Signal filtering method for electronic accelerator pedal of rice transplanter

Abstract

The invention relates to the technical field of agricultural equipment new energy driving, in particular to a rice transplanter electronic accelerator pedal signal filtering method, which comprises the following steps in each work cycle: S1, accelerator pedal signal acquisition and processing; s2, state prediction and variance prediction are carried out; s3, state updating and variance updating are carried out; and S4, outputting a filtering result. According to the method, an accelerator pedal signal is filtered by adopting a self-adaptive maximum correlation entropy Kalman filtering algorithm based on a student t kernel, and the algorithm comprises an innovative kernel bandwidth self-adaptive adjustment method of maximum correlation entropy Kalman filtering based on the student t kernel. Compared with a traditional filtering algorithm, the method has higher robustness on paddy field non-Gaussian noise, the robustness estimation capacity of the system on the opening degree of the accelerator pedal is effectively enhanced, and the control stability and precision of the electric drive system in the complex paddy field environment are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of new energy drive for agricultural equipment, and in particular to a method for filtering electronic accelerator pedal signals of a rice transplanter. Background Art

[0002] With the continuous advancement of new energy technologies, the electrification of rice transplanters is becoming a key trend in the development of agricultural machinery. Against this backdrop, the maturity and portability of electric drive systems and electronic control technologies, such as electronic accelerator pedals, already widely used in electric vehicles, offer a viable path for electrification and high-precision control of rice transplanters. However, when applied to rice transplanters, the uneven softness and hardness of paddy field soils can easily produce fine vibrations and slight bounces during operation. This can cause the driver's foot to tremble slightly when pressing the electronic accelerator pedal, resulting in noise in the electronic accelerator pedal signal. This problem affects the final calculated electronic accelerator pedal opening, preventing the driver's actual acceleration intention from being accurately reflected, reducing operation smoothness and efficiency. This phenomenon is more pronounced in paddy fields than on paved roads, and the noise it generates is often non-Gaussian. Therefore, a filtering method with greater robustness to non-Gaussian disturbances is needed to filter the electronic accelerator pedal signal of electric rice transplanters, thereby improving the control stability and accuracy of the transplanter's electric drive system. Currently, widely used accelerator pedal signal filtering methods include first-order low-pass filtering, sliding average filtering, median filtering, and innovative methods based on these methods. For example, patent CN110311651A discloses a vehicle accelerator pedal signal filtering method that adjusts the median filter window width based on the vehicle's vertical acceleration to improve its robustness against uneven terrain. These methods are simple in structure and easy to implement, but they have significant limitations in practical applications: they have limited ability to suppress outliers and pulse interference, which can easily cause output fluctuations and affect control accuracy. Some systems attempt to use the more robust traditional Kalman filter algorithm and its derivatives (such as the unscented Kalman filter and the adaptive Kalman filter) to dynamically estimate the accelerator pedal signal. For example, in the paper "Dynamic Smoothness Control for Dual-Motor-Independent-Drive Electric Vehicles Based on Kalman Filter" [Dong et al., Energy Procedia, 2016], researchers filter the accelerator pedal signal using a Kalman filter to reduce the dynamic impact caused by mechanical backlash. However, these methods all assume that system noise is Gaussian in distribution. In this context, the Maximum Correlation Entropy Kalman Filter (MCEKF), an algorithm that incorporates the correlation entropy criterion from information theory, can effectively improve the robustness of filters in non-Gaussian environments. Compared with traditional Kalman filtering, the MCEKF no longer relies on precise modeling of noise statistics. Instead, it enhances the system's adaptability to outliers by maximizing the correlation entropy between state estimates and observations. This approach can suppress output deviations caused by non-ideal inputs while maintaining estimation accuracy.Based on this characteristic, the maximum correlation entropy Kalman filter is well-suited for filtering accelerator pedal signals in rice transplanter control systems. It is expected to improve the stability of vehicle dynamic response and control system robustness while reducing filtering latency and enhancing the driving experience. Furthermore, while the traditional maximum correlation entropy Kalman filter uses a Gaussian kernel as its kernel function, the Student's t kernel has stronger non-Gaussian noise processing capabilities than the Gaussian kernel, making it more suitable for environments with significant non-Gaussian noise, such as paddy fields. Currently, the maximum correlation entropy Kalman filter has not been applied to accelerator pedal signal filtering. Therefore, a maximum correlation entropy Kalman filter based on the Student's t kernel can be used to filter the accelerator pedal signal. Furthermore, considering that the traditional fixed-kernel maximum correlation entropy Kalman filter based on the Student's t kernel has poorer tracking performance than the traditional Kalman filter, its kernel bandwidth can be adjusted in a direction that increases the sensitivity of the cost function to the observed state. This results in an innovative kernel bandwidth adaptive adjustment method for the maximum correlation entropy Kalman filter based on the Student's t kernel, further improving the accuracy of the maximum correlation entropy Kalman filter based on the Student's t kernel. This improves the filtering accuracy of the rice transplanter's electronic accelerator pedal signal in complex paddy field environments and provides key support for high-robustness control of agricultural electric equipment. Summary of the Invention

