Shield tunneling machine automatic driving method and system combining deep learning and PID control
By combining deep learning and PID control methods, the problem of large control boundary error in shield machine automatic driving technology is solved, and the autonomous excavation of shield machine is realized, and the excavation speed, efficiency and accuracy are improved.
Patent Information
- Application Number
- CN202411393918.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-08
- Publication Date
- 2024-12-06
Smart Images

Figure CN119087787A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of autonomous driving technology, and in particular to an autonomous driving method and system for a shield machine excavating along a designed axis by combining deep learning and PID control. Background Art
[0002] There are many studies on autonomous driving technology in the fields of automobile, high-speed rail, subway, ship driving, etc. Among them, the automobile autonomous driving technology is to enable the car to drive and operate without human intervention through computer systems and sensors. Relying on the perception system to obtain information about the surrounding environment, it is necessary to detect the existence of other vehicles, pedestrians and other related entities. High-speed rail or subway autonomous driving relies on high-speed rail track facilities and uses track detectors and position reference information to achieve autonomous driving. Ship autonomous driving technology relies on global satellite navigation systems, radar and laser rangefinders, cameras and image processing technology, and communication technology. However, the autonomous driving technology for shield machines is not perfect at present.
[0003] The working environment of the shield machine is deep underground, and some traditional tools for obtaining external information (such as cameras, GPS positioning systems, radars, etc.) cannot be used. The method adopted by the prior art is to combine the machine learning, deep learning and other technical training models in the field of artificial intelligence to control the posture of the shield machine. However, this method of the prior art does not meet the technical requirements of the shield machine's automatic driving forward excavation, and its boundary error is relatively large. If the shield machine excavation control fails, it is easy to cause the shield machine to fall, kowtow, surface settlement, landslides, tunnel floating, serpentine and other problems. Therefore, a reliable shield machine automatic driving model drives the shield machine to automatically dig forward, which is of great significance to improving the excavation speed, efficiency and excavation accuracy. Summary of the invention
[0004] In order to solve the above-mentioned problems, the present invention aims to provide an automatic driving method, system and storage medium for a shield machine to excavate along a designed axis by combining deep learning and PID control.
[0005] To achieve the above object, the present invention adopts the following technical solutions:
[0006] The present invention provides a shield machine automatic driving method combining deep learning and PID control, which is characterized by comprising the following steps:
[0007] Step S1: Obtaining parameters of the shield machine;
[0008] Step S2: fuzzy process all the parameters obtained in step S1 to obtain a set of new data;
[0009] Step S3: Process the data obtained in step S2 as a training set, and divide the data into input data and labels;
[0010] Step S4: input the data obtained in step S3 into the neural network model for iterative training;
[0011] Step S5: Use the output result of the model trained in step S4 as the test set data to determine whether the test set data meets the iteration number requirement. If not, feed the test set data back to the training set data in step S3, and then perform the operation of step S4 again until the training is terminated after the iteration number is reached to obtain a trained model.
[0012] Step S6: Using the trained model obtained in step S5 as a prediction model, and outputting a prediction result of the prediction model;
[0013] Step S7: input the prediction result obtained in step S6 into the PID algorithm controller as a process variable of PID control;
[0014] Step S8: Use PID algorithm controller to adjust the proportional parameter K online in real time P , integration parameter K I , differential parameter K D Make it reach the optimal value, and then use the optimal value to ultimately control the excavation work of the shield machine.
[0015] Furthermore, in the shield machine automatic driving method combining deep learning and PID control provided by the present invention, it can also have the following characteristics: wherein, the parameters acquired in step S1 include the shield machine's own parameters, external environment parameters, and latitude and longitude parameters.
[0016] Furthermore, in the shield machine automatic driving method combining deep learning and PID control provided by the present invention, there may also be such a feature: wherein the specific process of step S3 is as follows:
[0017] Step S3-1: first convert the data matrix;
[0018] Step S3-2: filter out the wrong data and keep the correct data;
[0019] Step S3-3: The current data of the five directional positions and the latitude and longitude parameters are used as labels, and the remaining parameters are used as input data.
