitudinal control of autonomous driving, resulting in unexpected driving or braking.

[0004] Therefore, how to provide a method that does not rely on an accurate vehicle model, solves the stability problem of the longitudinal control of the autonomous driving vehicle when the vehicle model parameters are missing or the model parameters are uncertain, and prevents unexpected driving or braking caused by interference from factors such as program abnormalities and communication blockages, is a problem that technical personnel in this field need to solve. Summary of the invention

[0005] The purpose of the present invention is to provide a multi-mode redundant model-free longitudinal control method for an autonomous driving vehicle, so as to solve the stability problem of the longitudinal control of the autonomous driving vehicle when the vehicle model parameters are missing or the model parameters are uncertain, and to prevent unexpected driving or braking caused by interference from factors such as program anomalies and communication congestion.

[0006] A multi-mode redundant model-free longitudinal control method for an autonomous driving vehicle, comprising:

[0007] Step S1, collecting longitudinal control signal sources of different types and sources;

[0008] Step S2, establishing an offline driving torque database and an offline braking pressure database based on the longitudinal control signal source;

[0009] Step S3, designing a model-free vehicle speed control strategy based on the longitudinal control signal source and combining the offline driving torque database and the offline braking pressure database to calculate the vehicle speed expected driving torque and the vehicle speed expected braking pressure;

[0010] Step S4, based on the longitudinal control signal source and in combination with the offline driving torque database, a model-free acceleration control strategy is designed to calculate the acceleration expected driving torque;

[0011] Step S5, based on the longitudinal control signal source and in combination with the offline brake pressure database, a model-free deceleration control strategy is designed to calculate the deceleration expected brake pressure;

[0012] Step S6, designing a model-free longitudinal control mode arbitration strategy based on the expected driving torque at vehicle speed, the expected braking pressure at vehicle speed, the expected driving torque at acceleration, and the expected braking pressure at deceleration, outputting the requested driving torque and the requested braking pressure, and realizing the longitudinal operation of the autonomous driving vehicle.

[0013] Compared with the prior art, the present invention has the following beneficial effects:

[0014] (1) The control method of the present invention does not rely on the precise physical model of the autonomous driving vehicle system. In the case of a real vehicle, the longitudinal stability control of the autonomous driving vehicle can be achieved by importing limited real vehicle experimental data. This can effectively solve the problem of longitudinal control stability when the autonomous driving vehicle model parameters are missing or the model parameters are uncertain.

[0015] (2) The feedforward control and conditional PID feedback control designed in the present invention are integrated to perform driving control and braking control, which can effectively suppress vehicle vibration caused by changes in driving torque or braking pressure, avoid the integral saturation phenomenon caused by continuous accumulation of negative driving torque or negative braking pressure, and contribute to the longitudinal stability control of the autonomous driving vehicle.

[0016] (3) The model-free longitudinal control mode arbitration designed by the present invention not only takes into account the unexpected driving or braking caused by interference from factors such as program anomalies and communication congestion, but also takes into account the accuracy error of the actual calculation and the difference in the calculation mode. Therefore, it can solve the problems of safety and effectiveness and has strong applicability. It can be applied to the autonomous valet parking induction control of a single intelligent vehicle or multiple intelligent vehicles at the same time, which is more efficient.

[0017] (4) Since the present invention is not affected by the accuracy of the vehicle physical model, it has strong practicality and flexibility and can be efficiently applied to various types of autonomous driving vehicles, including autonomous driving cars, autonomous driving buses, autonomous driving special vehicles, etc., which can speed up the development progress and reduce R&D costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flow chart of a multi-mode redundant model-free longitudinal control method for an autonomous driving vehicle provided by the present invention;

[0019] Figure 2 is a schematic diagram of an exemplary result of a step speed real vehicle test;

[0020] Figure 3 It is another exemplary schematic diagram of the results of a step speed actual vehicle test. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] See also Figure 1 , an embodiment of the present invention provides a multi-mode redundant model-free longitudinal control method for an autonomous driving vehicle, comprising steps S1 to S6:

[0023] Step S1, collecting longitudinal control signal sources of different types and sources.

[0024] Specifically, the longitudinal control signal source includes the actual acceleration a e , expected acceleration a r 、Expected accelerator pedal opening a p , actual vehicle speed v e 、Expected vehicle speed v r , actual driving torque T e , actual brake pressure F e , Expected brake pedal opening b p , expected deceleration b r , actual deceleration b e .

[0025] Step S2: establishing an offline driving torque database and an offline braking pressure database based on the longitudinal control signal source.

[0026] In step S2, three offline driving torque databases are established, namely, a vehicle speed-driving torque database, a vehicle speed-acceleration-driving torque database, and an accelerator pedal opening-driving torque database;

[0027] The vehicle speed-driving torque database is a one-dimensional discrete data table T vf (v), the interpolation method uses cubic spline curve to achieve third-order continuity, T vfThe input of (v) is the vehicle speed, and the output is the vehicle speed torque obtained by looking up the table;

[0028] The vehicle speed-acceleration-drive torque database is a two-dimensional discrete data table T af (v, a), the interpolation method uses a cubic spline curve to achieve third-order continuity, T af The input of (v, a) is vehicle speed and acceleration, and the output is the acceleration torque obtained by looking up the table;

[0029] The accelerator pedal opening-driving torque database is a one-dimensional discrete data table T pf (a), interpolation method uses hyperbolic tangent function, T pf The input of (a) is the accelerator pedal opening, and the output is the pedal torque obtained by looking up the table.

