CN115492928A - Economic, dynamic and safety comprehensive optimal gear shifting rule optimization method - Google Patents

Abstract

The invention relates to an optimization method of an economic, dynamic and safety comprehensive optimal gear shifting rule, which respectively establishes a comprehensive performance evaluation function to optimize the gear shifting rule according to different performance expectation combinations of a driver in an accelerator pedal operation process and a brake pedal operation process: (1) aiming at the operation process of an accelerator pedal, the opening degree of the accelerator pedal and the vehicle speed are taken as control parameters, and the gear shifting rule is optimized by comprehensively considering the economy and the dynamic property; (2) aiming at the operation process of the brake pedal, the opening degree of the brake pedal and the vehicle speed are used as control parameters, and the gear shifting rule is optimized by comprehensively considering the economy and the safety. The invention considers the running conditions of level roads and ramps, so that the optimized gear shifting rule meets the individual requirements of the driver and better conforms to the actual driving operation condition.

Description

Economic, dynamic and safety comprehensive optimal gear shifting rule optimization method

Technical Field

The invention relates to the technical field of automatic control of automobiles, in particular to an automatic speed change control method for a vehicle, which comprehensively considers the gear change performance expectation of a driver from three dimensions of economy, dynamic performance and safety and optimizes the gear change rule of a stepped automatic speed change vehicle so as to meet the personalized gear change performance requirement of the driver.

Background

In a method for optimizing a gear shifting rule of an automobile based on a dynamic programming algorithm (CN 108333921A), an engine fuel consumption characteristic MAP is established, three driving cycle working conditions and automobile parameter data are given, gears are used as state variables, a gear shifting control command is used as a control variable, the fuel consumption of the whole driving cycle is used as an optimal objective function, a dynamic equation is established, the dynamic programming algorithm is used for respectively solving gear transmission ratio optimal sequences of the cycle working conditions, the optimal sequences are led into an established AMEstim whole automobile simulation model, the speed and the accelerator pedal opening degree data corresponding to each cycle working condition are extracted, the maximum coincidence area of the gear working points under the three cycle working conditions is analyzed and found, and a gear shifting rule curve is made according to the maximum coincidence area. And comparing the accelerator opening data obtained by the gear shifting rule based on dynamic programming and the gear shifting rule based on dynamic property under the same working condition, and adjusting the gear shifting curve based on dynamic programming to enable the accelerator opening to be as close to the accelerator opening based on the dynamic gear shifting rule as possible on the premise of ensuring the fuel economy, so that the driving dynamic property of the automobile is ensured, and finally the gear shifting rule curve meeting both the dynamic property and the fuel economy is obtained.

In a pure electric vehicle gear shifting rule multi-target optimization method of a two-gear automatic gearbox based on a braking working condition (CN 110667395A), the invention obtains motor regenerative braking torque under different gears and braking strength through whole vehicle braking force analysis under the braking working condition, takes the vehicle speed, the motor regenerative braking torque and the braking strength as optimization variables of the gear shifting rule, takes regenerative braking recovered energy and the impact of a whole vehicle during braking as a gear shifting rule optimization target, takes the motor maximum torque value constraint, the regenerative braking low-speed cut-off point constraint and the battery charge state constraint as constraint conditions for whether to brake and shift gears, and adopts a multi-target cuckoo optimization algorithm to optimize the gear shifting rule.

In a driving intention recognition method based on an improved HMM and SVM double-layer algorithm (CN 106971194A), the improved HMM and SVM double-layer algorithm is adopted for off-line training so as to recognize driving intentions of a driver for lane change from the fast left, lane change from the normal left, lane keeping, lane change from the normal right and lane change from the fast right.

In a driver intention identification method (CN 103318181A), a multidimensional discrete hidden Markov model is adopted to establish a double-layer identification structure, and the acceleration/braking behavior and the steering/lane changing behavior of a driver are identified.

In a method for quantifying automobile dynamic property and economic property expectation (CN 111027618A), a fuzzy neural network is adopted to establish a driving intention dynamic property expectation quantification model, a driver operation parameter and a vehicle running state parameter are input into the model to obtain a dynamic property expectation value, the economic property expectation value is further calculated, and the continuous quantification of the driver dynamic property expectation and the economic property expectation is realized.

In summary, from the perspective of improving the comprehensive performance, the existing method for formulating the gear shifting rule of the step-variable automatic transmission vehicle rarely considers the safety of the vehicle in the driving process when optimizing the gear shifting rule, and cannot meet the requirement of a driver on the driving safety of the vehicle. In addition, no matter the accelerator pedal is operated or the brake pedal is operated, the motion state of the automobile is influenced, so that the performance of the whole automobile is changed, in the prior art, the speed and the opening degree of the accelerator pedal are mostly optimized as control parameters of a gear shifting rule, and the opening degree of the brake pedal is rarely optimized as the control parameters of the gear shifting rule.

In the aspect of identification of driving intentions, the prior art mainly performs classification identification on driving behaviors or operation intentions of drivers such as lane changing and braking, and most of the driving behaviors or operation intentions belong to qualitative identification. And a few techniques to quantify driver performance expectations: on one hand, the quantification of the economic expectation and/or the dynamic expectation focuses on the quantification of the economic expectation and/or the dynamic expectation of a driver, so that the continuous quantification of the safety expectation of the driver cannot be realized, the safety requirement of the driver is difficult to reflect in a performance optimization strategy (such as a gear shifting rule and the like), and the quantification index is not comprehensive enough; on the other hand, the existing performance expectation quantization technology generally adopts a fuzzy inference method, and the accuracy of a quantization result needs to be improved because the membership function cannot be optimized. In addition, the prior art does not take the road information (such as the gradient and the like) into consideration in the quantification of the performance expectation of the driver, so that the accuracy of the quantification result is limited.

Disclosure of Invention

The invention aims to provide an economic, dynamic and safety comprehensive optimal gear shifting rule optimization method, and aims to establish a driver performance expectation quantification model under various working conditions including ramp driving from three dimensions of economic, dynamic and safety, comprehensively consider the driver performance expectation, and optimize the gear shifting rule which meets the vehicle economic, dynamic and safety comprehensive optimal conditions, so that the economic, dynamic and safety comprehensive optimal personalized automatic gear shifting control of the vehicle is realized.

The purpose of the invention is realized by the following steps: an optimization method for an optimal comprehensive gear shifting rule with economy, dynamic performance and safety comprises the following steps:

(1) Determining the form of each partial objective function

Constructing each branch objective function by adopting an integral variable limit function summation mode: respectively taking the lower speed limit and the upper speed limit of a gear shifting speed optimizing interval as an integral lower limit and an integral upper limit, performing definite integration on a low gear performance index, respectively taking the gear shifting speed and the upper speed limit of the gear shifting speed optimizing interval as an integral lower limit and an integral upper limit, performing definite integration on a high gear performance index, and finally summing two integral values, wherein the calculation formula is as follows

Where f (x) is a sub-objective function for a certain performance, x is the shift speed, v _min 、v _max Lower and upper speed limits, y, for the shift speed optimization interval _k 、y _k+1 Performance index values of a gear k and a gear k +1 at the corresponding vehicle speed are obtained;

(2) Establishing an economic, dynamic and safety index subobjective function

(1) Establishing an economic objective function

An engine oil consumption rate neural network model and an engine forced idling oil consumption rate model which are established according to experimental data adopt an integral variable limit function summation mode, and the segmented integral summation of the oil consumption rate of a gear shifting vehicle speed interval to the vehicle speed is used as an economic subobjective function respectively aiming at an accelerator pedal operation process and a brake pedal operation process;

a) Establishing an economic objective function of the accelerator pedal operation process

Wherein x is the shift speed, Q _e1 The fuel consumption rate of the engine is shown, k is a gear, and thr1 is the opening degree of an accelerator pedal; v is the vehicle speed;

b) Establishing an economic objective function for a brake pedal actuation process

In the formula, Q _e2 Forced idling fuel consumption rate for the engine;

(2) establishing dynamic objective function

An engine torque neural network model and an engine anti-drag torque model which are established according to experimental data adopt a mode of integral variable limit function summation, and the sectional integral summation of the acceleration of a gear shifting vehicle speed interval to the vehicle speed is used as a dynamic performance sub-objective function

F _i ＝mg sinθ (6)

F _f ＝mgf cosθ (7)

In the formula, F _t (k) Driving force for k gear, F _i As ramp resistance, F _f To rolling resistance, F _w As air resistance, T _e Is engine output torque, i _g (k) For the transmission in k-speed ratio, i ₀ Is the main reducer transmission ratio eta _T For driveline mechanical efficiency, r is wheel radius, m is vehicle mass, θ is road slope angle (uphill positive, downhill negative), f is rolling resistance coefficient, C _D The calculation formula is as follows, wherein A is windward area, v is vehicle speed, and delta (k) is k-gear rotating mass conversion coefficient

In the formula I _w Is the moment of inertia of the wheel, I _e Is the engine flywheel moment of inertia;

(3) establishing security objective functions

In the braking process, the mode of integral variable limit function summation is adopted, the integral summation of the deceleration of the gear shifting vehicle speed interval to the vehicle speed in sections is used as a safety subobjective function,

in the formula, F _eb (k) For k-gear equivalent braking force, F _b Is the braking force. F _eb (k) Is as follows

In the formula, T _ft The engine is used for resisting the drag torque;

(4) quadratic processing of an objective function

And (3) carrying out secondary processing on each branch objective function: the partial objective functions are converted from a maximum value type to a minimum value type, and the range of each partial objective function value is limited to a [0,1] interval by normalization.

