GNSS / INS fusion positioning device and method for harvester considering adaptive compensation of vibration error

Abstract

The present invention provides a GNSS / INS fusion positioning device and method for a harvester considering adaptive compensation of vibration errors, belonging to the field of intelligent agricultural machinery navigation and positioning technology. The fusion positioning method includes: a vibration sensor and GNSS / INS collect vibration information, positioning information, and orientation information of the harvester and transmit them to a host; the host establishes a vibration error compensation model under different working conditions; the host uses an RBF neural network to identify the current working condition according to the vibration characteristics, and then selects a corresponding vibration error compensation model to output the compensated GNSS data; the host uses a random forest algorithm to fuse the compensated GNSS data with the INS data; and the output fusion data provides positioning information for the harvester. The present invention can improve the positioning accuracy of the harvester by establishing a vibration error compensation model to compensate for the GNSS data, and fuse the compensated GNSS data with the INS data, thereby promoting the development of agricultural machinery navigation and positioning technology.

Description

Technical Field

[0001] The present invention belongs to the technical field of agricultural machinery navigation and positioning, and in particular relates to a GNSS / INS fusion positioning device and method for a harvester considering adaptive compensation of vibration errors. Background Art

[0002] With the acceleration of agricultural modernization, precision agriculture technology has gradually become the key to improving agricultural production efficiency and operation quality. In precision agriculture, the precise navigation and positioning of harvesters are the key to improving crop harvesting efficiency and reducing resource waste. The traditional harvester positioning system mainly relies on the combination of global navigation satellite system (GNSS) and inertial navigation system (INS). GNSS provides global coverage positioning services, but its signals are easily affected by the external environment, such as tree occlusion and building interference. Especially when the harvester is operating, due to the vibration and bumps of the machine itself, the reception quality of GNSS signals will further decrease, resulting in reduced positioning accuracy. INS can provide continuous navigation information, but in the absence of external input, its cumulative error will increase over time, affecting the long-term positioning accuracy. In addition, the vibration and impact generated by the harvester during operation will interfere with the INS sensor, further exacerbating the positioning error. At present, the GNSS / INS fusion positioning system of the harvester lacks consideration of vibration error, which will aggravate the positioning error of the GNSS / INS fusion positioning system and affect the operation of the harvester. Therefore, how to compensate for the impact of vibration errors on GNSS / INS positioning to improve the positioning accuracy of the fusion system has become an urgent problem to be solved. Summary of the invention

[0003] In view of the deficiencies in the prior art, the present invention provides a GNSS / INS fusion positioning device and method for a harvester taking into account adaptive compensation of vibration errors. By establishing a vibration error compensation model to compensate GNSS data, and fusing the compensated GNSS data with INS data, the positioning accuracy of the harvester is improved, thereby promoting the development of agricultural machinery navigation and positioning technology.

[0004] The present invention achieves the above technical objectives through the following technical means.

[0005] A GNSS / INS fusion positioning method for a harvester considering adaptive compensation of vibration errors includes the following processes:

[0006] Step 1: The host controls the vibration sensor and the GNSS / INS fusion positioning system to start, collect the harvester's vibration information, positioning information, and orientation information, and transmit them to the host;

[0007] Step 2: The host establishes a vibration error compensation model for the harvester under three working conditions: idling, transporting, and harvesting;

[0008] Step 3: The host uses the RBF model to identify the current working condition based on the vibration characteristics, then selects the corresponding vibration error compensation model and outputs the compensated GNSS data;

[0009] Step 4: The host uses the random forest algorithm to fuse the compensated GNSS data with the INS data;

[0010] Step 5: Output the fused data to provide positioning information for the harvester.

[0011] Furthermore, in step 2, when the harvester is in an idling state, the vibration error has a linear characteristic:

[0012]

[0013] The compensation formula is:

[0014]

[0015] in, represents linear error; Indicates that the harvester is in idling state; is the proportionality factor related to the vibration amplitude; is a measurement related to vibration; It is the GNSS measurement value when the harvester is idling; It is the value after GNSS compensation when the harvester is idling.