[0030] Under the preset environmental conditions, in this embodiment, specifically under the environmental conditions of an empty vehicle, a dry asphalt straight road with an adhesion coefficient of 0.8 and no wind, the rated driving torque range [T min , T max ] is divided into n+1 equal parts. In this embodiment, n is 100, T min is the minimum rated driving torque, T max is the maximum rated driving torque, and the driving torque step length ΔT is obtained, ΔT=(T max -T min ) / n, according to the test drive torque from T min At the beginning, the driving torque is increased by a step size ΔT each time until the test driving torque increases to T max Until then, record the time when the vehicle reaches a constant speed in a steady state after each test drive torque is issued, so as to obtain the test drive torque sequence and speed sequence ,in, represents the i-th test driving torque, represents the i-th uniform vehicle speed, then the vehicle speed sequence and test drive torque sequence T vf (v) The horizontal and vertical axes.

[0031] Under the preset environmental conditions (unloaded vehicle, 0.8 adhesion coefficient, dry asphalt straight road, no wind), the rated drive torque range [T min , T max ] is divided into n+1 equal parts, and the driving torque step length ΔT is obtained. According to the test driving torque from T min At the beginning, the driving torque is increased by a step size ΔT each time until the test driving torque increases to T maxUntil then, record the time when the vehicle reaches a steady-state uniform speed after each test drive torque is issued. At this time, the test drive torque sequence is sequentially Extract a driving torque increment and add it to the test driving torque, record the average acceleration at this time, and thus obtain the average acceleration sequence and requested drive torque sequence , , represents the jth driving torque increment, Indicates based on and The requested driving torque is obtained, then the vehicle speed sequence , average acceleration series , request drive torque sequence T af The horizontal, vertical and vertical coordinates of (v, a);

[0032] Divide the accelerator pedal opening range [0%, 100%] into n+1 equal parts to obtain the accelerator pedal opening step Δa, Δa=100 / n. The accelerator pedal opening increases by one Δa from 0% to form the corresponding accelerator pedal opening. Solve the corresponding test accelerator pedal driving torque according to the hyperbolic tangent function to obtain the test accelerator pedal driving torque sequence and accelerator pedal opening sequence , represents the ith accelerator pedal opening, represents the accelerator pedal driving torque of the i-th test, , tanh represents the hyperbolic tangent function, then, the accelerator pedal opening sequence and test the accelerator pedal actuation torque sequence T pf (a) The horizontal and vertical axes.

[0033] In step S2, two offline brake pressure databases are established, namely, a vehicle speed-deceleration-brake pressure database and a brake pedal opening-brake pressure database;

[0034] The vehicle speed-deceleration-brake pressure database is a two-dimensional discrete data table F bf (v, b), the interpolation method uses a cubic spline curve, F bf The input of (v, b) is vehicle speed and deceleration, and the output is vehicle speed minus deceleration brake pressure obtained by looking up the table;

[0035] The brake pedal opening-brake pressure database is a one-dimensional discrete data table F pf (b), the interpolation method uses the hyperbolic tangent function, F pfThe input of (b) is the brake pedal opening, and the output is the opening brake pressure obtained by looking up the table.

[0036] Under the preset environmental conditions (unloaded vehicle, 0.8 adhesion coefficient, dry asphalt straight road, no wind), the rated brake pressure range [F min , F max ] is divided into n+1 equal parts, F min is the minimum rated brake pressure, F max is the maximum rated braking pressure, and the braking pressure step ΔF is obtained, ΔF=(F max -F min ) / n, according to the test brake pressure F min At the beginning, the brake pressure is increased by one step ΔF each time until the test brake pressure increases to F max Until then, record the time when the vehicle reaches a constant speed in a steady state after each test brake pressure is issued, so as to obtain the initial brake pressure sequence , represents the i-th initial test brake pressure. In addition, the initial brake pressure sequence is Extract a brake pressure increment and add it to the initial brake pressure, record the average deceleration at this time, and thus obtain the average deceleration sequence and test brake pressure sequence , , represents the jth brake pressure increment, Indicates based on and The test brake pressure obtained is represents the jth average deceleration, then the vehicle speed sequence , average deceleration sequence , Test brake pressure sequence They constitute F bf The horizontal, vertical and vertical coordinates of (v, b);

[0037] Divide the brake pedal opening range [0%, 100%] into n+1 equal parts to obtain the brake pedal opening step Δp, Δp=100 / n. The brake pedal opening increases by one Δp from 0% to form the corresponding brake pedal opening. Solve the corresponding test brake pedal pressure according to the hyperbolic tangent function to obtain the test brake pedal pressure sequence and brake pedal opening sequence , represents the ith accelerator pedal opening, represents the brake pedal pressure of the ith test, , then, the brake pedal opening sequence and test brake pedal pressure sequence They constitute F pf (b) The horizontal and vertical axes.