Economic subobjective function after secondary processing

In which for the accelerator pedal operation and brake pedal operation process f _e (x) Are respectively f _e1 (x) And f _e2 (x)；

Dividing the economy under the optimized vehicle speed range into the maximum value and the minimum value of an objective function;

dynamic subobjective function after secondary processing

In the formula (I), the compound is shown in the specification,

dividing the maximum value and the minimum value of the dynamic performance target function in the optimized vehicle speed range;

security sub-objective function after secondary processing

In the formula (I), the compound is shown in the specification,

dividing the maximum value and the minimum value of an objective function for safety in an optimized vehicle speed range;

(3) Constructing a comprehensive evaluation function

A square weighting and ideal point method is adopted to construct a comprehensive evaluation function which gives consideration to economy, dynamic property and safety

F(x)＝ω _e (F _e (x)-F _e ^* ) ² +ω _d (F _d (x)-F _d ^* ) ² +ω _s (F _s (x)-F _s ^* ) ² (15)

In the formula, F _e ^* Sub-objective function values after secondary processing corresponding to economically optimal shift points, F _d ^* Sub-objective function values after secondary processing, F, for the drivability-optimized shift points _s ^* Sub-objective function value after secondary processing, omega, for safety-optimized shift points _e 、ω _d 、ω _s The method comprises the following steps of (1) obtaining an economic weight value, namely an economic expectation, a dynamic weight value, namely a dynamic expectation, and a safety weight value, namely a safety expectation;

(4) Establishing optimal constraint conditions of gear shifting rules

(1) Establishing optimal constraint conditions of gear shifting rules in accelerator pedal operation process

a) The vehicle acceleration after the gear shift cannot be less than 0, i.e.

b) The engine speed should be at the lowest stable speed n corresponding to the opening degree of the accelerator pedal _eαmin And a maximum rotation speed n _eαmax In between, i.e.

c) Aiming at the downhill process, according to the design vehicle speed upper limit corresponding to each gradient in the table 1, determining the additional constraint of the downhill upshift regular optimization, namely: if the current accelerator pedal opening is kept, the upshifting can cause the automobile to accelerate to be higher than the current ramp design vehicle speed upper limit, the upshifting is forbidden,

g ₄ (x)＝v _max (i,k+1)-v _max (i)≤0 (i＜0) (19)

in the formula, v _max (i, k + 1) is the maximum vehicle speed that can be achieved by driving on a slope with a gradient i with a gear of k +1 under the current accelerator pedal opening, v _max (i) Designing an upper limit of the vehicle speed for driving the vehicle on the slope with the gradient i;

TABLE 1 upper limit of design vehicle speed corresponding to different slopes

(2) Establishing optimal constraint conditions of gear shifting rules in brake pedal operation process

a) The vehicle acceleration after the gear shift cannot be greater than 0, i.e.

b) The engine speed should be at the lowestSteady speed n _emin And a forced idling maximum speed n _eimax In between, i.e

(5) Establishing a mathematical model of a gear shifting law optimization problem

(1) Establishing a mathematical model of a gear shifting rule optimization problem in the operation process of an accelerator pedal

(2) Establishing a mathematical model of a gear shifting law optimization problem in a brake pedal operation process

(6) Optimizing shift schedules

Uniformly traversing possible value ranges of the opening degree of an accelerator pedal and the opening degree of a brake pedal by setting a slope (the slope of a flat road is 0) and a gear shifting performance expected combination, solving the established optimization problem mathematical model by utilizing a particle swarm optimization algorithm, solving an upshift rule of the accelerator pedal in the operation process and a downshift rule of the brake pedal in the operation process, and calculating the downshift rule of the accelerator pedal in the operation process (or the upshift rule of the brake pedal in the operation process) according to the optimally solved upshift rule (or downshift rule);

specifically, taking a particle swarm optimization algorithm as an example, the specific steps are as follows:

1) Taking a comprehensive evaluation function as a fitness function, taking the vehicle speed as a particle position, calling a particle swarm optimization algorithm, and solving the up/down gear vehicle speed under the given gradient, the expected combination of gear shifting performance and the opening degree of an accelerator/brake pedal;

2) Traversing each determined accelerator/brake pedal opening, repeating the step 1), and solving the upshifting/downshifting speed u of different accelerator/brake pedal openings under the given gradient and the expected combination of gear shifting performance _aup /u _adown Forming an up/down shift curve under the given gradient and the expected combination of the gear shifting performance;

3) For the operation process of the accelerator pedal, a downshift curve under the given combination of gradient and expected gear shifting performance is calculated according to the following formula

In the formula u _adown (k) The gear speed is reduced for the gear k + 1; u. of _aup (k) The gear-up speed is k gear-up and k +1 gear-up; k is a radical of _max The highest gear is set; an ₁ 、An ₂ An is a downshift coefficient when the opening of An accelerator pedal is less than or equal to 30 percent ₁ ＝0.4、An ₂ =0.6, and An is when the opening degree of An accelerator pedal is more than 30 percent ₁ ＝0.15、An ₂ ＝0.4；

4) For the brake pedal operation process, an upshift curve under a given slope and expected gear shifting performance combination is calculated according to the following formula

In the formula, bn ₁ 、Bn ₂ Taking Bn as the upshift factor ₁ ＝0.1、Bn ₂ ＝0.4；

5) For different expected combinations of gear shifting performance at a given gradient (the operation process of the accelerator pedal is a combination of an economical weight and a dynamic weight; the operation process of the brake pedal is the combination of the economical weight and the safety weight), and the steps 1) to 4) are repeated to obtain gear shifting curves corresponding to different expected combinations of gear shifting performance under a given slope;

6) Setting different gradient values (the gradient of the flat road is 0), and repeating the steps 1) to 5) to obtain gear shifting curves corresponding to different expected combinations of gear shifting performance under different gradients;

through the steps, the gear shifting rules under different gradients, different economic weights and dynamic weights in the accelerator pedal operation process or the gear shifting rules under different gradients, different economic weights and safety weights in the brake pedal operation process can be obtained;

(7) Automatic transmission control for vehicle

During actual running of the stepped automatic speed changing vehicle, firstly, gear shifting rules under different gradients, different economic expectations and dynamic expectations (an accelerator pedal operation process) or different economic expectations and safety expectations (a brake pedal operation process) obtained through offline optimization calculation are stored in a gear shifting rule database in a TCU in the form of a data table, a transmission control unit TCU collects and processes signals of the gradients, the vehicle speed, the accelerator pedal opening and the brake pedal opening in real time, the economic, dynamic and safety expectations of a driver are calculated by utilizing a quantitative model of the economic, dynamic and safety expectations of the driver, for the accelerator pedal operation process, according to the gradients, the economic expectations and the power expectations, gear shifting rules matched with the current gradients, the economic expectations and the power expectations are selected from the gear shifting rule database stored in the TCU, then according to the vehicle speed and the accelerator pedal opening, target gears are determined through the matched gear shifting rules, for the brake pedal operation process, according to the gradient, economic expectations and safety expectations, gear shifting rules matched with the current gradient, economic expectations and safety expectations stored in the TCU are selected, and then the optimal gear shifting rules and the optimal gear shifting rules of the vehicle speed and the gear shifting performance are determined through comprehensive control system.

Compared with the prior art, the invention has the beneficial effects that:

(1) The invention constructs the target functions of economy, dynamic property and safety by adopting the mode of integral variable limit function summation, and comprehensively considers the gradient, the economy, the dynamic property and the safety into the optimization of the gear shifting rule, wherein the speed and the opening degree of an accelerator pedal are used as gear shifting rule control parameters in the operation process of the accelerator pedal, and the speed and the opening degree of a brake pedal are used as the gear shifting rule control parameters in the operation process of the brake pedal.

Due to the fact that the economical efficiency, dynamic performance and safety expectation (weight) are considered in the comprehensive evaluation function, and the economic efficiency, dynamic performance and safety related performance sub-objective functions are considered, the optimized gear shifting rule can reflect individual requirements of a driver on economical efficiency, dynamic performance and safety, and meanwhile the stepped automatic transmission vehicle achieves comprehensive optimization of economical efficiency, dynamic performance and safety.

(2) Because the branch objective function is constructed in the form of integral variable limit function summation, the influence of the shift point on the overall performance can be more accurately reflected, and the process-based comprehensive performance optimization can be realized;

(3) Aiming at an accelerator pedal operation process and a brake pedal operation process respectively, different gear-shifting performance expectation combinations and different gear-shifting control parameters are adopted to optimize a gear-shifting rule, namely, the accelerator pedal operation process considers the economic expectation and the dynamic expectation of a driver, the vehicle speed and the accelerator pedal opening degree are taken as control parameters, the brake pedal operation process considers the economic expectation and the safety expectation of the driver, and the vehicle speed and the brake pedal opening degree are taken as control parameters, so that the optimized gear-shifting rule meets the individual requirements and better accords with the actual driving operation condition.

(4) The technical scheme provided by the invention considers the ramp driving working condition, so that the method is suitable for optimizing the flat road gear shifting rule and the ramp gear shifting rule, and the gear decision of the stepped automatic speed changing vehicle can better adapt to the change of the actual road condition while meeting the individual requirement.

(5) According to the invention, the road gradient and the brake pedal opening degree are introduced into a quantitative scheme of the driver performance expectation, the economic, dynamic and safety expectations of the driver are comprehensively considered, and the defects that the input information is incomplete and the economic, dynamic and safety expectations of the driver cannot be quantized simultaneously in the prior quantitative technology are solved.

(6) According to the technical scheme provided by the invention, through continuous quantification of the economic, dynamic and safety expectations of the driver, a foundation is laid for accurately realizing the automobile personalized control comprehensively considering the economic, dynamic and safety expectations of the driver.