[0016] Furthermore, in step 2, when the harvester is in the grain transport state, the vibration error is approximately a linear error, but is affected by the vibration frequency:

[0017]

[0018] The compensation formula is:

[0019]

[0020] in, is the error related to the vibration frequency, Indicates that the harvester is in transport status; is the vibration frequency Function of is a measurement related to vibration; It is the GNSS measurement value when the harvester transports grain; It is the value after GNSS compensation when the harvester transports grain.

[0021] Furthermore, in step 2, when the harvester is in the grain harvesting state, the vibration error is a nonlinear error:

[0022]

[0023] The compensation formula is:

[0024]

[0025] in, is the nonlinear vibration error, Indicates that the harvester is in harvesting state; is a measurement related to vibration; , , are model parameters; It is the GNSS measurement value when the harvester is harvesting grain; It is the value after GNSS compensation when the harvester is harvesting grain.

[0026] Furthermore, in step 3, the specific method for the host to identify the current working condition using the RBF model according to the vibration characteristics is as follows:

[0027] Step 3.1: First, the vibration sensor collects the vibration frequency, amplitude, and vibration acceleration of the harvester under different working conditions as sample data and sends it to the host. The host preprocesses the received data and extracts the vibration acceleration. ,amplitude ;

[0028] Step 3.2: Host vibration acceleration Integrate to get the vibration speed , and then get the vibration eigenvector , , is the vibration sensor working time;

[0029] Step 3.3: Manually classify the vibration conditions of the harvester according to the vibration feature vector and present the classification results and the vibration eigenvector The RBF training model is input together, and the center vector, width parameter, and weight of the hidden layer nodes in the RBF model are obtained by using the self-organizing center selection learning method;

[0030] Step 3.4: The host constructs the RBF model based on the results of step 3.3;

[0031] Step 3.5: The host inputs the vibration feature vector in the RBF model input layer to obtain the radial basis function output value ;

[0032] Step 3.6: In the output layer For further processing:

[0033]

[0034] in, It is The weights of the hidden layer nodes; is the output value of the output layer; ( =1, 2, 3, …, ) is a hidden layer node; is the number of hidden layer nodes;

[0035] Step 3.7: Normalize to get the probability of each category:

[0036]

[0037] in, is the output layer corresponding to The output value of the hidden layer nodes; is the softmax function;

[0038] Compare The probability value of each category in is used to determine the working condition category to which the input vibration feature vector belongs.

[0039] Furthermore, the specific method of step 4 is as follows:

[0040] Step 4.1: First, use linear interpolation to align the timestamps of the INS data and the compensated GNSS data:

[0041]

[0042]

[0043]

[0044] in, is the GNSS timestamp; is the INS timestamp; is the slope, which indicates the average increase in the GNSS timestamp for each unit increase in the INS timestamp; is the intercept, which indicates the initial offset of the GNSS timestamp when the INS timestamp is 0; Corresponding GNSS timestamp ;INS timestamp Corresponding GNSS timestamp ;

[0045] Step 4.2: Extract INS and compensated GNSS feature data and construct random forest input features :

[0046]

[0047] in, , , It is the characteristic data of GNSS and INS;

[0048] Step 4.3: Build a random forest model:

[0049]

[0050]

[0051] in, Based on features The prediction function of is the number of trees in the random forest; It is a tree prediction; It is Prediction of a tree; yes The voting result of all trees;

[0052] Step 4.4: GNSS / INS fusion positioning system and vibration sensor collect sample data to train the random forest model;

[0053] Step 4.5: Use mean square error (MSE) to evaluate the performance of the trained random forest model.

[0054] Step 4.6: Input the constructed random forest into the features As a feature Substitute the qualified random forest model and solve it to obtain the fused data.