[0020] Furthermore, in the shield machine automatic driving method combining deep learning and PID control provided by the present invention, it can also have the following characteristics: wherein, step S4 takes the first 80% of the data obtained in step S3 as a training set, and the last 20% of the data as a verification set, and the training set is input into the MLP neural network model for iterative training, the input layer sets the 24 input data as 24 neurons, maps them to the hidden layer, and uses the ReLU activation function, and finally the output layer reduces the data to 8 neurons, namely, the current data of 5 directional positions and 3 longitude and latitude parameters.
[0021] Furthermore, in the shield machine automatic driving method combining deep learning and PID control provided by the present invention, it can also have the following characteristics: wherein, the MLP neural network model is provided with 3 hidden layers, the first layer has 64 neurons, the second layer has 128 neurons, and the third layer has 64 neurons.
[0022] Furthermore, in the shield machine automatic driving method combining deep learning and PID control provided by the present invention, it can also have the following characteristics: wherein, the MLP neural network model is provided with 3 hidden layers, the first layer has 128 neurons, the second layer has 128 neurons, and the third layer has 128 neurons.
[0023] Furthermore, in the shield machine automatic driving method combining deep learning and PID control provided by the present invention, it can also have the following characteristics: wherein, the prediction result obtained in step S6 is 5 current prediction values and 3 longitude and latitude high prediction values, and the 5 current prediction values and 3 longitude and latitude high prediction values are used as process variables of PID control, and the expected output of the PID algorithm controller is the thrust value, which is the target value to be adjusted for the shield machine.
[0024] Furthermore, in the shield machine automatic driving method combining deep learning and PID control provided by the present invention, it can also have the following characteristics: wherein step S8 adjusts the proportional parameter K online in real time P , integration parameter K I , differential parameter K D The specific operation is: First, initialize the parameter K P , K I , K D , use the real-time collected data to calculate the system error, which is the input of the PID algorithm controller, and then the new control output is calculated through the PID algorithm.
[0025] The present invention also provides a shield machine automatic driving system, which is used in the above-mentioned shield machine automatic driving method combining deep learning and PID control, and is characterized in that it includes: a parameter acquisition module, used to obtain the parameters of the shield machine; an MLP module, which has a preset trained MLP neural network model, takes the 24 parameters of the shield machine as input data, and outputs 5 current prediction values and 3 longitude and latitude high prediction values; a PID module, which runs a PID control algorithm, and uses the 5 current prediction values and 3 longitude and latitude high prediction values obtained by the MLP module as process variables for PID control, and the expected output of the PID algorithm controller is the thrust value; a result display module, which is used to display the calculation results of the PID module; and a shield machine execution control module, which controls the shield machine according to the calculation results of the PID module.
[0026] The present invention also provides a computer-readable storage medium on which computer program instructions are stored. When the computer program instructions are executed by a processor, the above-mentioned shield machine automatic driving method combining deep learning and PID control is implemented.
[0027] Function and effect of the present invention:
[0028] The present invention combines deep learning and PID control of the shield machine automatic driving method, through the shield machine's own parameters, external parameters, geological data and the driver's operation of the shield machine, these parameters are substituted into different soil conditions for training and learning, giving full play to the prediction method of MLP deep learning, maintaining the robustness, real-time and interpretability of the algorithm, and improving the accuracy by combining the PID control algorithm, so as to realize the shield machine in the case of unmanned operation through the trained model can autonomously forward dig tunnel behavior. The algorithm has a high accuracy rate and a low shield boundary error, and it is concluded that the combination of MLP and PID control algorithms can accurately dig under different soil conditions. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] Figure 1 It is a diagram of different construction soil layer types;
[0030] Figure 2 It is a flow chart of a construction model of a shield machine automatic driving method combining deep learning and PID control according to the present invention;
[0031] Figure 3 This is a PID calculation process diagram of the shield machine automatic driving method combining deep learning and PID control in the present invention. DETAILED DESCRIPTION
[0032] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the following embodiments are combined with the accompanying drawings to specifically illustrate the technical solutions of the present invention.
[0033] See also Figure 2 The present invention provides a shield machine automatic driving method combining deep learning and PID control, comprising the following steps:
[0034] Step S1: Obtain the parameters of the shield machine, which are 32 in total, including the shield machine's own parameters, external environment parameters, and latitude and longitude parameters.