(7) The method adopts the Radial Basis Function (RBF) neural network to establish the quantitative model of the economic, dynamic and safety expectations of the driver, can adjust the internal parameters of the neural network according to the collected physical quantity of the driving control characteristic, the vehicle running state information, the gradient information and the subjective evaluation data of the driver, and can quantify the economic, dynamic and safety expectations of the driver, and the quantification result is more accurate.

The invention provides the method for quantizing the economical, dynamic and safety expectations of the driver based on the radial basis neural network, which aims to solve the defects that the continuous quantization indexes of the performance expectations of the driver are not comprehensive enough and the quantization precision needs to be improved in the prior art, considers the important index of the safety expectations of the driver in the identification of the driving intention, can quantize the expectations of the driver to the economical, dynamic and safety of the automobile in real time in the driving process, and lays a foundation for better realizing the personalized gear shifting of the automatic speed changing vehicle.

Drawings

Fig. 1 is a flowchart of an optimization procedure of an economic, dynamic and safety comprehensive optimal shift schedule according to the present invention.

Fig. 2 is a torque curve of the engine.

FIG. 3 is a graph of engine oil consumption.

Fig. 4 is a graph of equivalent braking force (plotted against engine drag test data).

FIG. 5a (ω) _e ＝0.8，ω _d = 0.2), fig. 5b (ω) _e ＝0.5，ω _d = 0.5), fig. 5c (ω) _e ＝0.2，ω _d = 0.8) are respectively an upshift curve and a downshift curve of the vehicle speed and the accelerator pedal opening under the working condition of a flat road.

FIG. 5d (ω) _e ＝0.8，ω _s = 0.2), fig. 5e (ω) _e ＝0.5，ω _s = 0.5), fig. 5f (ω) _e ＝0.2，ω _s = 0.8) are respectively the upshift curve and the downshift curve of the vehicle speed and the brake pedal opening under the flat road working condition.

FIG. 6a (4% grade uphill, ω) _e ＝0.5，ω _d = 0.5), fig. 6b (7% slope uphill, ω _e ＝0.5，ω _d = 0.5), fig. 6c (10% slope uphill, ω: = 0.5) _e ＝0.5，ω _d = 0.5), respectively an upshift curve and a downshift curve of the vehicle speed and the accelerator pedal opening under different slope conditions.

FIG. 6d (4% slope downhill, ω) _e ＝0.5，ω _s = 0.5), fig. 6e (7% slope downhill, ω _e ＝0.5，ω _s = 0.5), fig. 6f (10% slope downhill, ω _e ＝0.5，ω _s And = 0.5) are respectively an upshift curve and a downshift curve of the vehicle speed and the brake pedal opening under different slope working conditions.

FIG. 7 is a schematic diagram of an implementation method of the present invention.

Fig. 8 is a technical solution for quantifying the economic, dynamic and safety expectations.

FIG. 9 is a schematic diagram of a process for establishing a quantitative model of the economic, dynamic and safety expectations of a driver.

FIG. 10 is a schematic diagram of the RBF neural network structure.

FIG. 11 is a flow of quantitative model training for driver economic, dynamic, and safety expectations.

FIG. 12 is a process for quantifying driver economic, dynamic, and safety expectations.

FIG. 13 is a graph of the error variation of the RBF neural network training process.

Fig. 14 is a statistical chart of the respective expected absolute errors.

FIG. 15 is a flow chart of the driver's economic, dynamic and safety expectations quantification model.

Detailed Description

The terms explain (1) driver economic, power, and safety expectations: the degree of the driver's tendency to the economical, dynamic and safety performance of the automobile is reflected by the driving manipulation characteristic physical quantity (such as the opening degree of an accelerator pedal, the opening degree of a brake pedal, etc.), the vehicle running state information (such as the speed of the automobile, etc.) and the road information (such as the gradient, etc.). In the present invention, the quantized values of the economic expectation, the dynamic expectation, and the safety expectation are referred to as an economic expectation, a dynamic expectation, and a safety expectation, respectively. In the present invention, the economy and dynamics expectations are taken into account for the accelerator pedal operation process, and the economy and safety expectations are taken into account for the brake pedal operation process.

An optimization method for comprehensive optimal shift schedule of economy, power and safety (the flow is shown in figure 1) comprises the following steps:

(1) Determining the form of each partial objective function

Constructing each branch objective function by adopting an integral variable limit function summation mode: and respectively taking the lower speed limit and the upper speed limit of the gear shifting speed optimizing interval as the lower integral limit and the upper integral limit, performing fixed integration on the low-gear performance index, respectively taking the gear shifting speed and the upper speed limit of the gear shifting speed optimizing interval as the lower integral limit and the upper integral limit, performing fixed integration on the high-gear performance index, and finally summing the two integral values. The calculation formula is as follows

Where f (x) is a sub-objective function for a certain performance, x is the shift speed, v _min 、v _max Lower and upper speed limits, y, for the shift speed optimization interval _k 、y _k+1 Performance index values for the k gear and the k +1 gear at the corresponding vehicle speed are obtained.

(2) Establishing an economic, dynamic and safety index subobjective function

(1) Establishing an economic objective function

An engine oil consumption rate neural network model and an engine forced idling oil consumption rate model which are established according to experimental data adopt an integral variable limit function summation mode, and the segmented integral summation of the oil consumption rate of a gear shifting vehicle speed interval to the vehicle speed is used as an economic subobjective function respectively aiming at an accelerator pedal operation process and a brake pedal operation process.

b) Establishing an economic objective function of the accelerator pedal operation process

Wherein x is the shift speed, Q _e1 The fuel consumption rate of the engine is shown, k is a gear, and thr1 is the opening degree of an accelerator pedal; and v is the vehicle speed.

c) Establishing an economic objective function for a brake pedal actuation process

In the formula, Q _e2 And the forced idling fuel consumption rate of the engine is obtained.

(2) Establishing dynamic subobjective function

An engine torque neural network model and an engine anti-drag torque model which are established according to experimental data adopt a mode of integral variable limit function summation, and the sectional integral summation of the acceleration of a gear shifting vehicle speed interval to the vehicle speed is used as a dynamic subobjective function

F _i ＝mg sinθ (6)

F _f ＝mgf cosθ (7)

In the formula, F _t (k) For k-gear driving force, F _i As ramp resistance, F _f To rolling resistance, F _w For air resistance, T _e For engine transmissionOutput torque, i _g (k) For the transmission k-speed gear ratio, i ₀ Is the main reducer transmission ratio eta _T For the mechanical efficiency of the drive train, r is the wheel radius, m is the overall vehicle mass, θ is the road slope angle (uphill positive, downhill negative), f is the rolling resistance coefficient, C _D The calculation formula is as follows, wherein A is windward area, v is vehicle speed, and delta (k) is k-gear rotating mass conversion coefficient

In the formula I _w Is the moment of inertia of the wheel, I _e Is the flywheel moment of inertia of the engine.

(3) Establishing security objective functions

In the braking process, an integral variable limit function summation mode is adopted, and the integral summation of the deceleration of the gear shifting vehicle speed interval to the vehicle speed in sections is used as a safety subobjective function.

In the formula, F _eb (k) For k-gear equivalent braking force, F _b Is the braking force. F _eb (k) Is as follows

In the formula, T _ft The engine anti-drag torque.

(4) Quadratic processing of an objective function

The sub-objective functions are subjected to secondary processing (conversion and normalization processing) to convert the sub-objective functions from a maximum value type to a minimum value type, and the range of each sub-objective function value is limited to the [0,1] interval by normalization.

Economic subobjective function after secondary processing

In which for the accelerator pedal operating process and the brake pedal operating process, f _e (x) Are respectively f _e1 (x) And f _e2 (x)；f _emax 、f _emin And dividing the maximum value and the minimum value of the objective function for optimizing the economy in the vehicle speed range.

Dynamic subobjective function after secondary processing

In the formula (f) _dmax 、f _dmin The maximum value and the minimum value of the dynamic performance sub-objective function in the optimized vehicle speed range are obtained.

Security sub-objective function after secondary processing

In the formula (f) _smax 、f _smin And dividing the maximum value and the minimum value of the objective function for optimizing the safety in the vehicle speed range.

(3) Constructing a comprehensive evaluation function

A square weighting and ideal point method is adopted to construct a comprehensive evaluation function which gives consideration to economy, dynamic property and safety

F(x)＝ω _e (F _e (x)-F _e ^* ) ² +ω _d (F _d (x)-F _d ^* ) ² +ω _s (F _s (x)-F _s ^* ) ² (15)

In the formula, F _e ^* Sub-objective function values after secondary processing for economically optimal shift points, F _d ^* The branch objective function value after the secondary processing corresponding to the dynamic optimal gear shifting point, F _s ^* Sub-objective function value after secondary processing, omega, for safety-optimized shift points _e 、ω _d 、ω _s An economic weight (i.e., economic expectation), a dynamic weight (i.e., dynamic expectation), and a safety weight (i.e., safety expectation).

(4) Establishing optimal constraint conditions of gear shifting rules

(3) Establishing optimal constraint conditions of gear shifting rules in accelerator pedal operation process

d) The vehicle acceleration after the gear shift cannot be less than 0, i.e.

e) The engine speed should be at the lowest stable speed n corresponding to the opening degree of the accelerator pedal _eαmin And the maximum rotational speed n _eαmax In between, i.e

f) Aiming at the downhill process, according to the design vehicle speed upper limit corresponding to each gradient in the table 1, determining the additional constraint of the downhill upshift regular optimization, namely: and if the current accelerator pedal opening is kept, the upshifting can cause the automobile to accelerate to be more than the current ramp design vehicle speed upper limit, and the upshifting is forbidden.

g ₄ (x)＝v _max (i,k+1)-v _max (i)≤0 (i＜0) (19)

In the formula, v _max (i, k + 1) is the maximum vehicle speed that can be achieved by driving on a slope with a gradient i with k +1 gear at the current accelerator pedal opening, v _max (i) And designing an upper limit of the vehicle speed for driving the vehicle on the slope with the gradient i.