[0055] A fusion positioning device for implementing the above-mentioned harvester GNSS / INS fusion positioning method considering vibration error adaptive compensation is fixed to the agricultural machine by welding or bolt and nut connection, and includes a control module, a data transmission module, a sensor, and a support structure;

[0056] The supporting structure includes an equipment box centrally mounted on the aluminum profile frame, a cover of the equipment box is hinged on the top of the equipment box, and a hole is opened on the side wall of the equipment box as a line outlet of the data transmission module; two mushroom-head antenna supports are respectively mounted at both ends of the aluminum profile frame for mounting the data transmission module;

[0057] The sensors include a vibration sensor installed on the cover of the equipment box and a GNSS / INS fusion positioning system installed in the corner of the equipment box; the vibration sensor is used to obtain the vibration information of the harvester, and the GNSS / INS fusion positioning system is used to obtain the positioning information and orientation information of the harvester;

[0058] The control module includes a host installed above the GNSS / INS fusion positioning system and supported by a boss inside the equipment box; the data transmission module includes a 4G communication module installed on the side wall of the equipment box, on which a 4G module antenna is installed, and also includes a mushroom head antenna installed on a mushroom head antenna support.

[0059] The present invention has the following beneficial effects:

[0060] (1) For the first time, a GNSS / INS fusion positioning device and method for harvesters considering adaptive compensation of vibration errors was proposed. By establishing a vibration error compensation model and fusing GNSS / INS data, the navigation accuracy of agricultural machinery in the field was significantly improved.

[0061] (2) By integrating the RBF algorithm and the softmax function, the working state of agricultural machinery can be identified in real time, and the vibration error compensation model can be adaptively changed according to the working condition of the agricultural machinery, thus avoiding the problem of mismatch between the vibration working condition and the vibration error compensation model.

[0062] (3) A vibration error compensation model for agricultural machinery under idling, transportation, and harvesting conditions was established to compensate for the impact of vibration on navigation accuracy during operation of agricultural machinery and to solve the problem of excessive vibration influence on navigation and positioning of traditional harvesters. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 The overall structural diagram of the harvester GNSS / INS fusion positioning device installed on the harvester considering the adaptive compensation of vibration errors;

[0064] Figure 2 The overall structural diagram of the harvester GNSS / INS fusion positioning device considering the adaptive compensation of vibration errors;

[0065] Figure 3 This is the flow chart of the GNSS / INS fusion positioning method for harvesters considering adaptive compensation of vibration errors;

[0066] Figure 4 Flowchart for vibration condition identification.

[0067] In the figure: 1-vibration sensor; 2-device box cover; 3-4G module antenna; 4-4G communication module; 5-host; 6-GNSS / INS fusion positioning system; 7-device box; 8-mushroom head antenna; 9-mushroom head antenna support; 10-built-in corner groove connector; 11-aluminum profile frame. DETAILED DESCRIPTION

[0068] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments, but the scope of protection of the present invention is not limited thereto. It should be noted that: unless otherwise defined, the technical terms or scientific terms used herein should be the usual meanings understood by people with ordinary skills in the field to which the present invention belongs. In the description used in the patent application specification and claims of the present invention, it should be emphasized that, unless otherwise clearly specified and limited, words such as "one", "an" or "the" do not indicate a quantitative limitation, but indicate that there is at least one instance. Similarly, similar words such as "include" or "comprise" indicate the listed elements and possible other elements; elements or objects include the elements or objects listed after "include" or "comprise" and their equivalents, but do not exclude other elements or objects with the same function.

[0069] like Figure 1 As shown, the harvester GNSS / INS fusion positioning device taking into account the adaptive compensation of vibration errors described in the present invention can be installed near the cockpit of the field agricultural machinery so as to connect the line to the cockpit. The fusion positioning device can be fixed to the agricultural machinery by welding or bolt and nut connection; the fusion positioning device includes a control module, a data transmission module, a sensor, and a support structure.