[0038] Step S3, based on the longitudinal control signal source, combined with the offline driving torque database and the offline braking pressure database, a model-free vehicle speed control strategy is designed to calculate the vehicle speed expected driving torque and the vehicle speed expected braking pressure.

[0039] In step S3, the model-free vehicle speed control strategy includes drive control and brake control, and both drive control and brake control adopt a fusion method of feedforward control and conditional PID feedback control;

[0040] In the drive control, the feedforward control part is to convert the desired vehicle speed v r Substitute T vf (v) The vehicle speed feedforward driving torque T is obtained 1 ; The conditional PID feedback control part is based on the expected vehicle speed v r Subtract the actual vehicle speed v e The obtained speed error e v , and then calculate the vehicle speed proportional torque T 2 =P v ×e v , vehicle speed integral torque T 3 =I v ×e vm , vehicle speed differential torque T 4 =D v ×e vd , vehicle speed condition PID feedback control torque T 5 = T 2 +T 3 + T 4 , P v ,I v , D v They are the vehicle speed proportional coefficient, vehicle speed integral coefficient, and vehicle speed differential coefficient, respectively. vm is the accumulated error of vehicle speed, e vm is the total speed error e from the time the system is started to the current time v The sum of vd is the speed error increment, e vd is the difference between the current speed error and the previous speed error. The output speed torque T of the model-free speed control 6 = T 1 + T 5 ; In order to suppress the vehicle shaking caused by the change of driving torque, a torque limiting and anti-shaking strategy is designed to calculate the driving speed torque T 7 = T e +sign(T 6 -T e)×min(|T 6 -T e |, ΔT max ), where sign represents the sign function, ΔT max is the maximum increment of the set driving torque, and min represents the minimum value; when T 7 ≥0, the vehicle speed expected driving torque T v =T 7 When T 7 <0, the vehicle speed expected driving torque T v = 0, and at the same time, the cumulative error clearing module is triggered, that is, at the current moment, the vehicle speed cumulative error e vm It is set to 0 to avoid the phenomenon of integral saturation caused by the continuous accumulation of negative driving speed torque.

[0041] In braking control, the vehicle speed error e v Differentiate to get the deceleration b v The feedforward control part is to convert the desired vehicle speed v r , Deceleration b v Substitute F bf (v, b) to obtain vehicle speed-deceleration feedforward brake pressure F 1 ; The conditional PID feedback control part is based on the deceleration error e vb , and then calculate the vehicle speed-deceleration ratio brake pressure F 2 =P vb ×e vb , vehicle speed-deceleration integral brake pressure F 3 =I vb *e vbm , Vehicle speed-deceleration differential brake pressure T 4 =D vb *e vbd , Vehicle speed-deceleration condition PID feedback control pressure F 5 = F 2 +F 3 +F 4 , P vb ,I vb , D vb They are the vehicle speed-deceleration proportional coefficient, vehicle speed-deceleration integral coefficient, and vehicle speed-deceleration differential coefficient, respectively. vb =b v - b e , e vbm is the vehicle speed-deceleration cumulative error, e vbm is the total deceleration error e from the time the system starts to the current time vb The sum of vbd is the vehicle speed-deceleration error increment, e vbdis the difference between the deceleration error at the current moment and the deceleration error at the previous moment; the output speed of the model-free speed control - deceleration brake pressure F 6 = F 1 + F 5 ; In order to suppress the vehicle shaking caused by the change of brake pressure, a pressure limiting and de-shaking strategy is designed to calculate the vehicle speed-deceleration brake pressure F 7 = F e +sign(F 6 -F e )×min(|F 6 -F e |, ΔF max ), where ΔF max is the maximum increment of the set brake pressure; when F 7 ≥0, the vehicle speed expected brake pressure F v = F 7 When F 7 <0, the expected braking pressure F v = 0, at the same time, the cumulative error clearing module is triggered, that is, at the current moment, the vehicle speed-deceleration cumulative error e vbm It is set to 0 to avoid the phenomenon of integral saturation caused by the continuous accumulation of negative vehicle speed-deceleration brake pressure.

[0042] Step S4, based on the longitudinal control signal source and in combination with the offline driving torque database, a model-free acceleration control strategy is designed to calculate the acceleration expected driving torque.