TABLE 1 upper limit of design vehicle speed corresponding to different slopes

(4) Establishing optimal constraint conditions of gear shifting rules in brake pedal operation process

c) The vehicle acceleration after the gear shift cannot be greater than 0, i.e.

d) The engine speed should be at the lowest stable speed n _emin And a forced idling maximum speed n _eimax (taking 4500r/min in the embodiment of the application), namely

(5) Establishing a mathematical model of a gear shifting law optimization problem

(1) Establishing a mathematical model of a gear shifting law optimization problem in an accelerator pedal operation process

(2) Establishing a mathematical model of a gear shifting law optimization problem in a brake pedal operation process

(6) Optimizing shift schedules

The method comprises the steps of uniformly traversing possible value ranges of the opening degree of an accelerator pedal and the opening degree of a brake pedal by setting a slope (the slope of a flat road is 0) and a gear shifting performance expectation combination, solving an established optimization problem mathematical model by utilizing an optimization algorithm (such as a particle swarm optimization algorithm, a genetic algorithm, a simulated annealing algorithm and the like), solving an upshift rule of the accelerator pedal in the operation process and a downshift rule of the brake pedal in the operation process, and calculating the downshift rule of the accelerator pedal in the operation process (or the upshift rule of the brake pedal in the operation process) according to the optimally solved upshift rule (or downshift rule).

Specifically, taking a particle swarm optimization algorithm as an example, the method specifically comprises the following steps:

1) Taking a comprehensive evaluation function as a fitness function, taking the vehicle speed as a particle position, calling a particle swarm optimization algorithm, and solving the upshift/downshift vehicle speed under the given gradient, the expected gear shifting performance combination and the accelerator/brake pedal opening degree;

2) Traversing each determined accelerator/brake pedal opening, repeating the step 1), and solving the upshifting/downshifting speed u of different accelerator/brake pedal openings under the given gradient and the expected combination of gear shifting performance _aup /u _adown Forming an up/down gear curve under the given gradient and the expected combination of gear shifting performance;

3) For the operation process of the accelerator pedal, a downshift curve under the given combination of gradient and expected gear shifting performance is calculated according to the following formula

In the formula u _adown (k) The gear speed is reduced for the gear k + 1; u. u _aup (k) The gear-up speed is k gear-up and k +1 gear-up; k is a radical of formula _max The highest gear is set; an ₁ 、An ₂ An is a downshift coefficient when the opening degree of An accelerator pedal is less than or equal to 30% ₁ ＝0.4、An ₂ =0.6, an when the opening degree of An accelerator pedal is more than 30% ₁ ＝0.15、An ₂ ＝0.4。

4) For the brake pedal operation process, an upshift curve under a given slope and expected gear shifting performance combination is calculated according to the following formula

In the formula, bn ₁ 、Bn ₂ For the upshift factor, takeBn ₁ ＝0.1、Bn ₂ ＝0.4。

5) For different expected combinations of gear shifting performance at a given gradient (the operation process of the accelerator pedal is a combination of an economical weight and a dynamic weight; the operation process of the brake pedal is the combination of the economical weight and the safety weight), and the steps 1) to 4) are repeated to obtain the gear shifting curves corresponding to different expected gear shifting performance combinations under the given gradient.

6) And (4) setting different gradient values (the gradient of the flat road is 0), and repeating the steps 1) to 5) to obtain the gear shifting curves corresponding to different gear shifting performance expected combinations under different gradients.

Through the steps, gear shifting rules under different slopes, different economic weights and dynamic weights (accelerator pedal operation process) or different economic weights and safety weights (brake pedal operation process) can be obtained.

It should be noted that, the step 1) calls a particle swarm optimization algorithm to solve the upshift/downshift vehicle speed under the given gradient, the expected shift performance combination and the accelerator/brake pedal opening degree, and the steps are to solve the upshift vehicle speed under the given gradient, the expected shift performance combination and the accelerator pedal opening degree, and to solve the downshift vehicle speed under the given gradient, the expected shift performance combination and the brake pedal opening degree.

Calling a particle swarm optimization algorithm to solve the upshift speed under the given gradient, the expected gear shifting performance combination and the accelerator pedal opening degree, and comprising the following steps of:

a) Initializing the particle population and setting an acceleration constant c ₁ ＝2、c ₂ =2; setting the maximum iteration number M =100; setting a population size N =80; setting the maximum value w of the inertial weight _max =0.9, minimum value w _min ＝0.4。

b) The initial position p of the first particle is given as the initial value of the position and velocity of all particles _l ⁰ At [ adjacent high gear lowest speed u ] _hmin Adjacent low gear maximum speed u _lmax ]Internal random values (l =1, \8230; N); initial velocity v of the first particle _l ⁰ In the [0,0.9]The internal random value (l =1, \8230;, N). According to the comprehensive evaluation function (the vehicle speed in the function is taken as the particle position)Position) calculating the fitness corresponding to the initial position of each particle; respectively taking the current particle position and the corresponding fitness as the local optimal particle position pbest _l And local optima pbestFitness _l (ii) a And respectively taking the minimum value of the fitness of all the particles and the positions of the particles corresponding to the minimum value as a global optimal value gbestFitness and a global optimal particle position gbest.

c) For each particle, iterative optimization starts with n =1 and ends with n = M. And for the nth iteration, the method comprises the following steps:

c1 Computing inertial weights w

c2 For each particle in the population), repeating the steps of:

c21 In [0,1)]Taking a random value r within the range ₁ 、r ₂ 。

c22 Update the velocity and position of the particle

v _l ⁿ ＝wv _l ^n-1 +c ₁ r ₁ (pbest _l -p _l ^n-1 )+c ₂ r ₂ (gbest-p _l ^n-1 ) (28)

Wherein v is _l ^k The speed of the nth iteration particle l; v. of _l ^n-1 The velocity of the (n-1) th iteration particle l; p is a radical of formula _l ^n-1 The position of the particle/is iterated for the (n-1) th time.

p _l ⁿ ＝p _l ^n-1 +v _l ^n-1 (29)

Wherein p is _l ⁿ The position of the particle/for the nth iteration.

c23 Combining with the gear shifting vehicle speed constraint condition to process the position boundary crossing of the particles: if p is _l ⁿ ＜u _hmin Let p stand for _l ⁿ ＝u _hmin (ii) a If p is _l ⁿ ＞u _lmax Let p stand for _l ⁿ ＝u _lmax 。

c24 For calculating the particle lFitness pbestFitness _l ⁿ 。

c25 ) if pbestFitness _l ⁿ ＜pbestFitness _l Then pbestFitness _l ＝pbestFitness _l ⁿ ，pbest _l ＝p _l ⁿ Otherwise pbestFitness _l And pbest _l Remain unchanged.

c26 If pbestFitness _l < gbestFitness, then gbestFitness = pbestFitness _l ，gbest＝pbest _l 。

c3 Determining the upshift speed, and after all the particles are optimized according to the step c 2), taking the global optimal particle position gbest corresponding to the global optimal value gbestFitness as the upshift speed under the given gradient, the expected combination of the gear shifting performance (the economic weight and the dynamic weight) and the opening degree of the accelerator pedal, namely u _aup ＝gbest。

The method comprises the following steps of calling a particle swarm optimization algorithm, solving the substep of solving the downshift vehicle speed under the given slope, the expected gear shifting performance combination and the brake pedal opening degree, and is the same as the substep of calling the particle swarm optimization algorithm to solve the upshift vehicle speed under the given slope, the expected gear shifting performance combination and the accelerator pedal opening degree, except that the expected gear shifting performance combination is expressed by an economic weight and a safety weight, and a dynamic weight in a comprehensive evaluation function (formula (15)) for calculating the fitness is zero.

The technical scheme of the invention is implemented and explained as follows:

the technical scheme of the invention is implemented for an automobile provided with a 5-gear stepped automatic transmission, and the flat road and ramp gear shifting rules comprehensively considering economy, dynamic property and safety are optimized.

The main parameters of the whole vehicle and the power transmission system of the vehicle are shown in the table 2.

TABLE 2 Main parameters of the whole vehicle and the power transmission system

An engine torque and oil consumption rate model based on a BP neural network is established by combining engine experimental data, and corresponding torque curved surfaces and oil consumption rate curved surfaces are respectively shown in fig. 2 and fig. 3.

And (3) drawing an equivalent braking force curve by combining the engine back-dragging experimental data as shown in figure 4.

Optimization example 1: by using the technical scheme of the invention, the shifting rule of the flat road (with the gradient of 0) with the economic weight and the power/safety weight combination of (0.8,0.2), (0.5) and (0.2,0.8) is optimized, and the result is shown in fig. 5.

Optimization example 2: by using the technical scheme of the invention, the slope gear shifting rule under different slopes with the combination of the economical weight and the dynamic/safety weight of (0.5) is optimized, and the result is shown in FIG. 6.