[0070] like Figure 2 As shown, the supporting structure includes an aluminum profile frame 11, a device box 7, a device box cover 2, a mushroom head antenna support 9, and a built-in angle slot connector 10; the device box 7 is centrally installed on the aluminum profile frame 11, and the device box 7 and the aluminum profile frame 11 are connected and fastened by the built-in angle slot connector 10; the top of the device box 7 is connected to the device box cover 2 by a hinge; a plurality of holes are opened on the side wall of the device box 7 as line outlets of subsequent data transmission modules; the mushroom head antenna support 9 is made of iron, and two mushroom head antenna supports 9 are respectively installed at both ends of the aluminum profile frame 11 for subsequent installation of the data transmission module, and at the same time can also improve the stability of the fusion positioning device.

[0071] like Figure 2 As shown, the sensor includes a vibration sensor 1 (vibrationsensor, VS) installed on the cover 2 of the equipment box and a GNSS / INS fusion positioning system 6 installed in the corner of the equipment box 7; the vibration sensor 1 is used to obtain the vibration information of the harvester, and the GNSS / INS fusion positioning system 6 is used to obtain the positioning information and orientation information (heading angle, pitch angle and roll angle) of the harvester.

[0072] like Figure 2 As shown, the control module includes a host 5, which is installed above the GNSS / INS fusion positioning system 6 and supported by the internal boss of the equipment box 7.

[0073] like Figure 2As shown, the data transmission module includes a 4G communication module 4 installed on the side wall of the device box 7, on which a 4G module antenna 3 is installed, and also includes two mushroom head antennas 8 installed on a mushroom head antenna support 9. The mushroom head antenna 8 includes an antenna body, a screw, and a magnetic base. The middle part of the mushroom head antenna support 9 is provided with a round hole that matches the screw thread. The mushroom head antenna 8 can be adsorbed on the mushroom head antenna support 9 through the magnetic base, or the screw can cooperate with the round hole in the middle part of the mushroom head antenna support 9 to play the role of fixing the mushroom head antenna 8; the aluminum profile frame 11 can be fixed with a cable tie.

[0074] Reference Figure 3 , 4 As shown, the GNSS / INS fusion positioning method for a harvester considering vibration error adaptive compensation using the above-mentioned harvester GNSS / INS fusion positioning device considering vibration error adaptive compensation includes the following steps:

[0075] Step 1: Start the fusion positioning device. Under the control of the host 5, start the vibration sensor 1 and the GNSS / INS fusion positioning system 6 respectively to collect the harvester vibration information, positioning information and orientation information, and transmit them to the host 5.

[0076] Step 2: The host 5 establishes a vibration error compensation model under different working conditions, as follows:

[0077] When the harvester is in the three working conditions of idling, transporting and harvesting, the error types caused by vibration are different;

[0078] When the harvester is in idling state, the vibration error has obvious linear characteristics. Therefore, a linear model is used to describe the error:

[0079]

[0080] The compensation formula is:

[0081]

[0082] in, represents linear error; Indicates that the harvester is in idling state; is the proportionality factor related to the vibration amplitude; is a measurement related to vibration; It is the GNSS measurement value when the harvester is idling; It is the value after GNSS compensation when the harvester is idling.

[0083] When the harvester is in the state of transporting grain, the vibration error is approximately a linear error, but is affected by the vibration frequency. Therefore, a frequency-based error model is used to describe the error:

[0084]

[0085] The compensation formula is:

[0086]

[0087] in, is the error related to the vibration frequency, Indicates that the harvester is in transport status; is the vibration frequency Function of is a measurement related to vibration; It is the GNSS measurement value when the harvester transports grain; It is the value after GNSS compensation when the harvester transports grain.

[0088] When the harvester is in the state of harvesting grain, the positioning error type affected by vibration is a nonlinear error. Therefore, a nonlinear error model is used to describe the error:

[0089]

[0090] The compensation formula is:

[0091]

[0092] in, is the nonlinear vibration error, Indicates that the harvester is in harvesting state; is a measurement related to vibration; , , are model parameters; It is the GNSS measurement value when the harvester is harvesting grain; It is the value after GNSS compensation when the harvester is harvesting grain.