[0003] To address the above technical issues, the present invention provides a method for filtering the electronic accelerator pedal signal for a rice transplanter. This method employs an adaptive maximum correlation entropy Kalman filter algorithm based on a Student's t kernel to filter the accelerator pedal signal. The algorithm incorporates an innovative method for adaptively adjusting the kernel bandwidth of the maximum correlation entropy Kalman filter, also based on the Student's t kernel. Compared to traditional filtering algorithms, this method exhibits greater robustness to non-Gaussian noise in paddy fields, effectively enhancing the system's robust estimation of accelerator pedal opening, and improving the control stability and precision of the electric drive system in complex paddy field environments.

[0004] In order to achieve the above object, the present invention provides the following technical solutions:

[0005] A rice transplanter electronic accelerator pedal signal filtering method comprises the following steps in each working cycle:

[0006] S1, accelerator pedal signal acquisition and processing;

[0007] The controller of the rice transplanter drive system receives the analog accelerator pedal signal sent by the electronic accelerator pedal, and converts the analog accelerator pedal signal sent by the accelerator pedal into a digital accelerator pedal signal as the observed value z of the accelerator pedal signal at the kth moment. k ;

[0008] S2. Perform state prediction and variance prediction;

[0009] The state prediction process is:

[0010]

[0011] In formula 1, is the estimated value of the accelerator pedal signal at the k-1th moment, in V; is the prior estimate of the accelerator pedal signal at the kth moment, in V;

[0012] The variance prediction process is:

[0013]

[0014] In formula 2, p k-1 is the accelerator pedal signal error variance at the k-1th moment, in V 2 ; is the priori accelerator pedal signal error variance at the kth moment, in V 2 ; q is the process noise variance of the accelerator pedal signal, in V 2 ;

[0015] S3, perform status update and variance update;

[0016] The status update process is as follows:

[0017]

[0018] The variance update process is:

[0019]

[0020] In formula 3 and formula 4, is the estimated value of the accelerator pedal signal at the kth moment, in V; is the prior estimate of the accelerator pedal signal at the kth moment, in V; k k is the Kalman gain; z k is the observed value of the accelerator pedal signal at the kth moment, that is, the digital accelerator pedal signal obtained after being processed by the controller at the kth moment, in V; p k is the accelerator pedal signal error variance at the kth moment, in V 2 ; is the priori accelerator pedal signal error variance at the kth moment, in V 2 ; r is the observed noise variance of the accelerator pedal signal, in V 2 ;

[0021] Among them, the Kalman gain k k It is calculated using the adaptive maximum correlation entropy Kalman filter method based on the Student t kernel, specifically including:

[0022] S3.1, kernel function selection;

[0023] The Student t kernel function is defined as:

[0024]

[0025] In formula 5, T β,σ (x,z) represents the Student t kernel function; x is the accelerator pedal signal, in V; z is the observed value of the accelerator pedal signal, in V; σ represents the kernel bandwidth, σ>0; β represents the degree of freedom, β>0;

[0026] S3.2, adaptive Kalman gain calculation;

[0027] The adaptive Kalman gain calculation formula of the maximum correlation entropy Kalman filter algorithm based on the Student's t kernel is:

[0028]

[0029] In formula 6, k k is the Kalman gain; is the priori accelerator pedal signal error variance at the kth moment, in V 2 ;z k is the observed value of the accelerator pedal signal at the kth moment, in V; is the prior estimate of the accelerator pedal signal at the kth moment, in V; β represents the degree of freedom, β>0; r is the observed noise variance of the accelerator pedal signal, in V 2 ; σ k is the kernel bandwidth at the kth moment, which is calculated using Formula 7:

[0030]

[0031] In formula 7, z k is the observed value of the accelerator pedal signal at the kth moment, in V; is the prior estimate of the accelerator pedal signal at the kth moment, in V; β represents the degree of freedom, β>0; σ c is the correction factor, σ c =1;

[0032] S4, filtering result output;

[0033] The estimated value of the accelerator pedal signal at the kth moment obtained by filtering Output, as the accelerator pedal signal value x at the kth moment k , for use in the electric drive system of rice transplanter.