Implementation method

Fig. 7 illustrates how the stepped automatic transmission vehicle performs the function of the present invention during actual running. First, gear shifting rules under different gradients, different economic expectations and dynamic expectations (accelerator pedal operation process) or different economic expectations and safety expectations (brake pedal operation process) obtained through offline optimization calculation are stored in a gear shifting rule database in the TCU in the form of a data table. And a Transmission Control Unit (TCU) acquires and processes signals of gradient, vehicle speed, accelerator pedal opening and brake pedal opening in real time, and calculates the economic, dynamic and safety expectations of the driver by using a quantitative model of the economic, dynamic and safety expectations of the driver. And selecting a gear shifting rule matched with the current gradient, the economic expected value and the dynamic expected value from a gear shifting rule database stored in a TCU according to the gradient, the economic expected value and the dynamic expected value in the accelerator pedal operation process, and determining a target gear according to the matched gear shifting rule according to the vehicle speed and the accelerator pedal opening. According to the gradient, the economic expectation and the safety expectation in the brake pedal operation process, a gear shifting rule matched with the current gradient, the economic expectation and the safety expectation is selected from a gear shifting rule database stored in a TCU, and then the target gear is determined according to the vehicle speed and the brake pedal opening degree through the matched gear shifting rule. And then the prior art is used for enabling the vehicle to realize the optimal gear shifting operation with the comprehensive performance which is consistent with the expected performance of the driver through a gear shifting control system.

The method for establishing the quantitative model of the economic, dynamic and safety expectations of the driver comprises the following specific steps of:

the specific technical scheme of the invention mainly comprises the following steps:

s1 data acquisition and processing

S11, different types of drivers are selected, the automobile is driven to run on roads with different gradients respectively under the working conditions of constant speed, acceleration and deceleration, and driving operation characteristic physical quantities (opening degree of an accelerator pedal and opening degree of a brake pedal), vehicle running state information (vehicle speed), road information (gradient) and subjective evaluations of the drivers on economy, dynamic performance and safety are collected.

Different types of driver selection schemes: the driving styles of the economic type, the dynamic type and the safe type are respectively 25%, the driving style of the compromise type is 25%, and the total number of people is not less than 40.

The slope coverage is-15 ° (downhill) to +15 ° (uphill).

The driver's subjective evaluations of economy, dynamics, and safety are all expressed in numbers of "0-10," 0 "indicating the lowest expected value for the performance, and" 10 "indicating the highest expected value for the performance.

Each test requires 1 driver, 1 driver-on-board recorder and 1 security officer. A driver carries out driving operation under corresponding working conditions and describes subjective evaluation of the driver on economy, dynamic performance and safety in real time; the vehicle-mounted recorder records vehicle state, roads, objective driving operation data (vehicle speed, gradient, accelerator pedal opening and brake pedal opening) and subjective evaluation data of a driver; a safety worker observes the surrounding environment of the experimental vehicle on the vehicle and informs a driver in advance when a dangerous condition is foreseen so as to guarantee the safety of the experimental process.

1) And (3) under a uniform working condition: the driver drives the vehicle at the speed of 20km/h, 40km/h, 60km/h and 80km/h respectively and keeps driving for 3s at a constant speed.

2) And (3) acceleration working condition: the driver drives the vehicle and respectively accelerates slowly, moderately and rapidly at 0-80km/h according to the driving habit of the driver.

3) And (3) deceleration working condition: the driver drives the vehicle to decelerate from 80km/h to 0km/h with slow deceleration, medium deceleration and rapid deceleration, respectively.

S12, screening out dead spots in the data to obtain original test data.

S13 determines training data and test data. The method comprises the steps of dividing the vehicle speed, the gradient, the accelerator pedal opening and the brake pedal opening and corresponding economical, dynamic and safety subjective evaluation into regions according to the vehicle speed, the gradient, the accelerator pedal opening and the brake pedal opening within possible value ranges of 10 and the like, randomly selecting the processed vehicle speed, gradient, accelerator pedal opening and brake pedal opening according to the proportion of 60%, 20% and 20% in each region, forming three types of data A, B and C with the corresponding economical, dynamic and safety subjective evaluation, forming training data by the type A and C data, and forming test data by the type B and C data.

S2, establishing an economical, dynamic and safety expectation quantification model of a driver

It mainly comprises the following steps:

and S21, normalizing the data. And (3) normalizing the training data and the test data, wherein the normalized data range is [ -1,1], and the formula is as follows.

In the formula (I), the compound is shown in the specification,

v, z-values before and after sample data normalization, respectively;

v _min 、v _max -minimum and maximum values of the sample data, respectively.

S22, training of the quantitative model of the economical efficiency, the dynamic performance and the safety expectation of the driver. And training the RBF neural network by using the normalized training data until the training error meets the requirement.

The adopted RBF neural network structure is shown in FIG. 3, and is a three-layer static feedforward neural network, which comprises an input layer, a hidden layer and an output layer.

The mathematical description of the layers is as follows:

inputting a vector:

X＝(x ₁ ,x ₂ ,x ₃ ,K,x _n ) ^T (31)

in the formula, n is the number of input vector nodes.

The excitation function of the j node of the hidden layer adopts a Gaussian function as follows

In the formula, hj and σ _j Respectively representing the center and the width of a basis function at the jth node of the hidden layer, wherein m is the number of the hidden layer nodes.

Output vector

Y＝(y ₁ ,y ₂ ,y ₃ ,K,y _s ) ^T (33)

Wherein s is the number of output layer nodes.

The output layer neuron adopts a linear excitation function, and the k output of the RBF can be obtained by the function as follows:

in the formula, ω _jk (j =1,2, \8230;, m; k =1,2, \8230;, s) represents the weight between the jth node of the hidden layer and the kth node of the output layer.

The parameter needing to be learned in RBF neural network training is the center h of the basis function _j Variance σ _j And a connection weight w _jk The RBF neural network is adopted to carry out quantitative model training of the economical efficiency, dynamic performance and safety expectation of the driver, and a training process flow chart is as follows. The method comprises the following specific steps:

(1) Loading a normalized training sample set: assuming that the input set in the normalized training sample set is X ₁ ,X ₂ ,...,X _p P is the number of training samples, and the corresponding target output set is Y ₁ ,Y ₂ ,...,Y _p And respectively loading the normalized input set and the normalized output set into the RBF neural network model.

(2) Setting parameter values of the RBF neural network: setting an upper error limit epsilon, an initial value of the number m of neurons in the hidden layer and the maximum number m of neurons _max 。

(3) Initializing a cluster center, namely randomly extracting m different samples from a training sample set to be used as an initial cluster center h _j (b) (j =1, 2.... M), b is the number of iterations, let b =1.

(4) Calculating the center of the basis function:

1) Calculating the distance from all samples to the initial clustering center, i.e., | | X _i -h _j (b)||,i＝1,2,...,p。

2) For sample input X _i Classifying according to a minimum distance principle: namely when

When, X _i I.e. to be classified as jth.

3) Center of gravity is adjusted by the following formula

h _j (b+1)＝h _j (b)+η[X _i -h _j (b)] (35)

Where η is the learning rate and η > 0.

4) If h is _j (b+1)≠h _j (b) B = b +1 and go to step 1), otherwise, clustering is finished to obtain the final basis function center h _j (h _j ＝h _j (b))。

(5) Calculating the variance of the basis function according to the following formula

In the formula (d) _max Is the maximum distance between the selected centers.

(6) Calculating system output as the center h of the basis function _j And its variance σ _j After determination, ω is adjusted using a least squares method _jk Minimizing training error E by input samples and their corresponding actual output samples

To the connection weight value omega _jk To adjust

In the formula, e _jk Is the error signal between the desired output and the actual output.

The system output Y is obtained from equations (37) to (40) in combination with equation (34).

(7) Error detection, namely calculating errors (training errors E) of expected values of the economical efficiency, the dynamic performance and the safety of the driver and actual output values of the network according to the formulas (37) to (38), ending the training if the training errors E are less than or equal to epsilon, otherwise, judging whether the number m of hidden neurons reaches the maximum value m _max If the maximum value is reached, stopping training, and if the maximum value is not reached, increasing one neuron number of the hidden layer for next training.

S23, carrying out quantitative model test on economical efficiency, dynamic performance and safety expectation of a driver. Setting a test error upper limit value gamma, taking an input set in the normalized test sample as the input of the trained RBF neural network, obtaining an output result after the RBF neural network is used for calculation, and performing inverse normalization on the output result according to the following formula

In the formula, V is the result after the sample is subjected to inverse normalization.

Then normalized by

In the formula, beta-beta = e, d, s respectively represent economy, power and safety;

ω _β -a desired quantification of the property β;

V _β -an output value of the RBF neural network corresponding to the expectation of the property β;

V _e 、V _d 、V _s the result is output by the RBF neural network in economy, dynamic performance and safety.

Comparing the result after normalization and standardization with the output set in the original test sample after standardization, and if the test errors of 90% of the test samples are less than or equal to the upper limit value gamma of the test errors, storing the RBF neural network after training; otherwise, training the network again, and testing until the testing requirements are met.

Through the steps, the quantitative model of the economical efficiency, the dynamic performance and the safety expectation of the driver can be obtained. S3, quantifying economical, dynamic and safety expectations of drivers in real time

The driver economy, dynamics, and safety expectations are quantified in real time as follows.

The method comprises the steps of collecting information of vehicle speed, gradient, accelerator pedal opening and brake pedal opening in real time, using the information as input, and calculating an economic expected value, a dynamic expected value and a safety expected value of a driver by using an established economic, dynamic and safety expected quantitative model of the driver.

