CN108629372A - Obtain experimental system and the driving style recognition methods of driving style characteristic parameter - Google Patents

Abstract

The present invention relates to an experimental system for obtaining characteristic parameters of driving style and a method for recognizing driving style. The experimental system includes a driver's console, a motor-to-drag test bench and a host computer. The driver's console is provided with a brake pedal, The accelerator pedal, the main controller and the simulated driving interaction module, the position signal output ends of the brake pedal and the accelerator pedal are respectively connected to the main controller, and the main controller is respectively connected to the simulated driving interactive module, the motor-to-drag test bench and the upper machine. Compared with the prior art, the present invention collects more accurate characteristic parameters through the experimental system and uses them in the neural network algorithm to accurately identify the driver's driving style, thereby improving the fuel economy of the vehicle.

Description

Experimental system for obtaining driving style characteristic parameters and driving style recognition method

technical field

The invention relates to the field of driver's driving style recognition, in particular to an experimental system for acquiring driving style characteristic parameters and a driving style recognition method.

Background technique

The concept of driving style refers to the behavior or habit of the driver when driving the vehicle. Therefore, the driving condition of the vehicle has a great relationship with the driver's personal state, values or behavior habits. Reckless drivers will frequently and significantly step on the Accelerate or brake the pedal, and its fuel economy will become worse. On the contrary, a moderate driver will lightly step on the pedal, so it is more fuel-efficient. That is to say, the driver's driving style will have a greater impact on the fuel economy of the car. Since each driver has his own unique driving style and the driving style is the behavior of the human brain after complex thinking, it cannot be described by simple mathematical formulas. The artificial neural network algorithm can simulate the way of thinking of the human brain and express it with mathematical results. Therefore, it can be combined with the artificial neural network algorithm for driving style recognition. Among them, the backpropagation neural network (BPNN) is currently the most widely used artificial neural network. Network model with powerful pattern recognition function. Therefore, this paper proposes an experimental system for obtaining characteristic parameters of driving style and a driving style recognition method based on neural network algorithm (BPNN). The driving style is recognized and applied to the control of the vehicle power system, thereby improving the fuel economy of the vehicle.

Contents of the invention

The purpose of the present invention is to provide an experimental system and a driving style recognition method for obtaining driving style characteristic parameters in order to overcome the above-mentioned defects in the prior art, in order to collect relatively accurate characteristic parameters through the experimental system for neural network algorithms, and then Accurately identify the driver's driving style to improve the fuel economy of the car.

The purpose of the present invention can be achieved through the following technical solutions:

An experimental system for obtaining driving style characteristic parameters, including a driver's console, a motor-to-drag test bench, and a host computer. The driver's console is equipped with a brake pedal, an accelerator pedal, a main controller, and a simulated driving interaction module , the position signal output ends of the brake pedal and the accelerator pedal are respectively connected to the main controller, and the main controller is respectively connected to the simulated driving interaction module, the motor-to-drag test bench and the host computer.

The simulated driving interaction module receives the selected type of simulated driving conditions, and simultaneously displays the preset driving state of the reference vehicle under the corresponding simulated driving conditions and the initial driving state of the operating vehicle in the same coordinate system according to the type of simulated driving conditions , the main controller outputs control commands according to the received position signals of the brake pedal and the accelerator pedal, the motor-to-drag test bench receives the control commands output by the main controller and outputs power and resistance data, and the main controller outputs according to the motor-to-drag test bench The real-time driving status of the operating vehicle is obtained from the power and resistance data of the operating vehicle, and the real-time driving status of the operating vehicle is fed back to the simulated driving interaction module. The simulated driving interactive module updates and displays the real-time driving status of the operating vehicle in real time to simulate the actual driving environment of the driver. .

The upper computer extracts the characteristic parameters according to the position signals of the brake pedal and the accelerator pedal forwarded by the main controller, and the characteristic parameters are used to identify the driver's style based on the neural network algorithm.

The motor-to-drag test bench includes a load motor, a flywheel, a gearbox, a two-speed transmission, a drive motor, a first speed torque sensor and a second speed torque sensor, and the output end of the load motor is connected to one end of the flywheel, so The other end of the flywheel is connected to one end of the gearbox, the other end of the gearbox is connected to one end of the two-speed transmission, the other end of the two-speed transmission is connected to the output end of the drive motor, and the first rotational speed torque sensor is located at Between the gearbox and the two-speed transmission, the second rotational speed torque sensor is arranged between the two-speed transmission and the drive motor, and the main controller is respectively connected to the control terminal of the load motor, the control terminal of the drive motor, and the first rotational speed The signal output end of the torque sensor and the signal output end of the second speed torque sensor, the power and resistance data correspond to the speed and torque of the driving motor output by the second speed torque sensor and the output of the first speed torque sensor The speed and torque of the load motor.