[0043] In step S4, the model-free acceleration control strategy adopts a fusion method of feedforward control and conditional PID feedback control, and the feedforward control consists of two parts, wherein the first part of the feedforward control is the actual vehicle speed v e Substitute T vf (v) Get the acceleration driving torque T 8 The second part of the feedforward control is to convert the actual vehicle speed v e , expected acceleration a r Substitute T af (v, a) to get the speed-acceleration drive torque T 9 ; The conditional PID feedback control part is based on the expected acceleration a r and the actual acceleration a e The acceleration error e a , and obtain the acceleration proportional torque T 10 =P a ×e a , acceleration integral torque T 11 =I a ×e am , Acceleration differential torque T 12 =D a ×ead , acceleration condition PID feedback control torque T 13 = T 11 + T 12 + T 10 , P a ,I a , D a They are respectively the acceleration proportional coefficient, the acceleration integral coefficient, and the acceleration differential coefficient, e am is the acceleration accumulation error, e am is the total acceleration error e from the time the system starts to the current time a The sum of ad is the acceleration error increment, e ad is the difference between the acceleration error at the current moment and the acceleration error at the previous moment; the output acceleration torque T of the model-free acceleration control 14 = T 13 + T 8 + T 9 ; In order to suppress the vehicle shaking caused by the change of driving torque, a torque limiting and shaking prevention strategy is designed to calculate the driving acceleration torque T 15 = T e +sign(T 14 -T e )×min(|T 14 -T e |, ΔT max ), when T 15 ≥0, the acceleration expected driving torque T a = T 15 When T 15 <0, the acceleration expected driving torque T a = 0, at the same time, the cumulative error clearing module is triggered, that is, at the current moment, the cumulative error e is accelerated am It is set to 0 to avoid the phenomenon of integral saturation caused by the continuous accumulation of negative driving acceleration torque.

[0044] Step S5, based on the longitudinal control signal source and in combination with the offline brake pressure database, a model-free deceleration control strategy is designed to calculate the deceleration expected brake pressure.

[0045] In step S5, the model-free deceleration control strategy adopts a fusion method of feedforward control and conditional PID feedback control. The feedforward control part is to calculate the actual vehicle speed v e , expected deceleration b r Substitute F bf (v, b) to get the deceleration feedforward brake pressure F 8 ; The conditional PID feedback control part is based on the deceleration error e b , and then calculate the deceleration ratio brake pressure F 9 =Pb ×e b , deceleration integral brake pressure F 10 =I b ×e bm , deceleration differential brake pressure F 11 =D b ×e bd , deceleration condition PID feedback control pressure F 12 =F 10 +F 11 + F 9 , P b ,I b , D b are respectively the deceleration proportional coefficient, deceleration integral coefficient, and deceleration differential coefficient, e b =b r - b e , e bm is the deceleration cumulative error, e bm is the total deceleration error e from the time the system starts to the current time b The sum of bd is the error increment, e bd is the difference between the deceleration error at the current moment and the deceleration error at the previous moment; in order to suppress the vehicle jitter caused by the change in brake pressure, a pressure limiting and de-jittering strategy is designed to calculate the deceleration brake pressure F 13 = F e +sign(F 12 -F e )×min(|F 12 -F e |, ΔF max ); when F 13 ≥0, the expected braking pressure F for deceleration a = F 13 When F 13 <0, deceleration expected brake pressure F a = 0, and at the same time, the cumulative error clearing module is triggered, that is, at the current moment, the cumulative error e bm It is set to 0 to avoid the phenomenon of integral saturation caused by the continuous accumulation of negative deceleration brake pressure.

[0046] Step S6, designing a model-free longitudinal control mode arbitration strategy based on the expected driving torque at vehicle speed, the expected braking pressure at vehicle speed, the expected driving torque at acceleration, and the expected braking pressure at deceleration, outputting the requested driving torque and the requested braking pressure, and realizing the longitudinal operation of the autonomous driving vehicle.

[0047] In step S6, the model-free longitudinal control mode arbitration strategy is executed in two levels, wherein the first level is torque arbitration and pressure arbitration, and the second level is over-limit arbitration;

[0048] In torque arbitration, according to the desired accelerator pedal opening a p Check T pf (a) Get the pedal driving torque T 16 The logic of torque arbitration is: Under normal circumstances, no matter which driving torque calculation mode is used, it should be able to truly reflect the longitudinal driving needs of the autonomous driving vehicle, that is, the pedal driving torque T 16 , vehicle speed expected driving torque T v and acceleration desired driving torque T a Theoretically, the three torques should be equal. Considering the calculation accuracy error and calculation mode difference, a logical threshold ΔT of the driving torque is set. mmax , when |T v -T a |<ΔT mmax ,|T v -T 16 |<ΔT mmax ,|T a -T 16 |<ΔT mmax When both conditions are met, the output arbitration torque T vaf =(T v +T a +T 16 ) / 3, and upload the driving torque normal signal to the fault handling module of the automatic driving; when |T v -T a |<ΔT mmax ,|T v -T 16 |<ΔT mmax ,|T a -T 16 |<ΔT mmax When the conditions are not met at the same time, the arbitration torque T vaf =0, and upload the abnormal driving torque signal to the fault handling module of the autonomous driving.