The technical scheme of the invention is illustrated by examples:

s1 data acquisition and processing

S11, different types of drivers are selected, the automobile is driven on roads with different gradients (three road test places are provided, wherein a Pi city area in the city of Sichuan province is subjected to level road test data acquisition, a Betula platyphylla road (close to a dragon fountain area) in a Jinjiang area in the city of Sichuan province and a Longchi tourist road near a purple plateau reservoir in the city of Sichuan province are subjected to ramp test data acquisition), the automobile runs under the working conditions of constant speed, acceleration and deceleration, and driving operation characteristic physical quantities (opening degree of an accelerator pedal and opening degree of a brake pedal), vehicle running state information (vehicle speed), road information (gradient) and subjective evaluation of the drivers on economy, power and safety are acquired.

S12, screening out dead spots in the data to obtain 8310 groups of original test data, wherein part of the original test data are shown in the table 1.

TABLE 1 original test data (parts)

In table 1, a positive value of the slope value indicates an upward slope, a value of 0 indicates a level road, and a negative value indicates a downward slope; the dereferencing range of the opening degree of the accelerator and the opening degree of the brake pedal is [0, 100]; the subjective evaluation value range of the driver performance expectation is [0, 10].

S13 determines training data and test data. The method comprises the steps of dividing the vehicle speed, the gradient, the accelerator pedal opening degree and the brake pedal opening degree and corresponding economical efficiency, dynamic performance and safety subjective evaluation into regions according to the vehicle speed, the gradient, the accelerator pedal opening degree and the brake pedal opening degree within possible value ranges of 10 and the like, randomly selecting the processed vehicle speed, gradient, accelerator pedal opening degree and brake pedal opening degree according to the proportion of 60%, 20% and 20% in each region, forming three types of data A, B and C with the corresponding economical efficiency, dynamic performance and safety subjective evaluation, forming training data by the type A and the type C data, forming test data by the type B and the type C data, wherein the training data totally comprise 4396 groups, and the test data totally comprise 3324 groups.

S2, establishing quantitative model of economical efficiency, dynamic performance and safety expectation of driver

S21, normalizing the data. The training data and the test data were normalized by equation (30), and the normalized partial data are shown in table 2.

TABLE 2 normalized data (parts)

S22, training of the quantitative model of the economical efficiency, the dynamic performance and the safety expectation of the driver. And training the RBF neural network by using the normalized training data until the training error meets the requirement. The RBF neural network carries out quantitative model training on the economical efficiency, dynamic performance and safety expectation of a driver, and the specific steps are as follows:

(1) Loading a normalized training sample set: the RBF neural network takes the speed, the gradient, the opening degree of an accelerator pedal and the opening degree of a brake pedal as input, namely an input vector X = (X) ₁ ,x ₂ ,x ₃ ,K,x _n ) ^T The value of n is 4, and the economic expectation, the dynamic expectation and the safety expectation of the driver are taken as the output of the RBF neural network model, namely an output vector Y = (Y) ₁ ,y ₂ ,y ₃ ,K,y _s ) ^T The value of s is 3, and an input set X corresponding to the normalized 4396 groups of training data is used ₁ ,X ₂ ,...,X _p (where p = 4396) and an output set Y ₁ ,Y ₂ ,...,Y _p And loading into an RBF neural network model.

(2) Setting parameter values of the RBF neural network: setting the upper limit of error epsilon to 0.01, initializing the number m of hidden layer neurons to 1, and setting the maximum number m of hidden layer neurons _max The value was 2000.

(3) Initializing a cluster center, namely randomly extracting m different samples from a training sample set to be used as an initial cluster center h _j (b) (j =1, 2.... M), b is the number of iterations, let b =1.

(4) Calculating the center h of the basis function _j 。

(5) Calculating the variance σ of the basis function as follows _j 。

(6) Computing system output Y = (Y) ₁ ,y ₂ ,y ₃ ) ^T I.e. driver economyExpectations, dynamics expectations, and safety expectations.

(7) And (3) error detection, namely calculating errors between expected values of the economical efficiency, the dynamic performance and the safety of the driver and actual output values of the network according to the formulas (8) to (9), ending the training if the training error E is less than or equal to 0.01, otherwise judging whether the number m of the hidden layer neurons reaches the maximum value of 2000, stopping the training if the number m of the hidden layer neurons reaches the maximum value, and increasing one hidden layer neuron number if the number m of the hidden layer neurons does not reach the maximum value, and performing the next training. By training when the number of hidden layer neurons is 520, the training error E =0.01 of the RBF neural network, and the error of the training process changes as shown in FIG. 13.

And S23, carrying out quantitative model test on the economical efficiency, dynamic performance and safety expectation of the driver. Setting the upper limit value gamma of the test error to be 0.1, taking the normalized test sample (3324 groups of data in total) as the input of the trained RBF neural network, calculating by the RBF neural network to obtain an output result, performing inverse normalization and normalization on the output result in turn according to the formulas (41) and (42), and comparing the obtained result with an output set corresponding to the normalized original test sample, wherein the result is shown in figure 14.

As can be seen from the figure, in terms of the quantitative result with desirable economy, the absolute error between the model quantitative result and the real vehicle test result after standardization accounts for 90.94% of the total data within 0.05%, and the absolute error accounts for 98.80% of the total data within 0.1%; in terms of the quantitative result of the dynamic expectation, the absolute error between the model quantitative result and the real vehicle test result after standardization accounts for 87.06% of the total data within 0.05, and the absolute error accounts for 96.57% of the total data within 0.1; in terms of the safety expectation quantification result, the absolute error between the model quantification result and the real vehicle test result after standardization accounts for 92.60% of the total data within 0.05, and the absolute error accounts for 98.35% of the total data within 0.1. The established quantitative models of the economical efficiency, the dynamic performance and the safety of the driver simultaneously meet the test requirements, and the network is saved.

Through the steps, the expected quantitative model of the economical efficiency, the dynamic performance and the safety of the driver is obtained.

S3 real-time quantification of economic, dynamic and safety expectations of drivers

The method comprises the steps of collecting information of vehicle speed, gradient, opening degree of an accelerator pedal and opening degree of a brake pedal in real time, using the information as input, and calculating an economic expected value, a dynamic expected value and a safety expected value of a driver by using an established economic, dynamic and safety expected quantitative model of the driver.

Implementation method

FIG. 15 is a flow chart of a driver economic, dynamic and safety expectation quantification model. Firstly, vehicle speed, gradient, accelerator pedal opening degree and brake pedal opening degree information and subjective evaluation of a driver on economy, dynamic performance and safety are collected through a road test, dead spots in the data are screened out, training data and test data are obtained, and a driver performance expectation quantification model is established in an off-line mode through the processed data. After the offline modeling is completed, the corresponding code is generated using the model and downloaded into the automatic Transmission Control Unit (TCU).

Claims (3)

1. An optimization method for an optimal comprehensive gear shifting rule with economy, dynamic performance and safety is characterized by comprising the following steps of:

(1) Determining the form of each partial objective function

Constructing each branch objective function by adopting an integral variable limit function summation mode: respectively taking the lower speed limit and the upper speed limit of a gear shifting speed optimizing interval as the lower integral limit and the upper integral limit, performing fixed integration on a low gear performance index, respectively taking the gear shifting speed and the upper speed limit of the gear shifting speed optimizing interval as the lower integral limit and the upper integral limit, performing fixed integration on a high gear performance index, and finally summing the two integral values, wherein the calculation formula is as follows

Where f (x) is a sub-objective function for a certain performance, x is the shift speed, v _min 、v _max Lower and upper vehicle speed limits, y, for the gear shift vehicle speed optimization interval _k 、y _k+1 Performance indexes of a gear k and a gear k +1 under corresponding vehicle speedsA value;

(2) Establishing an economic, dynamic and safety index subobjective function

(1) Establishing an economic objective function

An engine oil consumption rate neural network model and an engine forced idling oil consumption rate model which are established according to experimental data adopt a mode of integral variable-limit function summation, and the segmented integral summation of the oil consumption rate of a gear shifting speed interval to the speed is used as an economic subobjective function respectively aiming at an accelerator pedal operation process and a brake pedal operation process;

a) Establishing an economic objective function of the accelerator pedal operation process

Wherein x is the shift speed, Q _e1 The fuel consumption rate of the engine is shown, k is a gear, and thr1 is the opening degree of an accelerator pedal; v is the vehicle speed;

b) Establishing an economic objective function for a brake pedal actuation process

In the formula, Q _e2 The forced idling oil consumption rate of the engine is obtained;

(2) establishing dynamic subobjective function

An engine torque neural network model and an engine anti-drag torque model which are established according to experimental data adopt a mode of integral variable limit function summation, and the sectional integral summation of the acceleration of a gear shifting vehicle speed interval to the vehicle speed is used as a dynamic subobjective function

F _i ＝mgsinθ (6)

F _f ＝mgfcosθ (7)

In the formula, F _t (k) For k-gear driving force, F _i As ramp resistance, F _f As rolling resistance, F _w For air resistance, T _e For engine output torque, i _g (k) For the transmission in k-speed ratio, i ₀ Is the main reducer transmission ratio eta _T For driveline mechanical efficiency, r is wheel radius, m is vehicle mass, θ is road slope angle (uphill positive, downhill negative), f is rolling resistance coefficient, C _D The calculation formula is as follows, wherein A is windward area, v is vehicle speed, and delta (k) is k-gear rotating mass conversion coefficient

In the formula I _w Is the moment of inertia of the wheel, I _e Is the flywheel moment of inertia of the engine;

(3) establishing security objective functions

In the braking process, the mode of integral variable limit function summation is adopted, the integral summation of the deceleration of the gear shifting vehicle speed interval to the vehicle speed in sections is used as a safety subobjective function,

in the formula, F _eb (k) For k gear equivalent braking force, F _b Is the braking force; f _eb (k) Is calculated as follows