The main controller obtains the required power according to the received position signals of the brake pedal and the accelerator pedal, and controls the driving force of the driving motor and the resistance of the load motor according to the required power.

The characteristic parameters include the average value of the accelerator pedal opening, the standard deviation of the accelerator pedal opening, the average value of the change rate of the accelerator pedal opening, the standard deviation of the change rate of the accelerator pedal opening, the average value of the brake pedal opening, the brake pedal opening The standard deviation of the brake pedal opening degree, the average value of the change rate of the brake pedal opening degree and the standard deviation of the change rate of the brake pedal opening degree.

The main controller adopts Freescale MC9S12EQ512 microprocessor.

A driving style recognition method, realized by using the above-mentioned experimental system for obtaining characteristic parameters of driving style, the method includes the following steps:

S1: Use the experimental system to collect the characteristic parameters of different types of driving testers in different simulated driving conditions in multiple sets of set periods as training data. The characteristic parameters in the group setting period are used as test data;

S2: Normalize the training data;

S3: Establish a BPNN algorithm model with a mentor type, and set the training parameters of the BPNN algorithm model;

S4: Use the training data to train the BPNN algorithm model until the training target error is within the set range;

S5: Preprocessing the test data;

S6: Based on the BPNN algorithm model trained in step S4, identify the driving style of the driving tester according to the preprocessed test data.

The activation function of the BPNN algorithm model uses a logarithmic sigmoid function, the training target error is set to 0.01, the maximum number of iterations is 400, and the learning rate is set to 0.01.

The driving style identified by this method package is divided into reckless driving style and moderate driving style.

Compared with the prior art, the present invention has the following advantages: by using the designed experimental system, the characteristic parameters for neural network identification can be collected more accurately, which has breakthrough significance in the collection of characteristic parameters. Then use MATLAB software to identify the driving style based on the neural network algorithm, which improves the accuracy of the driving style identification, thereby improving the fuel economy of the car.

Description of drawings

Fig. 1 is a schematic diagram of the driver's console in the experimental system for obtaining characteristic parameters of driving style;

Figure 2 is a schematic diagram of the structure of the motor-to-drag test bench;

Fig. 3 is the working flow diagram of obtaining the characteristic parameters of driving style by the experimental system;

Fig. 4 is the interface design diagram of the simulated driving interaction module;

Figure 5 is a topological diagram of the experimental system;

Fig. 6 is the flowchart of identifying driver's style with BPNN algorithm;

Fig. 7 is the flowchart of BPNN algorithm;

Figure 8 is a program diagram of MATLAB driver style recognition.

In the figure, 1. Driver console, 11. Brake pedal, 12. Accelerator pedal, 13. Main controller, 14. Simulated driving interaction module, 2. Motor-to-drag test bench, 21. Load motor, 22. Flywheel , 23, gearbox, 24, two-speed transmission, 25, drive motor, 26, first speed torque sensor, 27, second speed torque sensor, 3, upper computer.

Detailed ways

The present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is carried out on the premise of the technical solution of the present invention, and detailed implementation and specific operation process are given, but the protection scope of the present invention is not limited to the following embodiments.

An experimental system for obtaining driving style characteristic parameters includes a driver's console 1, a motor-to-drag test bench 2, and a host computer 3. As shown in Figure 1, the driver's console 1 is provided with a brake pedal 11 and an accelerator pedal 12. , the main controller 13 and the simulated driving interaction module 14, the position signal output terminals of the brake pedal 11 and the accelerator pedal 12 are respectively connected to the main controller 13, as shown in Figure 2, the motor-to-drag test bench includes a load motor 21, a flywheel 22 , gearbox 23, two-speed transmission 24, drive motor 25, first rotational speed torque sensor 26 and second rotational speed torque sensor 27, the output end of load motor 21 is connected with one end of flywheel 22, and the other end of flywheel 22 is connected with gearbox One end of 23, the other end of gearbox 23 connects one end of two-speed transmission 24, the other end of two-speed transmission 24 connects the output end of driving motor 25, and the first rotational speed torque sensor 26 is located at gearbox 23 and two-speed transmission 24 Between, the second rotation speed torque sensor 27 is arranged between the two-speed transmission 24 and the driving motor 25, as shown in Figure 5, the main controller 13 is respectively connected to the simulated driving interaction module 14, the control end of the load motor 21, the driving motor 25, the signal output end of the first rotational speed torque sensor 26, the signal output end of the second rotational speed torque sensor 27 and the upper computer 3.