[0034] In step S3, the degree of freedom β=1, the initial value of the accelerator pedal signal estimation value The initial value of the error variance of the accelerator pedal signal is p0 = 0.01V 2 , the process noise variance of the accelerator pedal signal q=0.001V 2 , the measurement noise variance of the accelerator pedal signal r=0.001V 2 .

[0035] The rice transplanter electronic accelerator pedal signal filtering method further comprises:

[0036] S5. Calculating the accelerator pedal opening according to the filtering result of the accelerator pedal signal, specifically comprising the following steps:

[0037] S5.1, limiting processing;

[0038] According to the range of the electronic accelerator pedal output voltage, set the upper limit sampling value x of the accelerator pedal signal max and the lower limit sampling value x min ; If the accelerator pedal signal x at the kth moment k Less than the lower limit sampling value x of the accelerator pedal signal min , then x k =x min ; If the accelerator pedal signal x at the kth moment k Greater than the upper limit sampling value x of the accelerator pedal signal max , then x k =x max ;

[0039] S5.2. Calculate the accelerator pedal opening;

[0040]

[0041] In formula 8, θ k is the accelerator pedal opening at the kth moment; x k is the accelerator pedal signal at the kth moment, in V; x max is the upper limit sampling value of the accelerator pedal signal, in V; x min It is the lower limit sampling value of the accelerator pedal signal, in V.

[0042] The rice transplanter electronic accelerator pedal signal filtering method further comprises:

[0043] S6, calculation of output torque of driving motor;

[0044] Based on the functional relationship between the accelerator pedal opening and the torque load coefficient, Formula 9, the output torque of the drive motor is calculated and determined by Formula 10;

[0045] L(θ)=aθ 2 +bθ Formula 9

[0046] Tgoal =L×T max (n) Formula 10

[0047] In formula 9 and formula 10, L(θ) represents the relationship function between the accelerator pedal opening and the torque load coefficient; θ is the accelerator pedal opening; a is the quadratic gain coefficient, b is the primary gain coefficient; L is the torque load coefficient; T goal is the output torque of the driving motor, in N·m; T max (n) is the maximum output torque corresponding to the drive motor speed n, in N·m.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] The rice transplanter electronic accelerator pedal signal filtering method of the present invention is based on an innovative adaptive maximum correlation entropy Kalman filtering method based on the Student's t kernel. Compared with the traditional maximum correlation entropy Kalman filtering method, it has stronger robustness to non-Gaussian disturbances in paddy fields and stronger tracking performance than the traditional maximum correlation entropy Kalman filtering algorithm based on the Student's t kernel. Specifically, in the simulation experiment of filtering the accelerator pedal signal under non-Gaussian noise, the mean square error of this method is 0.0241V. 2 Compared with an adaptive maximum correlation entropy Kalman filter algorithm based on a Gaussian kernel, the proposed method reduces the tracking performance by 43.2%, compared with a traditional Kalman filter algorithm, it reduces the tracking performance by 65.4%, and compared with a traditional maximum correlation entropy Kalman filter algorithm based on a Student's t kernel, which has poor tracking performance, it reduces the tracking performance by 93.5%. This method can reduce the impact of non-Gaussian disturbances in paddy fields on the accelerator pedal opening during driving, providing a basis for precise control of electric drive systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 This is the working process of the rice transplanter electronic accelerator pedal signal filtering method in each working cycle of the present invention;

[0051] Figure 2 The figure compares the recognition results of a group of accelerator pedal signals with non-Gaussian disturbances using four methods: an embodiment of the method of the present invention, an adaptive maximum correlation entropy Kalman filter algorithm based on Gaussian kernel (Comparative Example 1), a traditional Kalman filter algorithm (Comparative Example 2), and a traditional maximum correlation entropy Kalman filter algorithm based on Student's t kernel (Comparative Example 3). DETAILED DESCRIPTION

[0052] The present invention will be further described below with reference to the accompanying drawings and examples.