In the formula, T _ft The engine is used for resisting the drag torque;

(4) quadratic processing of an objective function

Carrying out secondary processing on each branch objective function: converting the partial objective function from a maximum value type to a minimum value type, and limiting the range of each partial objective function value in a [0,1] interval through normalization;

economic subobjective function after secondary processing

In which f is the accelerator pedal operating process and the brake pedal operating process _e (x) Are respectively f _e1 (x) And f _e2 (x)；

Dividing the economy under the optimized vehicle speed range into the maximum value and the minimum value of an objective function;

dynamic subobjective function after secondary processing

In the formula (I), the compound is shown in the specification,

dividing the maximum value and the minimum value of the dynamic property under the optimizing vehicle speed range into a target function;

security sub-objective function after secondary processing

In the formula (I), the compound is shown in the specification,

dividing the maximum value and the minimum value of the objective function for the safety in the optimized vehicle speed range;

(3) Constructing a comprehensive evaluation function

A square weighting and ideal point method is adopted to construct a comprehensive evaluation function which gives consideration to economy, dynamic property and safety

In the formula, F _e ^* Sub-objective function values after secondary processing for economically optimal shift points, F _d ^* The branch objective function value after the secondary processing corresponding to the dynamic optimal gear shifting point, F _s ^* Sub-objective function value after secondary processing, omega, for safety-optimized shift points _e 、ω _d 、ω _s The method comprises the following steps of (1) obtaining an economic weight value, namely an economic expectation, a dynamic weight value, namely a dynamic expectation, and a safety weight value, namely a safety expectation;

(4) Establishing optimal constraint conditions of gear shifting rules

(1) Establishing optimal constraint conditions of gear shifting rules in accelerator pedal operation process

a) The vehicle acceleration after the gear shift cannot be less than 0, i.e.

b) The engine speed should be at the lowest stable speed n corresponding to the opening degree of the accelerator pedal _eαmin And the maximum rotational speed n _eαmax In between, i.e.

c) Aiming at the downhill process, according to the design vehicle speed upper limit corresponding to each gradient in the table 1, determining the additional constraint of the downhill upshift regular optimization, namely: if the current opening degree of the accelerator pedal is kept, the upshifting can lead the automobile to be accelerated to be more than the current ramp design vehicle speed upper limit, the upshifting is forbidden,

g ₄ (x)＝v _max (i,k+1)-v _max (i)≤0(i＜0) (19)

in the formula, v _max (i, k + 1) is the maximum vehicle speed that can be achieved by driving on a slope with a gradient i with k +1 gear at the current accelerator pedal opening, v _max (i) Designing an upper limit of the vehicle speed for driving the vehicle on the slope with the gradient i;

TABLE 1 upper limit of design vehicle speed corresponding to different slopes

(2) Establishing optimal constraint conditions of gear shifting rules in brake pedal operation process

a) The vehicle acceleration after the gear shift cannot be greater than 0, i.e.

b) The engine speed should be at the lowest stable speed n _emin And a forced idling maximum speed n _eimax In between, i.e

(5) Establishing a mathematical model of a gear shifting law optimization problem

(1) Establishing a mathematical model of a gear shifting law optimization problem in an accelerator pedal operation process

(2) Establishing a mathematical model of a gear shifting law optimization problem in a brake pedal operation process

(6) Optimizing shift schedules

Uniformly traversing possible value ranges of the opening degree of an accelerator pedal and the opening degree of a brake pedal by setting a slope (the slope of a flat road is 0) and a gear shifting performance expected combination, solving the established optimization problem mathematical model by utilizing a particle swarm optimization algorithm, solving an upshift rule of the accelerator pedal in the operation process and a downshift rule of the brake pedal in the operation process, and calculating the downshift rule of the accelerator pedal in the operation process (or the upshift rule of the brake pedal in the operation process) according to the optimally solved upshift rule (or downshift rule);

specifically, taking a particle swarm optimization algorithm as an example, the method specifically comprises the following steps:

1) Taking a comprehensive evaluation function as a fitness function, taking the vehicle speed as a particle position, calling a particle swarm optimization algorithm, and solving the up/down gear vehicle speed under the given gradient, the expected combination of gear shifting performance and the opening degree of an accelerator/brake pedal;

2) Traversing each determined accelerator/brake pedal opening, repeating the step 1), and solving the upshifting/downshifting speed u of different accelerator/brake pedal openings under the given gradient and gear shifting performance expectation combination _aup /u _adown Forming an up/down shift curve under the given gradient and the expected combination of the gear shifting performance;

3) For the operation process of the accelerator pedal, calculating a downshift curve under the combination of a given gradient and expected gear shifting performance according to the following formula

In the formula u _adown (k) The gear speed of the gear k is reduced for the gear k + 1; u. u _aup (k) The gear-up speed is k liter and k +1 gear; k is a radical of formula _max The highest gear is set; an ₁ 、An ₂ An is a downshift coefficient when the opening degree of An accelerator pedal is less than or equal to 30 percent ₁ ＝0.4、An ₂ =0.6, an when the opening of the accelerator pedal is greater than 30% ₁ ＝0.15、An ₂ ＝0.4；

4) For the brake pedal operation process, an upshift curve under a given slope and expected gear shifting performance combination is calculated according to the following formula

In the formula, bn ₁ 、Bn ₂ Taking Bn as the up-shift coefficient ₁ ＝0.1、Bn ₂ ＝0.4；

5) For different expected combinations of gear shifting performance at a given gradient (the operation process of the accelerator pedal is a combination of an economical weight and a dynamic weight; the operation process of the brake pedal is the combination of the economical weight and the safety weight), and the steps 1) to 4) are repeated to obtain gear shifting curves corresponding to different expected combinations of gear shifting performance under a given slope;

6) Setting different gradient values (the gradient of the flat road is 0), and repeating the steps 1) to 5) to obtain gear shifting curves corresponding to different expected combinations of gear shifting performance under different gradients;

through the steps, the gear shifting rules under different gradients, different economic weights and dynamic weights in the accelerator pedal operation process or the gear shifting rules under different gradients, different economic weights and safety weights in the brake pedal operation process can be obtained;

(7) Automatic vehicle transmission control

During actual running of the stepped automatic transmission vehicle, firstly, gear shifting rules under different gradients, different economic expectations and dynamic expectations (accelerator pedal operation process) or different economic expectations and safety expectations (brake pedal operation process) obtained through offline optimization calculation are stored in a gear shifting rule database in a TCU in the form of a data table, a transmission control unit TCU acquires and processes signals of the gradients, the vehicle speed, the accelerator pedal opening and the brake pedal opening in real time, the economic, dynamic and safety expectations of a driver are calculated by utilizing a quantitative model of the economic, dynamic and safety expectations of the driver, a target gear is determined through the matched gear shifting rules according to the gradients, the economic expectations and the power expectations from the gear shifting rule database stored in the TCU, the gear shifting rules matched with the current gradients, the economic expectations and the power expectations are selected according to the vehicle speed and the accelerator pedal opening, a target gear is determined according to the brake pedal operation process, the gear shifting rules are determined according to the gradients, the economic expectations and the safety expectations, the target gear shifting performance is determined according to the optimal gear shifting rules of the vehicle speed, the gear shifting laws, the target gear shifting performance is determined, and the target gear shifting performance is determined according to the optimal gear shifting laws and the optimal shift performance of the vehicle is determined by a comprehensive vehicle driving control system.

2. An economic, dynamic and safety combined optimal shift schedule optimization method according to claim 1, characterized by the steps of: the specific steps of calling the particle swarm optimization algorithm are as follows:

a) Initializing the particle population and setting an acceleration constant c ₁ 、c ₂ (ii) a Setting a maximum iteration number M; setting a population size N; setting the maximum value w of the inertial weight _max Minimum value w _min ；

b) The initial position p of the first particle is given as the initial value of the position and velocity of all particles _l ⁰ At [ adjacent high gear lowest vehicle speed u _hmin Phase of changeHighest speed u of adjacent low gear _lmax ]Internal random values (l =1, \8230;, N); initial velocity v of the first particle _l ⁰ In the [0,0.9 ]]Internal random values (l =1, \8230;, N); calculating the fitness corresponding to each particle initial position according to a comprehensive evaluation function (the vehicle speed in the function is taken as the particle position); respectively taking the current particle position and the corresponding fitness as the local optimal particle position pbest _l And local optima pbestFitness _l (ii) a Respectively taking the minimum value of all particle fitness values and the particle positions corresponding to the minimum value as a global optimal value gbestFitness and a global optimal particle position gbest;

c) For each particle, iteratively optimizing starting from n =1 and ending with n = M; and for the nth iteration, the method comprises the following steps:

c1 Computing inertial weights w

c2 For each particle in the population), repeating the steps of:

c21 In [0,1)]Taking random values r within the range ₁ 、r ₂ ，

c22 Update the velocity and position of the particle

v _l ⁿ ＝wv _l ^n-1 +c ₁ r ₁ (pbest _l -p _l ^n-1 )+c ₂ r ₂ (gbest-p _l ^n-1 ) (28)

Wherein v is _l ^k The velocity of the nth iteration particle l; v. of _l ^n-1 The velocity of the (n-1) th iteration particle l; p is a radical of _l ^n-1 For the position of the n-1 st iteration particle l,

p _l ⁿ ＝p _l ^n-1 +v _l ^n-1 (29)

wherein p is _l ⁿ Is the position of the nth iteration particle l;

c23 Combining with a gear shifting vehicle speed constraint condition, processing the position boundary crossing of the particles: if it isp _l ⁿ ＜u _hmin Let p stand for _l ⁿ ＝u _hmin (ii) a If p is _l ⁿ ＞u _lmax Let p stand for _l ⁿ ＝u _lmax ，

c24 Calculate fitness of particle l pbestFitness _l ⁿ ，

c25 If pbestFitness _l ⁿ ＜pbestFitness _l Then pbestFitness _l ＝pbestFitness _l ⁿ ，pbest _l ＝p _l ⁿ Else pbestFitness _l And pbest _l Keeping the same;

c26 If pbestFitness _l < gbestFitness, then gbestFitness = pbestFitness _l ，gbest＝pbest _l ；

c3 Determining the upshift speed, and after all the particles are optimized according to the step c 2), taking the global optimal particle position gbest corresponding to the global optimal value gbestFitness as the upshift speed under the given gradient, the expected combination of the gear shifting performance (the economic weight and the dynamic weight) and the accelerator pedal opening, namely u _aup ＝gbest；

The method comprises the following steps of calling a particle swarm optimization algorithm, solving the substep of solving the downshift vehicle speed under the given slope, the expected gear shifting performance combination and the brake pedal opening degree, and is the same as the substep of calling the particle swarm optimization algorithm to solve the upshift vehicle speed under the given slope, the expected gear shifting performance combination and the accelerator pedal opening degree, except that the expected gear shifting performance combination is expressed by an economic weight and a safety weight, and a dynamic weight in a comprehensive evaluation function (formula (15)) for calculating the fitness is zero.