The main controller 13 controls the load motor 21 and the driving motor 25 in the motor-towing test bench 2 through the CAN bus according to the position of the pedals, so as to simulate the situation of driving resistance and power, and the load motor 21 drives the flywheel 22 to transmit the speed and torque. to the gearbox 23, the two rotational speed torque sensors receive the signals from the drive motor 25 and the load motor 21 and transmit them to the CAN bus; the host computer 3 performs data collection and analysis to collect driving style characteristic parameters and perform Neural network analysis, as shown in Figure 3, the data acquisition workflow of the entire experimental system is as follows:.

a: The driver selects the type of simulated driving conditions through the interface of the simulated driving interaction module 14 (as shown in FIG. 4 ).

b: The main controller 13 returns confirmation information to the interface of the simulated driving interaction module 14, and the simulated driving interaction module 14 simultaneously displays the presets of the reference vehicle under the corresponding simulated driving conditions in the same coordinate system according to the type of the simulated driving conditions Driving state and initial driving state of the operating vehicle.

c: The driver operates the accelerator pedal 12 and the brake pedal 11 to control the running of the operating vehicle, so that the operating vehicle follows the reference vehicle.

d: The position signals of the two pedals are sent to the main controller 13, and then the main controller 13 sends the pedal position signals to the host computer 3. The upper computer 3 extracts the characteristic parameters according to the position signals of the brake pedal 11 and the accelerator pedal 12 forwarded by the main controller 13, and the characteristic parameters are used to identify the driver's style based on the neural network algorithm.

The characteristic parameters extracted in one cycle include the average value acc _avg of the opening degree of the accelerator pedal 12 , the standard deviation acc _sd of the opening degree of the accelerator pedal 12 , the average value accd _avg of the opening degree change rate of the accelerator pedal 12 , the standard deviation accd of the opening degree change rate of the accelerator pedal 12 _sd , the average value accd _sd of the opening degree of the brake pedal 11 , the standard deviation brk _sd of the opening degree of the brake pedal 11 , the average value of the change rate of the opening degree of the brake pedal 11 brkd _avg , and the standard deviation of the change rate of the opening degree of the brake pedal 11 brkd _sd .

e: The main controller 13 outputs control commands according to the pedal position signal and the historical signal to control the load motor 21 and the drive motor 25 in the motor-to-drag test bench 2, and the motor-to-drag test bench 2 simulates the power and resistance of driving. A speed torque sensor collects the speed and torque data of the driving motor 25 and the load motor 21 in real time. Wherein, the rotational speed and the torque of the drive motor 25 output by the second rotational speed torque sensor 27 reflect the power situation of the simulated driving, as the power data of the simulated driving, the rotational speed and the torque of the load motor 21 output by the first rotational speed torque sensor 26 Reflect the resistance of simulated driving as the resistance data of simulated driving.

In step e, the main controller 13 obtains the required power according to the received position signals of the brake pedal 11 and the accelerator pedal 12, and controls the driving force of the drive motor 25 and the resistance of the load motor 21 according to the required power.

f: The motor-to-drag test bench 2 returns the speed and torque data of the running drive motor 25 and load motor 21 to the main controller 13 .

g: The main controller 13 obtains the real-time driving state of the operating vehicle through calculation according to the speed and torque data, and sends the real-time driving state data such as the operating position of the operating vehicle to the simulated driving interaction module 14, which is displayed on the interface of the simulated driving interaction module 14 Real-time updates show the real-time driving status of the operating vehicle. Afterwards, return to step c and repeat this cycle. If the simulated working condition is selected again, start again from step a.

In Fig. 4, wherein A is the reference vehicle controlled by the main controller 13, it is set to run by the program, and can be operated according to the 8 cycle working conditions (NYCC, ECE, UDDS, EUDC, HWFET, LA92, 1015, IM240) in Fig. 4 ) to run, and the cycle working condition is selected by the driver. When the driver selects the working condition, the main controller 13 will delay 3 seconds after receiving the letter and notify the driver that the corresponding working condition has changed color. The information has been confirmed and starts to control the reference vehicle at the same time run. B is realized by the driver through the control of two pedals, and the driver presses the control of B vehicle to follow A vehicle to drive under the selected working conditions.

The purpose of the above process is to simulate the actual driving environment of the driver and collect the characteristic parameters of the neural network used for driving style recognition. The data acquisition system of the upper computer 3 is designed using the LabVIEW software of NI Company.