[0049] In pressure arbitration, according to the expected brake pedal opening b p Check F pf (b) Get the pedal brake pressure F 14 The logic of pressure arbitration is: Under normal circumstances, no matter which brake pressure calculation mode is used, it should be able to truly reflect the longitudinal braking needs of the autonomous driving vehicle, that is, the pedal brake pressure F 14 、Vehicle speed expected brake pressure F v and the desired braking pressure F a The three pressures should be equal in theory. Considering the calculation accuracy error and calculation mode difference, a logic threshold ΔF of the brake pressure is set. mmax , when |Fv -F a |<ΔF mmax ,|F v -F 14 |<ΔF mmax ,|F a -F 14 |<ΔF mmax When all the conditions are met, the output arbitration pressure F vaf =(F v +F a +F 14 ) / 3, and upload the brake pressure no abnormal signal to the fault handling module of the automatic driving; when |F v -F a |<ΔF mmax ,|F v -F 14 |<ΔF mmax ,|F a -F 14 |<ΔF mmax If the conditions are not met at the same time, the arbitration pressure F vaf =0, and upload the abnormal brake pressure signal to the fault handling module of the autonomous driving;

[0050] The logic of over-limit arbitration is: Under normal circumstances, it is impossible to request driving and braking at the same time, that is, only when the arbitration pressure F vaf =0, arbitration torque T vaf It can be non-zero and only when the arbitration torque T vaf =0, arbitration pressure F vaf It can be non-zero. If the arbitration pressure F vaf and arbitration torque T vaf At the same time, it is not 0, indicating that an abnormal situation has occurred and a fault must be reported. Considering the accuracy error of the calculation, a logical threshold ΔT of the driving torque is set for this purpose. mmin and the logic threshold ΔF of the brake pressure mmin , when T vaf <ΔT mmin When the required driving torque T=T vaf , requested brake pressure F=F vaf , and the fault handling module of the autonomous driving uploads no abnormal signal of longitudinal control; when T vaf >ΔT mmin , and F vaf <ΔF mmin When the required driving torque T=T vaf , requested brake pressure F=F vaf , and the fault handling module of the autonomous driving uploads no abnormal signal of longitudinal control; when T vaf >ΔT mmin , and Fvaf >ΔF mmin When the requested driving torque T=0, the requested braking pressure F=F vaf , and the fault handling module of the autonomous driving uploads the longitudinal control abnormality signal.

[0051] Verification case: Figure 2 As shown, it describes an exemplary scenario of a step speed actual vehicle test, in which the expected vehicle speed starts from 0, jumps to 0.8m / s at 0.8 seconds, and then remains at 0.8m / s. The traditional PID control method will have a relatively violent jitter from 0.8 seconds to 1 second, and then there will be a relatively large overshoot, and then approach the expected vehicle speed; after the implementation of the method proposed in the present invention, there is obviously no violent jitter and large overshoot.

[0052] like Figure 2 As shown, another exemplary scenario of a step speed actual vehicle test is described, wherein the expected vehicle speed starts from 0, jumps to 1.5 m / s at 1.4 seconds, and then remains at 1.5 m / s. The traditional PID control method will have a relatively violent jitter from 1.4 seconds to 1.6 seconds, and then there will be a relatively large overshoot, and then approach the expected vehicle speed; however, after the implementation of the method proposed in the present invention, there is obviously no violent jitter and large overshoot.

[0053] The above results show that the rail longitudinal control method of the present invention can reduce the tracking error and stabilize the longitudinal control.

[0054] In summary, the present invention has the following beneficial effects:

[0055] (1) The control method of the present invention does not rely on the precise physical model of the autonomous driving vehicle system. In the case of a real vehicle, the longitudinal stability control of the autonomous driving vehicle can be achieved by importing limited real vehicle experimental data. This can effectively solve the problem of longitudinal control stability when the autonomous driving vehicle model parameters are missing or the model parameters are uncertain.

[0056] (2) The feedforward control and conditional PID feedback control designed in the present invention are integrated to perform driving control and braking control, which can effectively suppress vehicle vibration caused by changes in driving torque or braking pressure, avoid the integral saturation phenomenon caused by continuous accumulation of negative driving torque or negative braking pressure, and contribute to the longitudinal stability control of the autonomous driving vehicle.

[0057] (3) The model-free longitudinal control mode arbitration designed by the present invention not only takes into account the unexpected driving or braking caused by interference from factors such as program anomalies and communication congestion, but also takes into account the accuracy error of the actual calculation and the difference in the calculation mode. Therefore, it can solve the problems of safety and effectiveness and has strong applicability. It can be applied to the autonomous valet parking induction control of a single intelligent vehicle or multiple intelligent vehicles at the same time, which is more efficient.

[0058] (4) Since the present invention is not affected by the accuracy of the vehicle physical model, it has strong practicality and flexibility and can be efficiently applied to various types of autonomous driving vehicles, including autonomous driving cars, autonomous driving buses, autonomous driving special vehicles, etc., which can speed up the development progress and reduce R&D costs.

[0059] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation 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 one or more embodiments or examples in a suitable manner.