3. The optimization method for the comprehensive optimal gear shifting schedule of the economy, the power performance and the safety as claimed in claim 1 is characterized in that the establishment of the quantitative model of the economical, the power performance and the safety expectation of the driver comprises the following specific steps:

s1 data acquisition and processing

S11, drivers of different types are selected, the driven automobile runs on roads with different gradients respectively under the working conditions of constant speed, acceleration and deceleration, and the driving and control characteristic physical quantity is collected: the method comprises the following steps of (1) carrying out subjective evaluation on the opening degree of an accelerator pedal, the opening degree of a brake pedal, the vehicle speed, the gradient and the economy, the dynamic property and the safety of a driver;

different types of driver selection schemes: the driving styles of the economic type, the dynamic type and the safe type are respectively 25%, the driving style of the compromise type is 25%, and the total number of people is not less than 40;

the slope coverage range is-15 degrees (downhill) to +15 degrees (uphill);

subjective evaluations of the driver for economy, dynamics, and safety are all expressed in numbers of "0-10", with "0" indicating the lowest expected value for the performance and "10" indicating the highest expected value for the performance;

each test needs 1 driver, 1 driver-mounted recorder and 1 safety driver, the driver carries out driving operation under corresponding working conditions and describes subjective evaluation of the driver on economy, dynamic performance and safety in real time; the vehicle-mounted recorder records the vehicle speed, the gradient, the opening degree of an accelerator pedal, the opening degree of a brake pedal and subjective evaluation data of a driver; a safety worker observes the surrounding environment of the experimental vehicle on the vehicle and informs a driver in advance when a dangerous condition is foreseen so as to ensure the safety of the experimental process;

1) And (3) under a uniform working condition: a driver drives a vehicle, and the vehicle keeps running for 3s at a constant speed of 20km/h, 40km/h, 60km/h and 80km/h respectively;

2) Acceleration condition: a driver drives a vehicle and respectively carries out slow acceleration, medium acceleration and rapid acceleration of 0-80km/h according to the driving habit of the driver;

3) And (3) deceleration working condition: the driver drives the vehicle and decelerates from 80km/h to 0km/h respectively through slow deceleration, medium deceleration and rapid deceleration;

s12, screening out dead pixels in the data to obtain original test data;

s13 determining training data and testing data

Dividing the vehicle speed, the gradient, the accelerator pedal opening and the brake pedal opening and corresponding economical, dynamic and safety subjective evaluation into intervals according to the vehicle speed, the gradient, the accelerator pedal opening and the brake pedal opening within possible value ranges of 10 and the like, randomly selecting the processed vehicle speed, gradient, accelerator pedal opening and brake pedal opening according to the proportion of 60%, 20% and 20% in each interval, and forming three types of data A, B and C with the corresponding economical, dynamic and safety subjective evaluation, forming training data by the type A and C data, and forming test data by the type B and C data;

s2, establishing quantitative model of economical efficiency, dynamic performance and safety expectation of driver

The establishment process of the quantitative model of the economic, dynamic and safety expectations of the driver comprises the following steps:

s21, data normalization; the training data and the test data are normalized, the normalized data range is [ -1,1], and the formula is as follows:

in the formula (I), the compound is shown in the specification,

v, z-values before and after sample data normalization, respectively;

v _min 、v _max -minimum and maximum values of the sample data, respectively;

s22 driver economic, dynamic and safety expectation quantitative model training

Training the RBF neural network by using the normalized training data until the training error meets the requirement;

the adopted RBF neural network structure is a three-layer static feedforward neural network and comprises an input layer, a hidden layer and an output layer;

the mathematical description of the layers is as follows:

inputting a vector:

X＝(x ₁ ,x ₂ ,x ₃ ,K,x _n ) ^T (31)

in the formula, n is the number of input vector nodes;

the excitation function of the j node of the hidden layer adopts a Gaussian function as follows

In the formula, h _j And σ _j Respectively representing the center and the width of a basis function at the jth node of the hidden layer, wherein m is the number of the hidden layer nodes;

output vector

Y＝(y ₁ ,y ₂ ,y ₃ ,K,y _s ) ^T (33)

In the formula, s is the number of nodes of an output layer;

the output layer neuron adopts a linear excitation function, and the k output of the RBF can be obtained by the function as follows:

in the formula, omega _jk (j =1,2, \8230;, m; k =1,2, \8230;, s) represents a weight between the jth node of the hidden layer and the kth node of the output layer;

the parameter needing to be learned in RBF neural network training is the center h of the basis function _j Variance σ _j And a connection weight w _jk The RBF neural network is adopted to carry out quantitative model training of the economical efficiency, dynamic performance and safety expectation of a driver, and the training comprises the following specific steps:

(1) Loading a normalized training sample set: assuming that the input set in the normalized training sample set is X ₁ ,X ₂ ,...,X _p P is the number of training samples, and the corresponding target output set is Y ₁ ,Y ₂ ,...,Y _p Respectively loading the normalized input set and output set into the RBF neural network model;

(2) Setting parameter values of the RBF neural network: setting an upper limit epsilon of an error, an initial value of the number m of neurons in an implicit layer and the maximum number m of neurons thereof _max ；

(3) Initializing a cluster center, namely randomly extracting m non-random numbers from a training sample setThe same sample is taken as an initial clustering center h _j (b) (j =1,2,.., m), b is the number of iterations, let b =1;

(4) Calculating the center of the basis function:

1) Calculating the distance from all samples to the initial clustering center, i.e., | | X _i -h _j (b)||,i＝1,2,...,p；

2) For sample input X _i Classifying according to a minimum distance principle: namely when

When, X _i I.e., classified as jth;

3) Center of gravity is adjusted by the following formula

h _j (b+1)＝h _j (b)+η[X _i -h _j (b)] (35)

In the formula, eta is the learning rate, and eta is more than 0;

4) If h is _j (b+1)≠h _j (b) B = b +1 and go to step 1), otherwise clustering is finished to obtain the final basis function center h _j (h _j ＝h _j (b))；

(5) Calculating the variance of the basis function according to

In the formula (d) _max For the maximum distance between the selected centers,

(6) Computing system output basis function center h _j And its variance σ _j After the determination, the method adopts a least square method to adjust omega _jk Input samples and corresponding actual output samples are enabled to minimize training error E

To the connection weight omega _jk To adjust

In the formula, e _jk For an error signal between the desired output and the actual output,

obtaining a system output Y according to equations (37) to (40) in combination with equation (34);

(7) Error detection, namely calculating the error of the economic, dynamic and safety expected value of the driver and the actual output value of the network, namely a training error E according to the formulas (37) to (38), if the training error E is less than or equal to epsilon, ending the training, otherwise, judging whether the number m of hidden neurons reaches the maximum value m _max If the maximum value is reached, stopping training, and if the maximum value is not reached, increasing the number of the neurons of the hidden layer by one for carrying out next training;

s23 driver economic, dynamic and safety expectation quantitative model test

Setting a test error upper limit value gamma, taking an input set in the normalized test sample as the input of the trained RBF neural network, obtaining an output result after the RBF neural network is used for calculation, and performing inverse normalization on the output result according to the following formula

In the formula, V is the result of sample inverse normalization,

then normalized by

In the formula, beta-beta = e, d, s respectively represent economy, power and safety;

ω _β -a desired quantification of the property β;

V _β -an output value of the RBF neural network corresponding to the expectation of the property β;

V _e 、V _d 、V _s the result is output by the RBF neural network in economy, dynamic performance and safety;

comparing the result after normalization and standardization with the output set in the original test sample after normalization, and if the test errors of 90% of the test samples are less than or equal to the upper limit value gamma of the test errors, storing the RBF neural network after training; otherwise, training the network again, and testing again until the testing requirements are met;

through the steps, an expected quantitative model of the economical efficiency, the dynamic performance and the safety of the driver can be obtained;

s3 real-time quantification of economic, dynamic and safety expectations of drivers

The real-time quantification process of the economic, dynamic and safety expectations of the driver is as follows: the method comprises the steps of collecting information of vehicle speed, gradient, accelerator pedal opening and brake pedal opening in real time, using the information as input, and calculating an economic expected value, a dynamic expected value and a safety expected value of a driver by using an established economic, dynamic and safety expected quantitative model of the driver.

Images