In this embodiment, the display interface screen of the simulated driving interaction module 14 is a DMT10600T102_02WT type high-definition touch industrial LCD screen. This screen supports RS232 real-time serial port communication to facilitate real-time data interaction with the main controller 13. The main controller 13 adopts Freescale MC9S12EQ512 microprocessor, which has good performance in terms of stability, precision, speed and other performances. As shown in Figure 5, the pedal position (voltage signal) is collected by the ADC (analog-to-digital converter) of the main controller 13, uses RS232 communication between the simulated driving interaction module 14 and the main controller 13, the main controller 13, the motor pair The towing test bench 2 and the host computer 3 are connected together on the CAN (controller area network) for data communication.

The present invention proposes a driving style recognition method based on a neural network algorithm on the basis of utilizing the above-mentioned experimental system to obtain driving style characteristic parameters, as shown in Figure 6, specifically comprising the following steps:

S1: Use the experimental system to collect the characteristic parameters of different types of driving testers in different simulated driving conditions in multiple sets of set periods as training data. Set the characteristic parameters in the set period as the test data.

S2: Normalize the training data.

S3: Establish a BPNN algorithm model with a mentor type, and set the training parameters of the BPNN algorithm model.

S4: Use the training data to train the BPNN algorithm model until the training target error is within the set range.

S5: Preprocessing the test data.

S6: Based on the BPNN algorithm model trained in step S4, the driving style of the driving tester is identified according to the preprocessed test data, and the driving style is divided into a reckless driving style and a moderate driving style.

The following uses MATLAB to realize the driving style recognition method as an example.

1) Based on the above-mentioned experimental system, a reckless driving tester was selected to conduct a simulated driving test. For a certain driving condition, the characteristic parameters were extracted with a cycle of 60 seconds. The results are shown in Table 1. Among them, NYCC and UDDS represent the typical working conditions of urban congestion and the typical working conditions of suburbs in foreign countries.

Table 1 Characteristic parameters of reckless driving style

2) In the same way, choose a moderate driving tester to carry out the simulated driving test, and extract the characteristic parameters for the same driving condition with a cycle of 60 seconds. The results are shown in Table 2.

Table 2 Characteristic parameters of mild driving style

3) Select the BPNN algorithm type with tutor type for identification, as shown in Figure 7 as its algorithm flow chart. The collected data is processed, which can be divided into two processing processes: training process and testing process. For the selection of training data, for the reckless driver and the moderate driver, select the average value of 4 time period data as a group in the driving data of each working condition, and select two groups for each working condition, that is, two Each driver corresponds to 20 sets of training data. For test data selection, similar to the above training data, 4 sets of data are selected for each driving style as test data.

Step S4: Use MATLAB to write recognition codes for driver style recognition through the data after training and testing, and complete the whole recognition process so far. The MATLAB driver style recognition code is shown in Figure 8.

Wherein, 1. The training data is constructed as follows, each row of the matrix dataTrain (8 rows and 40 columns) represents a characteristic parameter, that is, the 8 rows correspond to the above-mentioned 8 driver style parameters respectively. The first 20 lists are 20 sets of data for reckless drivers, and the last 20 lists are 20 sets of data for moderate drivers. For example, a _(3,2) represents parameter 3 (accd _avg ) in the second group of test data for reckless drivers; b _(4,5) represents parameter 4 (accd _sd ) in the fifth group of test data for moderate drivers.

Perform normalization processing on the matrix dataTrain, and linearly change each row to the [-1,1] interval to obtain a normalized matrix. The normalization process is as follows: Take the first row as an example, its minimum value is min1, and its maximum value is max1. The relationship between a certain element a and the transformed a ₁ is as follows:

Set the result output matrix output (2 rows and 40 columns). The first row consists of 20 1s and 20 0s, and the second row consists of 20 0s and 20 1s. The 1 in the first row corresponds to the output of the reckless driving style, and the 0 corresponds to the mild type; the 0 in the second row corresponds to the reckless type, and 1 corresponds to the moderate type.

Then configure the parameters of BPNN. In this embodiment, the activation function of the BPNN algorithm model uses a logarithmic sigmoid function, the training target error is set to 0.01, the intermediate result period is 40, the maximum number of iterations is 400, and the learning rate is set to 0.01.

2. The settings of the test data matrix dataTest (8 rows and 8 columns) are as follows, and the meanings of its elements are the same as those of dataTrain.

Preprocess the matrix dataTest, taking the first row as an example, its minimum value is min2, and its maximum value is max2 (the minimum value min1 and maximum value of the first row of dataTrain before, and the maximum value is max1), and one of the elements a and the processed a ₁ The relationship looks like this:

In summary, the experimental system can combine the collection of neural network characteristic parameters with the use of MATLAB software to complete the recognition of driving style, and propose an experimental system for driving environment simulation and driving style recognition.