[0093] Step 3: The host 5 uses a radial basis function (RBF) neural network to identify the current working condition according to the vibration characteristics, and then selects a corresponding vibration error compensation model to output the compensated GNSS data;

[0094] Among them, refer to Figure 4 ,The specific method of using the classification algorithm to identify the current working condition based on the vibration characteristics is as follows:

[0095] Step 3.1: First, the vibration sensor 1 collects the vibration frequency, amplitude, and vibration acceleration of the harvester under different working conditions as sample data and sends it to the host 5. The host 5 pre-processes the received data and extracts the vibration acceleration. ,amplitude .

[0096] Step 3.2: Host 5 for vibration acceleration Integrate to get the vibration speed , and then get the vibration eigenvector ;

[0097]

[0098] in, is the working time of vibration sensor 1.

[0099] Step 3.3: Manually classify the vibration conditions of the harvester according to the vibration feature vector and present the classification results and the vibration eigenvector The RBF training model is input together, and the self-organizing center selection learning method is used to solve the center vector, width parameter, and weight of the hidden layer nodes in the radial basis function.

[0100] Step 3.3.1: Use K-means algorithm to find the center vector of hidden layer nodes;

[0101] Host 5 according to the manual classification results The vibration eigenvector of the sample data Divided into Group, from this Extract a vibration feature vector from each group of data As cluster center ;

[0102] For each vibration eigenvector , calculate its Euclidean distance to each cluster center:

[0103]

[0104] in, ( =1, 2, 3, …, ) is a hidden layer node; is the number of hidden layer nodes; It is the first Vibration eigenvectors ( =1,2,…); It is Cluster centers; is the Euclidean distance between the vibration eigenvector and the cluster center;

[0105] The vibration eigenvector Assign it to the nearest cluster center and recalculate the latest cluster center:

[0106]

[0107] in, is assigned to The vibration eigenvector of is assigned to The number of vibration eigenvectors;

[0108] Repeatedly assign vibration feature vectors, calculate the latest cluster center, and obtain the center vector of the hidden layer node .

[0109] Step 3.3.2: Calculate the width parameter of the hidden layer node by the following formula:

[0110]

[0111] in, is the maximum distance between selected cluster centers; is the width parameter of the hidden layer nodes.

[0112] Step 3.3.3: Use the loss function to solve the weights of the hidden layer nodes:

[0113]

[0114] in, is the weight of the hidden layer nodes.

[0115] Step 3.4: Host 5 builds the RBF model based on the results of step 3.3, which mainly includes the input layer, hidden layer, and output layer:

[0116]

[0117] in, Output value for radial basis function; It is The center vector of each hidden layer node.

[0118] Step 3.5: The vibration sensor 1 collects the harvester vibration frequency, amplitude, and vibration acceleration and sends them to the host 5. The host 5 solves the characteristic vector through steps 3.1 and 3.2. , input the vibration feature vector into the RBF model input layer , and the solution is .

[0119] Step 3.6: In the output layer For further processing:

[0120]

[0121] in, It is The weights of the hidden layer nodes; is the output value of the output layer.

[0122] Step 3.7: Output value for the output layer , use the softmax function to normalize to get the probability of each category:

[0123]

[0124] in, is the output layer corresponding to The output value of the hidden layer nodes; is the softmax function;

[0125] By comparison The probability value of each category in the input vibration feature vector can be determined The operating condition category.

[0126] Step 4: The host 5 uses the random forest algorithm to fuse the compensated GNSS data and INS data. The specific method is as follows:

[0127] Step 4.1: First, use linear interpolation to align the timestamps of the INS data and the compensated GNSS data:

[0128]

[0129]

[0130]

[0131] in, is the GNSS timestamp; is the INS timestamp; is the slope, which indicates the average increase in the GNSS timestamp for each unit increase in the INS timestamp; is the intercept, which indicates the initial offset of the GNSS timestamp when the INS timestamp is 0; ( , ): INS timestamp Corresponding GNSS timestamp ; ( , ): INS timestamp Corresponding GNSS timestamp .