[0053] like Figure 1As shown, the present invention provides a rice transplanter electronic accelerator pedal signal filtering method, which includes the following steps in each working cycle:

[0054] S1, accelerator pedal signal acquisition and processing;

[0055] The controller of the rice transplanter drive system receives the analog accelerator pedal signal sent by the electronic accelerator pedal, and converts the analog accelerator pedal signal sent by the accelerator pedal into a digital accelerator pedal signal as the observed value z of the accelerator pedal signal at the kth moment. k ;

[0056] S2. Perform state prediction and variance prediction;

[0057] The state prediction process is:

[0058]

[0059] In formula 1, is the estimated value of the accelerator pedal signal at the k-1th moment, in V; is the prior estimate of the accelerator pedal signal at the kth moment, in V;

[0060] The variance prediction process is:

[0061]

[0062] In formula 2, p k-1 is the accelerator pedal signal error variance at the k-1th moment, in V 2 ; is the priori accelerator pedal signal error variance at the kth moment, in V 2 ; q is the process noise variance of the accelerator pedal signal, in V 2 ;

[0063] S3, perform status update and variance update;

[0064] The status update process is as follows:

[0065]

[0066] The variance update process is:

[0067]

[0068] In formula 3 and formula 4, is the estimated value of the accelerator pedal signal at the kth moment, in V; is the prior estimate of the accelerator pedal signal at the kth moment, in V; k k is the Kalman gain; z kis the observed value of the accelerator pedal signal at the kth moment, that is, the digital accelerator pedal signal obtained after being processed by the controller at the kth moment, in V; p k is the accelerator pedal signal error variance at the kth moment, in V 2 ; is the priori accelerator pedal signal error variance at the kth moment, in V 2 ; r is the observed noise variance of the accelerator pedal signal, in V 2 ;

[0069] Among them, the Kalman gain k k It is calculated using the adaptive maximum correlation entropy Kalman filter method based on the Student t kernel, specifically including:

[0070] S3.1, kernel function selection;

[0071] The kernel function in the maximum correlation entropy Kalman filter determines how noise is processed. Compared to the Gaussian kernel used in traditional maximum correlation entropy Kalman filters, the Student's t kernel has stronger non-Gaussian noise processing capabilities. Therefore, the maximum correlation entropy Kalman filter based on the Student's t kernel can better cope with non-Gaussian noise caused by paddy field working environments. The Student's t kernel function is defined as:

[0072]

[0073] In formula 5, T β,σ (x,z) represents the Student t kernel function; x is the accelerator pedal signal, in V; z is the observed value of the accelerator pedal signal, in V; σ represents the kernel bandwidth, σ>0; β represents the degree of freedom, β>0;

[0074] S3.2, adaptive Kalman gain calculation;

[0075] The adaptive Kalman gain calculation formula of the maximum correlation entropy Kalman filter algorithm based on the Student t kernel is:

[0076]

[0077] In formula 6, k k is the Kalman gain; is the priori accelerator pedal signal error variance at the kth moment, in V 2 ;z k is the observed value of the accelerator pedal signal at the kth moment, in V; is the prior estimate of the accelerator pedal signal at the kth moment, in V; β represents the degree of freedom, β>0; r is the observed noise variance of the accelerator pedal signal, in V 2 ; σ k is the kernel bandwidth at the kth moment, which is calculated using Formula 7:

[0078]

[0079] In formula 7, z k is the observed value of the accelerator pedal signal at the kth moment, in V; is the prior estimate of the accelerator pedal signal at the kth moment, in V; β represents the degree of freedom, β>0; σ c is the correction factor, σ c =1.