[0060] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

Claims

1. A multi-mode redundant model-free longitudinal control method for an autonomous driving vehicle, characterized in that: include: Step S1, collecting longitudinal control signal sources of different types and sources; Step S2, establishing an offline driving torque database and an offline braking pressure database based on the longitudinal control signal source; Step S3, based on the longitudinal control signal source, combined with the offline driving torque database and the offline braking pressure database, design a model-free vehicle speed control strategy, calculate the expected vehicle speed driving torque and the expected vehicle speed braking pressure, wherein the model-free vehicle speed control strategy includes driving control and braking control, and both driving control and braking control adopt a fusion method of feedforward control and conditional PID feedback control; Step S4, based on the longitudinal control signal source and in combination with the offline driving torque database, a model-free acceleration control strategy is designed to calculate the expected driving torque for acceleration, wherein the model-free acceleration control strategy adopts a fusion method of feedforward control and conditional PID feedback control; Step S5, based on the longitudinal control signal source and in combination with the offline brake pressure database, a model-free deceleration control strategy is designed to calculate the expected deceleration brake pressure, wherein the model-free deceleration control strategy adopts a fusion method of feedforward control and conditional PID feedback control; Step S6, designing a model-free longitudinal control mode arbitration strategy according to the vehicle speed expected driving torque, the vehicle speed expected braking pressure, the acceleration expected driving torque, and the deceleration expected braking pressure, outputting the requested driving torque and the requested braking pressure, and realizing the longitudinal operation of the automatic driving vehicle; In step S6, the model-free longitudinal control mode arbitration strategy is executed in two levels, wherein the first level is torque arbitration and pressure arbitration, and the second level is over-limit arbitration; In torque arbitration, according to the desired accelerator pedal opening a p Check the accelerator pedal opening-driving torque database to get the pedal driving torque T 16 The logic of torque arbitration is: set a logical threshold ΔT of the driving torque mmax , when |T v -T a |<ΔT mmax ,|T v -T 16 |<ΔT mmax ,|T a -T 16 |<ΔT mmax When both conditions are met, the output arbitration torque T vaf =(T v +T a +T 16 ) / 3, and upload the driving torque normal signal to the fault handling module of the automatic driving; when |T v -T a |<ΔT mmax ,|T v -T 16 |<ΔT mmax ,|T a -T 16 |<ΔT mmax When the conditions are not met at the same time, the arbitration torque T vaf =0, and upload the abnormal driving torque signal to the fault handling module of the autonomous driving, where T v represents the expected driving torque at vehicle speed, T a It represents the expected driving torque for acceleration; In pressure arbitration, according to the expected brake pedal opening b p Check the brake pedal opening-brake pressure database to get the pedal brake pressure F 14 The logic of pressure arbitration is: set a logic threshold value ΔF of brake pressure mmax , when |F v -F a |<ΔF mmax ,|F v -F 14 |<ΔF mmax ,|F a -F 14 |<ΔF mmax When all the conditions are met, the output arbitration pressure F vaf =(F v +F a +F 14 ) / 3, and upload the brake pressure no abnormal signal to the fault handling module of the automatic driving; when |F v -F a |<ΔF mmax ,|F v -F 14 |<ΔF mmax ,|F a -F 14 |<ΔF mmax If the conditions are not met at the same time, the arbitration pressure F vaf =0, and upload the abnormal brake pressure signal to the fault handling module of the autonomous driving, where F v Indicates the expected braking pressure at vehicle speed, F a Indicates the brake pressure expected for deceleration; The logic of over-limit arbitration is: set a logical threshold ΔT of the driving torque mmin and the logic threshold ΔF of the brake pressure mmin , when T vaf <ΔT mmin When the required driving torque T=T vaf , requested brake pressure F=F vaf , and the fault handling module of the autonomous driving uploads no abnormal signal of longitudinal control; when T vaf >ΔT mmin , and F vaf <ΔF mmin When the required driving torque T=T vaf , requested brake pressure F=F vaf , and the fault handling module of the autonomous driving uploads no abnormal signal of longitudinal control; when T vaf >ΔT mmin , and F vaf >ΔF mmin When the requested driving torque T=0, the requested braking pressure F=F vaf , and the fault handling module of the autonomous driving uploads the longitudinal control abnormality signal.

2. The multi-mode redundant model-free longitudinal control method for an autonomous driving vehicle according to claim 1, characterized in that: The longitudinal control signal source includes the actual acceleration a e , expected acceleration a r 、Expected accelerator pedal opening a p , actual vehicle speed v e 、Expected vehicle speed v r , actual driving torque T e , actual brake pressure F e , Expected brake pedal opening b p , expected deceleration b r , actual deceleration b e .

3. The multi-mode redundant model-free longitudinal control method for an autonomous driving vehicle according to claim 2, characterized in that: In step S2, three offline driving torque databases are established, namely, a vehicle speed-driving torque database, a vehicle speed-acceleration-driving torque database, and an accelerator pedal opening-driving torque database; The vehicle speed-driving torque database is a one-dimensional discrete data table T vf (v), the interpolation method uses cubic spline curve to achieve third-order continuity, T vf The input of (v) is the vehicle speed, and the output is the vehicle speed torque obtained by looking up the table; The vehicle speed-acceleration-drive torque database is a two-dimensional discrete data table T af (v, a), the interpolation method uses a cubic spline curve to achieve third-order continuity, T af The input of (v, a) is vehicle speed and acceleration, and the output is the acceleration torque obtained by looking up the table; The accelerator pedal opening-driving torque database is a one-dimensional discrete data table T pf (a), interpolation method uses hyperbolic tangent function, T pf The input of (a) is the accelerator pedal opening, and the output is the pedal torque obtained by looking up the table.