[0132] Step 4.2: Extract INS and compensated GNSS feature data and construct random forest input features :

[0133]

[0134] in, , , It is the characteristic data of GNSS and INS.

[0135] Step 4.3: Build a random forest model:

[0136]

[0137]

[0138] in, Based on features The prediction function of is the number of trees in the random forest; It is a tree prediction; It is Prediction of a tree; yes The result of the common vote of the trees.

[0139] Step 4.4: The GNSS / INS fusion positioning system 6 and the vibration sensor 1 collect sample data to train the random forest model.

[0140] Step 4.5: Evaluate the performance of the trained random forest model using Mean Squared Error (MSE):

[0141]

[0142] in, is the true value, is the predicted value, is the sample size.

[0143] Step 4.6: Use the qualified random forest model to fuse INS data and compensated GNSS data: Input the constructed random forest into the feature As a feature Substitute the qualified random forest model and solve it to obtain the fused data.

[0144] Step 5: Output the fused data to provide positioning information for the harvester.

[0145] The embodiments are preferred implementations of the present invention, but the present invention is not limited to the above-mentioned implementations. Any obvious improvements, substitutions or modifications that can be made by those skilled in the art without departing from the essential content of the present invention belong to the protection scope of the present invention.

[0146] As described above, although the present invention has been shown and described with reference to specific preferred embodiments, it should not be construed as limiting the present invention itself. Various changes in form and details may be made without departing from the spirit and scope of the present invention as defined in the appended claims.

Claims

1. A GNSS / INS fusion positioning method for harvesters considering adaptive compensation of vibration errors, characterized in that: The process includes the following: Step 1: The host (5) controls the vibration sensor (1) and the GNSS / INS fusion positioning system (6) to start, collect the harvester vibration information, positioning information and orientation information and transmit them to the host (5); Step 2: The host (5) establishes a vibration error compensation model of the harvester under three working conditions: idling, transporting, and harvesting; Step 3: The host (5) uses the RBF model to identify the current working condition according to the vibration characteristics, then selects the corresponding vibration error compensation model and outputs the compensated GNSS data; Step 4: The host (5) uses the random forest algorithm to fuse the compensated GNSS data with the INS data; Step 5: Output fusion data to provide positioning information for the harvester; In step 2, when the harvester is in an idling state, the vibration error has a linear characteristic: e I =k I x The compensation formula is: and I 'and I -and I Among them, e I Indicates linear error; I indicates that the harvester is in idle state; k I is the proportionality factor related to the vibration amplitude; x is the measured value related to the vibration; y I is the GNSS measurement value when the harvester is idling; I ′ is the value after GNSS compensation when the harvester is idling; In step 2, when the harvester is in the state of transporting grain, the vibration error is approximately a linear error, but is affected by the vibration frequency: e T =H(f)*x The compensation formula is: and' T =and T -and T Among them, e T is the error related to the vibration frequency, T indicates that the harvester is in the transport state; H(f) is a function of the vibration frequency f; y T is the GNSS measurement value when the harvester transports grain; y′ T It is the value after GNSS compensation when the harvester transports grain; In step 2, when the harvester is in the grain harvesting state, the vibration error is a nonlinear error: e H =Ax 2 +Bx+C The compensation formula is: and' H =and H -and H Among them, e H is the nonlinear vibration error, H indicates that the harvester is in the harvesting state; A, B, C are model parameters; y H is the GNSS measurement value when the harvester is harvesting grain; y′ H It is the value after GNSS compensation when the harvester is harvesting grain.