Claims (8)

1. An experimental system for obtaining driving style characteristic parameters is characterized in that it comprises a driver's console, a motor-to-drag test bench and an upper computer, and the driver's console is provided with a brake pedal, an accelerator pedal, a main control device and simulated driving interaction module, the position signal output terminals of the brake pedal and the accelerator pedal are respectively connected to the main controller, and the main controller is respectively connected to the simulated driving interactive module, the motor-to-drag test bench and the host computer; The simulated driving interaction module receives the selected type of simulated driving conditions, and simultaneously displays the preset driving state of the reference vehicle under the corresponding simulated driving conditions and the initial driving state of the operating vehicle in the same coordinate system according to the type of simulated driving conditions , the main controller outputs control commands according to the received position signals of the brake pedal and the accelerator pedal, the motor-to-drag test bench receives the control commands output by the main controller and outputs power and resistance data, and the main controller outputs according to the motor-to-drag test bench The real-time driving status of the operating vehicle is obtained from the power and resistance data of the operating vehicle, and the real-time driving status of the operating vehicle is fed back to the simulated driving interaction module. The simulated driving interactive module updates and displays the real-time driving status of the operating vehicle in real time to simulate the actual driving environment of the driver. ; The upper computer extracts the characteristic parameters according to the position signals of the brake pedal and the accelerator pedal forwarded by the main controller, and the characteristic parameters are used to identify the driver's style based on the neural network algorithm. 2. The experimental system for obtaining driving style characteristic parameters according to claim 1, wherein the motor-to-drag test bench includes a load motor, a flywheel, a gearbox, a two-speed transmission, a driving motor, a first rotational speed torque sensor and the second speed torque sensor, the output end of the load motor is connected to one end of the flywheel, the other end of the flywheel is connected to one end of the gearbox, and the other end of the gearbox is connected to one end of the two-speed transmission, the two The other end of the gear transmission is connected to the output end of the drive motor, the first rotational speed torque sensor is arranged between the gearbox and the two-speed transmission, and the second rotational speed torque sensor is arranged between the two-speed transmission and the drive motor, The main controller is respectively connected to the control terminal of the load motor, the control terminal of the driving motor, the signal output terminal of the first rotational speed torque sensor and the signal output terminal of the second rotational speed torque sensor, and the power and resistance data correspond to the first The rotational speed and torque of the driving motor output by the second rotational speed torque sensor and the rotational speed and torque of the load motor output by the first rotational speed torque sensor. 3. The experimental system for obtaining driving style characteristic parameters according to claim 2, wherein the main controller obtains the required power according to the received position signals of the brake pedal and the accelerator pedal, and controls the drive motor according to the required power The driving force and the resistance of the load motor. 4. The experimental system for obtaining driving style characteristic parameters according to claim 1, wherein said characteristic parameters include accelerator pedal opening average value, accelerator pedal opening degree standard deviation, accelerator pedal opening degree change rate average value, The standard deviation of the change rate of the accelerator pedal opening, the average value of the brake pedal opening, the standard deviation of the brake pedal opening, the average value of the change rate of the brake pedal opening, and the standard deviation of the change rate of the brake pedal opening. 5. The experimental system for obtaining driving style characteristic parameters according to claim 1, wherein the main controller adopts a Freescale MC9S12EQ512 microprocessor. 6. A driving style identification method is characterized in that, utilizes the experimental system that obtains driving style characteristic parameter as claimed in claim 1 to realize, and the method comprises the following steps: S1: Use the experimental system to collect the characteristic parameters of different types of driving testers in different simulated driving conditions in multiple sets of set periods as training data. The characteristic parameters in the group setting period are used as test data; S2: Normalize the training data; S3: Establish a BPNN algorithm model with a mentor type, and set the training parameters of the BPNN algorithm model; S4: Use the training data to train the BPNN algorithm model until the training target error is within the set range; S5: Preprocessing the test data; S6: Based on the BPNN algorithm model trained in step S4, identify the driving style of the driving tester according to the preprocessed test data. 7. The driving style recognition method according to claim 6, wherein the activation function of the BPNN algorithm model uses a logarithmic sigmoid function, the training target error is set to 0.01, the maximum number of iterations is 400 times, and the learning rate is set to 0.01. is 0.01. 8 . The driving style recognition method according to claim 6 , wherein the driving style identified by the method includes a reckless driving style and a moderate driving style.