4. The multi-mode redundant model-free longitudinal control method for an autonomous driving vehicle according to claim 3, characterized in that: Under the preset environmental conditions, the rated drive torque range [T min , T max ] is divided into n+1 equal parts, T min is the minimum rated driving torque, T max is the maximum rated driving torque, and the driving torque step length ΔT is obtained, ΔT=(T max -T min ) / n, according to the test drive torque from T min At the beginning, the driving torque is increased by a step size ΔT each time until the test driving torque increases to T max Until then, record the time when the vehicle reaches a constant speed in a steady state after each test drive torque is issued, so as to obtain the test drive torque sequence and speed sequence ,in, represents the i-th test driving torque, represents the i-th uniform vehicle speed, then the vehicle speed sequence and test drive torque sequence T vf (v) the abscissa and ordinate; Under the preset environmental conditions, the rated drive torque range [T min , T max ] is divided into n+1 equal parts, and the driving torque step length ΔT is obtained. According to the test driving torque from T min At the beginning, the driving torque is increased by a step size ΔT each time until the test driving torque increases to T max Until then, record the time when the vehicle reaches a steady-state uniform speed after each test drive torque is issued. At this time, the test drive torque sequence is sequentially Extract a driving torque increment and add it to the test driving torque, record the average acceleration at this time, and thus obtain the average acceleration sequence and requested drive torque sequence , , represents the jth driving torque increment, Indicates based on and The requested driving torque is obtained, then the vehicle speed sequence , average acceleration series , request drive torque sequence T af The horizontal, vertical and vertical coordinates of (v, a); Divide the accelerator pedal opening range [0%, 100%] into n+1 equal parts to obtain the accelerator pedal opening step Δa, Δa=100 / n. The accelerator pedal opening increases by one Δa from 0% to form the corresponding accelerator pedal opening. Solve the corresponding test accelerator pedal driving torque according to the hyperbolic tangent function to obtain the test accelerator pedal driving torque sequence and accelerator pedal opening sequence , represents the ith accelerator pedal opening, represents the accelerator pedal driving torque of the i-th test, , tanh represents the hyperbolic tangent function, then, the accelerator pedal opening sequence and test the accelerator pedal actuation torque sequence T pf (a) The horizontal and vertical axes.

5. The multi-mode redundant model-free longitudinal control method for an autonomous driving vehicle according to claim 4, characterized in that: In step S2, two offline brake pressure databases are established, namely, a vehicle speed-deceleration-brake pressure database and a brake pedal opening-brake pressure database; The vehicle speed-deceleration-brake pressure database is a two-dimensional discrete data table F bf (v, b), the interpolation method uses a cubic spline curve, F bf The input of (v, b) is the vehicle speed and deceleration, and the output is the vehicle speed minus the deceleration brake pressure obtained by looking up the table; The brake pedal opening-brake pressure database is a one-dimensional discrete data table F pf (b), the interpolation method uses the hyperbolic tangent function, F pf The input of (b) is the brake pedal opening, and the output is the opening brake pressure obtained by looking up the table.

6. The multi-mode redundant model-free longitudinal control method for an autonomous driving vehicle according to claim 5, characterized in that: Under the preset environmental conditions, the rated brake pressure range [F min , F max ] is divided into n+1 equal parts, F min is the minimum rated brake pressure, F max is the maximum rated braking pressure, and the braking pressure step ΔF is obtained, ΔF=(F max -F min ) / n, according to the test brake pressure F min At the beginning, the brake pressure is increased by one step ΔF each time until the test brake pressure increases to F max Until then, record the time when the vehicle reaches a constant speed in a steady state after each test brake pressure is issued, so as to obtain the initial brake pressure sequence , represents the i-th initial test brake pressure. In addition, the initial brake pressure sequence is Extract a brake pressure increment and add it to the initial brake pressure, record the average deceleration at this time, and thus obtain the average deceleration sequence and test brake pressure sequence , , represents the jth brake pressure increment, Indicates based on and The test brake pressure obtained is represents the jth average deceleration, then the vehicle speed sequence , average deceleration sequence , Test brake pressure sequence They constitute F bf The horizontal, vertical and vertical coordinates of (v, b); Divide the brake pedal opening range [0%, 100%] into n+1 equal parts to obtain the brake pedal opening step Δp, Δp=100 / n. The brake pedal opening increases by one Δp from 0% to form the corresponding brake pedal opening. Solve the corresponding test brake pedal pressure according to the hyperbolic tangent function to obtain the test brake pedal pressure sequence and brake pedal opening sequence , represents the ith accelerator pedal opening, represents the brake pedal pressure of the ith test, , then, the brake pedal opening sequence and test brake pedal pressure sequence They constitute F pf (b) The horizontal and vertical axes.