2. The harvester GNSS / INS fusion positioning method considering vibration error adaptive compensation according to claim 1 is characterized in that: In step 3, the specific method of the host (5) using the RBF model to identify the current working condition according to the vibration characteristics is as follows: Step 3.1: First, the vibration sensor (1) collects the vibration frequency, amplitude, and vibration acceleration of the harvester under different working conditions as sample data and sends it to the host (5). The host (5) pre-processes the received data and extracts the vibration acceleration a and amplitude A; Step 3.2: The host (5) integrates the vibration acceleration a to obtain the vibration velocity v, and then obtains the vibration characteristic vector t is the working time of the vibration sensor (1); Step 3.3: Manually classify the vibration working conditions of the harvester according to the vibration feature vector, input the classification results and the vibration feature vector into the RBF training model, and use the self-organizing center selection learning method to obtain the center vector, width parameter, and weight of the hidden layer node in the RBF model; Step 3.4: The host (5) constructs the RBF model according to the result of step 3.3; Step 3.5: The host (5) inputs the vibration feature vector into the RBF model input layer to obtain the radial basis function output value h j ; Step 3.6: In the output layer, h j For further processing: Among them, ω j is the weight of the jth hidden layer node; y RBF is the output value of the output layer; j is the hidden layer node, j = 1, 2, 3, ..., q; q is the number of hidden layer nodes; Step 3.7: y RBF Normalize to get the probability of each category: Among them, y j is the output value corresponding to the jth hidden layer node in the output layer; is the softmax function; Compare The probability value of each category in is used to determine the working condition category to which the input vibration feature vector belongs.

3. The harvester GNSS / INS fusion positioning method considering vibration error adaptive compensation according to claim 1 is characterized in that: The specific method of step 4 is as follows: Step 4.1: First, use linear interpolation to align the timestamps of the INS data and the compensated GNSS data: y tGNSS =mx tINS +b b=y tGNSS1 -mx tINS1 Among them, y tGNSS is the GNSS timestamp; x tINS is the INS timestamp; m is the slope; b is the intercept; INS timestamp x tINS1 Corresponding GNSS timestamp y tGNSS1 ;INS timestamp x tINS2 Corresponding GNSS timestamp y tGNSS2 ; Step 4.2: Extract INS and compensated GNSS feature data and construct random forest input feature X: X=[X1,X2,…,X n ] Among them, X1, X2, …, X n It is the characteristic data of GNSS and INS; Step 4.3: Build the Random Forest Model: T(x RF )=f(x RF ) Among them, f(x RF ) is based on the feature x RF prediction function; N is the number of trees in the random forest; T(x RF ) is the prediction of a tree; T i (x RF ) is the prediction of the i-th tree; is the common voting result of i trees; Step 4.4: The GNSS / INS fusion positioning system (6) and the vibration sensor (1) collect sample data to train the random forest model; Step 4.5: Use mean square error (MSE) to evaluate the performance of the trained random forest model. Step 4.6: Input the constructed random forest feature X as feature x RF Substitute the qualified random forest model into the solution to obtain the fused data.

4. A fusion positioning device for implementing the GNSS / INS fusion positioning method for a harvester considering adaptive compensation of vibration errors as described in any one of claims 1 to 3, fixed to the agricultural machine by welding or bolt and nut connection, and connected to the cockpit of the agricultural machine by lines, characterized in that: It includes support structure, data transmission module, sensor and control module; The supporting structure comprises a device box (7) centrally mounted on an aluminum profile frame (11); a device box cover (2) is hingedly connected to the top of the device box (7); and a side wall opening of the device box (7) serves as a line outlet of a data transmission module; two mushroom-head antenna supports (9) are respectively mounted at two ends of the aluminum profile frame (11) for mounting the data transmission module; the data transmission module comprises a 4G communication module (4) mounted on the side wall of the device box (7), a 4G module antenna (3) being mounted thereon, and a mushroom-head antenna (8) mounted on the mushroom-head antenna support (9); The sensor comprises a vibration sensor (1) installed on a cover (2) of an equipment box and a GNSS / INS fusion positioning system (6) installed at a corner of the equipment box (7); the vibration sensor (1) is used to obtain vibration information of the harvester, and the GNSS / INS fusion positioning system (6) is used to obtain positioning information and orientation information of the harvester; the control module comprises a host (5) installed above the GNSS / INS fusion positioning system (6) and supported by a boss inside the equipment box (7).

Images