[0080] In traditional maximum correlation entropy Kalman filter algorithms, the kernel bandwidth is usually a fixed value. The present invention uses formula 7 to adaptively adjust the kernel bandwidth, which can further improve the tracking performance of the maximum correlation entropy Kalman filter based on the Student's t kernel. The derivation process of formula 7 is as follows:

[0081] The cost function of the maximum correlation entropy Kalman filter based on the Student's t kernel is usually defined as:

[0082]

[0083] Among them, x k is the accelerator pedal signal value at the kth moment, in V. After the filtering is completed, will be x k , which is used for the subsequent calculation of the accelerator pedal opening. In order to improve the sensitivity of the cost function to the observed state changes, and thus improve the tracking performance of the maximum correlation entropy Kalman filter based on the Student t kernel, the present invention adjusts the kernel bandwidth in the direction of increasing the sensitivity of the cost function to the observed state changes. To achieve this goal, first assume that Then let the cost function J(x k )right Derivation, we get Then, let Finally, the kernel bandwidth calculation formula at the kth moment is obtained In addition, it is easy to prove is a negative value, and yes However, it is worth noting that when When , the Kalman gain will become a constant. At this time, the filtering algorithm will no longer be based on the maximum correlation entropy algorithm. Therefore, the correction factor σ is introduced. c =1, and make This is used as the final kernel bandwidth adaptive adjustment formula. At this time, σ k 2 Still distance The minimum point of is close, which can ensure the sensitivity of the filtering algorithm to the changes in the observed state.

[0084] Preferably, in step S3, the degree of freedom β=1, the initial value of the accelerator pedal signal estimation value The initial value of the error variance of the accelerator pedal signal is p0 = 0.01V 2 , the process noise variance of the accelerator pedal signal q=0.001V 2 , the measurement noise variance of the accelerator pedal signal r=0.001V 2 .

[0085] S4, filtering result output;

[0086] The estimated value of the accelerator pedal signal at the kth moment obtained by filtering Output, as the accelerator pedal signal value x at the kth moment k , for use in the electric drive system of rice transplanter.

[0087] The electronic accelerator pedal signal obtained by filtering in the present invention can be used for subsequent calculation of the accelerator pedal opening. Therefore, the rice transplanter electronic accelerator pedal signal filtering method further includes step S5, calculating the accelerator pedal opening based on the filtering result of the accelerator pedal signal, specifically comprising the following steps:

[0088] S5.1, limiting processing;

[0089] In order to ensure that the accelerator pedal opening can be reliably set to zero and eliminate the instability of the accelerator pedal signal sampling value when it is close to the maximum and minimum values, the upper limit sampling value x of the accelerator pedal signal is set according to the range of the electronic accelerator pedal output voltage. max and the lower limit sampling value x min ; If the accelerator pedal signal x at the kth moment k Less than the lower limit sampling value x of the accelerator pedal signal min , then x k =x min ; If the accelerator pedal signal x at the kth moment k Greater than the upper limit sampling value x of the accelerator pedal signal max , then x k =x max ;

[0090] In the embodiment of the present invention, the output voltage range of the electronic accelerator pedal is 0V-3.3V, and the upper limit sampling value x of the accelerator pedal signal is set. max =3.1V and lower limit sampling value x min =0.17V; if x k <x min , then x k =x min =0.17, if x k >x max , then x k =x max=3.1.

[0091] S5.2. Calculate the accelerator pedal opening;

[0092] The calculation formula is as follows:

[0093]

[0094] In formula 8, θ k is the accelerator pedal opening at the kth moment; x k is the accelerator pedal signal at the kth moment, in V; x max is the upper limit sampling value of the accelerator pedal signal, in V; x min It is the lower limit sampling value of the accelerator pedal signal, in V.

[0095] The rice transplanter electronic accelerator pedal signal filtering method further includes step S6, driving motor output torque calculation;

[0096] Based on the functional relationship between the accelerator pedal opening and the torque load coefficient, Formula 9, the output torque of the drive motor is calculated and determined by Formula 10;

[0097] L(θ)=aθ 2 +bθ Formula 9

[0098] T goal =L×T max (n) Formula 10

[0099] In formula 9 and formula 10, L(θ) represents the relationship function between the accelerator pedal opening and the torque load coefficient; θ is the accelerator pedal opening; a is the quadratic gain coefficient, b is the primary gain coefficient; L is the torque load coefficient; T goal is the output torque of the driving motor, in N·m; T max (n) is the maximum output torque corresponding to the driving motor speed n, in N·m;

[0100] For Formula 9, as a and b change, its function curve always passes through the two points (0,0) and (1,1). The relationship between the accelerator pedal opening and the torque load coefficient can be divided into power mode (L(0.5)>0.5), general mode (L(0.5)=0.5), and economic mode (L(0.5)<0.5). The characteristic of the general mode is that the torque changes more evenly under different accelerator pedal openings, taking into account both power and economy. Therefore, in the embodiment of the present invention, the general mode is used to calculate the output torque of the drive motor, taking a=0 and b=1.