7. The multi-mode redundant model-free longitudinal control method for an autonomous driving vehicle according to claim 6, characterized in that: In step S3, in the driving control of the model-free vehicle speed control strategy, the feedforward control part is to calculate the desired vehicle speed v r Substitute T vf (v) to obtain the vehicle speed feedforward drive torque T1; the conditional PID feedback control part is based on the desired vehicle speed v r Subtract the actual vehicle speed v e The obtained speed error e v , and then calculate the vehicle speed proportional torque T2=P v ×e v , vehicle speed integral torque T3 = I v ×e vm , vehicle speed differential torque T4 = D v ×e vd , vehicle speed condition PID feedback control torque T5 = T2 + T3 + T4, P v ,I v , D v They are the vehicle speed proportional coefficient, vehicle speed integral coefficient, and vehicle speed differential coefficient, respectively. vm is the accumulated error of vehicle speed, e vd is the speed error increment, the output speed torque T6 of the model-free speed control = T1 + T5; the driving speed torque T7 = T e +sign(T6- T e )×min(|T6- T e |, ΔT max ), where sign represents the sign function, ΔT max is the maximum increment of the set driving torque, and min represents the minimum value; when T7 ≥ 0, the vehicle speed expected driving torque T v = T7; when T7<0, the vehicle speed expected driving torque T v = 0; In braking control, the speed error e v Differentiate to get the deceleration b v The feedforward control part is to convert the desired vehicle speed v r , Deceleration b v Substitute F bf (v, b) to obtain the vehicle speed-deceleration feedforward brake pressure F1; the conditional PID feedback control part is based on the deceleration error e vb , and then obtain the vehicle speed-deceleration ratio brake pressure F2=P vb ×e vb , vehicle speed - deceleration integral brake pressure F3 = I vb *e vbm , vehicle speed - deceleration differential brake pressure T4 = D vb *e vbd , vehicle speed-deceleration condition PID feedback control pressure F5 = F2+F3+F4, P vb ,I vb , D vb They are the vehicle speed-deceleration proportional coefficient, vehicle speed-deceleration integral coefficient, and vehicle speed-deceleration differential coefficient, respectively. vb =b v - b e , e vbm is the vehicle speed-deceleration cumulative error, e vbd is the speed-deceleration error increment; the output of the model-free speed control, the speed-deceleration brake pressure F6 = F1 + F5; the speed-deceleration brake pressure F7 = F e +sign(F6- F e )×min(|F6- F e |, ΔF max ), where ΔF max is the maximum increment of the set brake pressure; when F7≥0, the vehicle speed expected brake pressure F v = F7; when F7<0, the vehicle speed expected brake pressure F v = 0.

8. The multi-mode redundant model-free longitudinal control method for an autonomous driving vehicle according to claim 7, characterized in that: In step S4, the feedforward control of the model-free acceleration control strategy consists of two parts, wherein the first part of the feedforward control is the actual vehicle speed v e Substitute T vf (v) to obtain the acceleration driving torque T8, the second part of the feedforward control is to convert the actual vehicle speed v e , expected acceleration a r Substitute T af (v, a) to obtain the speed-acceleration drive torque T9; the conditional PID feedback control part is based on the expected acceleration a r and the actual acceleration a e The acceleration error e a , and obtain the acceleration proportional torque T 10 =P a ×e a , acceleration integral torque T 11 =I a ×e am , Acceleration differential torque T 12 =D a ×e ad , acceleration condition PID feedback control torque T 13 = T 11 + T 12 + T 10 , P a ,I a , D a They are respectively the acceleration proportional coefficient, the acceleration integral coefficient, and the acceleration differential coefficient, e am is the acceleration accumulation error, e ad is the acceleration error increment; the output acceleration torque T of the model-free acceleration control 14 = T 13 + T8+ T9; driving acceleration torque T 15 = T e +sign(T 14 -T e )×min(|T 14 -T e |, ΔT max ), when T 15 ≥0, the acceleration expected driving torque T a = T 15 When T 15 <0, the acceleration expected driving torque T a = 0.

9. The multi-mode redundant model-free longitudinal control method for an autonomous driving vehicle according to claim 8, characterized in that: In step S5, the feedforward control part of the model-free deceleration control strategy is to convert the actual vehicle speed v e , expected deceleration b r Substitute F bf (v, b) to obtain the deceleration feedforward brake pressure F8; the conditional PID feedback control part is based on the deceleration error e b , and then calculate the deceleration ratio brake pressure F9=P b ×e b , deceleration integral brake pressure F 10 =I b ×e bm , deceleration differential brake pressure F 11 =D b ×e bd , deceleration condition PID feedback control pressure F 12 =F 10 +F 11 + F9, P b ,I b , D b are respectively the deceleration proportional coefficient, deceleration integral coefficient, and deceleration differential coefficient, e b =b r - b e , e bm is the deceleration cumulative error, e bd is the error increment; deceleration brake pressure F 13 = F e +sign(F 12 -F e )×min(|F 12 -F e |, ΔF max ); when F 13 ≥0, the expected braking pressure F for deceleration a = F 13 When F 13 <0, deceleration expected brake pressure F a = 0.

Images