[0101] Example

[0102] This embodiment simulates and verifies the accuracy and superiority of the rice transplanter electronic accelerator pedal signal filtering method proposed in the present invention. The experiment was completed by matlab, and the simulation step length was 1s. During the experiment, four methods, namely, the embodiment of the method of the present invention, an adaptive maximum correlation entropy Kalman filter algorithm based on Gaussian kernel (Comparative Example 1), a traditional Kalman filter algorithm (Comparative Example 2), and a traditional maximum correlation entropy Kalman filter algorithm based on Student t kernel (Comparative Example 3), were used to identify a group of accelerator pedal signals with non-Gaussian disturbances. Among them, the initial value of the observed state is 1V, and the state at the remaining moments is randomly generated between 0V and 3.3V. At the same time, random non-Gaussian noise is applied to the system, wherein the process noise distribution is: Q~0.9N(0,Q0)+0.1N(0,50Q0), wherein Q0=0.001V 2 ; The measurement noise distribution is: R~0.9N(0,R0)+0.1N(0,100R0), where R0=0.001V 2 In addition, in order to verify the superiority of the present invention's method in tracking performance compared to the traditional maximum correlation entropy Kalman filter method based on the Student's t kernel, the initial values ​​of all filtering methods are set to 1.1V, which is larger than the initial value of the actual observed state. At the same time, the initial values ​​of the error variance, process noise variance, and observation noise variance in all comparative examples are consistent with those of the present invention. In comparative example 2, the kernel bandwidth adjustment method is where σ c = 1. Finally, the comparison of the recognition results of the four methods of the present invention, an adaptive maximum correlation entropy Kalman filter algorithm based on Gaussian kernel (Comparative Example 1), a traditional Kalman filter algorithm (Comparative Example 2), and a traditional maximum correlation entropy Kalman filter algorithm based on Student's t kernel (Comparative Example 3) for a set of accelerator pedal signals with non-Gaussian disturbances is shown in the figure below. Figure 2 The statistical results of mean square error are shown in Table 1. Among them, the mean square error of the filtering result of the method of the present invention is 0.0241V 2 Compared with an adaptive maximum correlation entropy Kalman filter algorithm based on a Gaussian kernel (Comparative Example 1), the accuracy is reduced by 43.2%, compared with a traditional Kalman filter algorithm (Comparative Example 2), the accuracy is reduced by 65.4%, and compared with a traditional maximum correlation entropy Kalman filter algorithm based on a Student's t kernel (Comparative Example 3), which has poor tracking performance, the accuracy is reduced by 93.5%. As can be seen from the experimental results, the method of the present invention achieves higher recognition accuracy than traditional methods under non-Gaussian perturbations, and solves the problem of low recognition accuracy caused by the poor tracking performance of the traditional maximum correlation entropy Kalman filter based on the Student's t kernel.

[0103] Table 1

[0104]

Claims

1. A rice transplanter electronic accelerator pedal signal filtering method, characterized in that: The rice transplanter electronic accelerator pedal signal filtering method comprises the following steps in each working cycle: S1, accelerator pedal signal acquisition and processing; The controller of the rice transplanter drive system receives the analog accelerator pedal signal sent by the electronic accelerator pedal, and converts the analog accelerator pedal signal sent by the accelerator pedal into a digital accelerator pedal signal as the observed value z of the accelerator pedal signal at the kth moment. k ; S2. Perform state prediction and variance prediction; The state prediction process is: In formula 1, is the estimated value of the accelerator pedal signal at the k-1th moment, in V; is the prior estimate of the accelerator pedal signal at the kth moment, in V; The variance prediction process is: In formula 2, p k-1 is the accelerator pedal signal error variance at the k-1th moment, in V 2 ; is the priori accelerator pedal signal error variance at the kth moment, in V 2 ; q is the process noise variance of the accelerator pedal signal, in V 2 ; S3, perform status update and variance update; The status update process is as follows: The variance update process is: p k =(1-k k )p k - (1-k k )+k k rk k Formula 4 In Formula 3 and Formula 4, is the estimated value of the accelerator pedal signal at the kth moment, in V; is the prior estimate of the accelerator pedal signal at the kth moment, in V; k k is the Kalman gain; z k is the observed value of the accelerator pedal signal at the kth moment, that is, the digital accelerator pedal signal obtained after being processed by the controller at the kth moment, in V; p k is the accelerator pedal signal error variance at the kth moment, in V 2 ; is the priori accelerator pedal signal error variance at the kth moment, in V 2 ; r is the observed noise variance of the accelerator pedal signal, in V 2 ; Among them, the Kalman gain k k It is calculated using the adaptive maximum correlation entropy Kalman filter method based on the Student t kernel, specifically including: S3.1, kernel function selection; The Student t kernel function is defined as: In formula 5, T β,σ (x,z) represents the Student t kernel function; x is the accelerator pedal signal, in V; z is the observed value of the accelerator pedal signal, in V; σ represents the kernel bandwidth, σ>0; β represents the degree of freedom, β>0; S3.2, adaptive Kalman gain calculation; The adaptive Kalman gain calculation formula of the maximum correlation entropy Kalman filter algorithm based on the Student's t kernel is: In formula 6, k k is the Kalman gain; is the priori accelerator pedal signal error variance at the kth moment, in V 2 ;z k is the observed value of the accelerator pedal signal at the kth moment, in V; is the prior estimate of the accelerator pedal signal at the kth moment, in V; β represents the degree of freedom, β>0; r is the observed noise variance of the accelerator pedal signal, in V 2 ; σ k is the kernel bandwidth at the kth moment, which is calculated using Formula 7: In formula 7, z k is the observed value of the accelerator pedal signal at the kth moment, in V; is the prior estimate of the accelerator pedal signal at the kth moment, in V; β represents the degree of freedom, β>0; σ c is the correction factor, σ c =1; S4, filtering result output; The estimated value of the accelerator pedal signal at the kth moment obtained by filtering Output, as the accelerator pedal signal value x at the kth moment k , for use in the electric drive system of rice transplanter.

2. The rice transplanter electronic accelerator pedal signal filtering method according to claim 1, characterized in that: In step S3, the degree of freedom β=1, the initial value of the accelerator pedal signal estimation value The initial value of the error variance of the accelerator pedal signal is p0 = 0.01V 2 , the process noise variance of the accelerator pedal signal q=0.001V 2 , the measurement noise variance of the accelerator pedal signal r=0.001V 2 .

3. The rice transplanter electronic accelerator pedal signal filtering method according to claim 1, characterized in that: The rice transplanter electronic accelerator pedal signal filtering method further comprises: S5. Calculating the accelerator pedal opening according to the filtering result of the accelerator pedal signal, specifically comprising the following steps: S5.1, limiting processing; According to the range of the electronic accelerator pedal output voltage, set the upper limit sampling value x of the accelerator pedal signal max and the lower limit sampling value x min ; If the accelerator pedal signal x at the kth moment k Less than the lower limit sampling value x of the accelerator pedal signal min , then x k =x min ; If the accelerator pedal signal x at the kth moment k Greater than the upper limit sampling value x of the accelerator pedal signal max , then x k =x max ; S5.

2. Calculate the accelerator pedal opening; In formula 8, θ k is the accelerator pedal opening at the kth moment; x k is the accelerator pedal signal at the kth moment, in V; x max is the upper limit sampling value of the accelerator pedal signal, in V; x min It is the lower limit sampling value of the accelerator pedal signal, in V.

4. The rice transplanter electronic accelerator pedal signal filtering method according to claim 3, characterized in that: The rice transplanter electronic accelerator pedal signal filtering method further comprises: S6, calculation of output torque of driving motor; Based on the functional relationship between the accelerator pedal opening and the torque load coefficient, Formula 9, the output torque of the drive motor is calculated and determined by Formula 10; L(θ)=aθ 2 +bθ formula 9T goal =L×T max (n) Formula 10 In Formula 9 and Formula 10, L(θ) represents the relationship function between the accelerator pedal opening and the torque load coefficient; θ is the accelerator pedal opening; a is the quadratic gain coefficient, b is the primary gain coefficient; L is the torque load coefficient; T goal is the output torque of the driving motor, in N·m; T max (n) is the maximum output torque corresponding to the drive motor speed n, in N·m.

Images