[0035] There are 20 parameters of the shield machine, namely: cutter speed, cutter torque, upper push jack pressure, right push jack pressure, lower push jack pressure, left push jack pressure, advancement speed, upper push jack stroke, right push jack stroke, lower push jack stroke, left push jack stroke, upper joint jack stroke, right joint jack stroke, lower joint jack stroke, left joint jack stroke, upper current, right current, lower current, left current, and middle current.
[0036] There are 9 external environmental parameters of the shield machine, namely: upper soil pressure, right soil pressure, lower soil pressure, left soil pressure, upper grouting pressure, right grouting pressure, lower grouting pressure, left grouting pressure, and grouting volume.
[0037] There are three latitude, longitude and altitude parameters of the shield machine: longitude, latitude and altitude.
[0038] Step S2: Fuzzy process all the parameters obtained in step S1 to obtain a set of new data.
[0039] Step S3: Process the data obtained in step S2 and divide the data into input data and labels. The specific process is as follows:
[0040] S3-1: First, convert the data matrix to obtain matrix data that can be processed by the computer;
[0041] S3-2: Filter out the wrong data and keep the correct data;
[0042] S3-3: The current data of the five directional positions (upper current, right current, lower current, left current, middle current) and the three longitude and latitude parameters (longitude, latitude and altitude) in the above parameters are used as labels, and the remaining 24 parameters are used as input data.
[0043] Step S4: Input the data obtained in step S3 into the neural network model for iterative training.
[0044] The first 80% of the data is taken as the training set, and the last 20% of the data is taken as the validation set. The training set is used for model training, and the validation set is input into the trained model for testing to obtain the accuracy of the model.
[0045] The neural network model in this step adopts an MLP neural network model, and the MLP neural network model has 3 hidden layers. The setting of the hidden layer can be as in the form 1 in the subsequent embodiment 1: the first hidden layer has 64 neurons, the second hidden layer has 128 neurons, and the third hidden layer has 64 neurons; or as in the form 2 in the subsequent embodiment 2: the first hidden layer has 128 neurons, the second hidden layer has 128 neurons, and the third hidden layer has 128 neurons.
[0046] The input layer of the MLP neural network model sets the 24 input data to 24 neurons, maps them to the hidden layer, and uses the ReLU activation function. The output layer of the MLP neural network model reduces the data dimension to 8 neurons, namely the current data of 5 directional positions and 3 longitude and latitude parameters.
[0047] For form one, the data is reduced in dimension, and x′=V[:,:8]·x (Formula 1) is used to reduce the dimension of 64 neurons to 8 neurons. The 8 neurons are output data, namely 5 current size data (upper current, right current, lower current, left current, middle current) and 3 longitude and latitude parameters (longitude, latitude and altitude). In the above formula 1: x′ represents the neurons of the output layer (8 neurons: 5 current data and 3 longitude and latitude parameters); v represents the matrix; x represents the input neurons, namely 64 neurons; [:,:8] represents taking the first 8 columns of the matrix v; · represents matrix multiplication. The situation in form two is similar and will not be elaborated in detail.
[0048] Step S5: Use the output result of the model after training in step S4 as the test set data to determine whether the test set data meets the iteration number requirement. If not, feed the test set data back to the training set data in step S3, and then perform the operation of step S4 again until the training is terminated after the iteration number is reached to obtain a trained model.
[0049] Feeding the output predicted current data back to the model for further training can improve accuracy. The current can be adjusted by calculating the position information deviation by operating the output latitude and longitude data with the actual latitude and longitude data.
[0050] Step S6: Use the trained model obtained in step S5 as a prediction model, and predict the 5 current prediction values and 3 longitude and latitude high prediction values of the prediction model.
[0051] Step S7: Input the five current prediction values and three latitude and longitude prediction values of the prediction results obtained in step S6 into the PID algorithm controller as process variables of PID control. The expected output of the PID algorithm controller is the thrust value, which is the target value to be adjusted for the shield machine.
[0052] Step S8: Use PID algorithm controller to adjust the proportional parameter K online in real time P , integration parameter K I , differential parameter K D Make it reach the optimal value, and then use the optimal value to ultimately control the excavation work of the shield machine.
[0053] See also Figure 3 , online real-time adjustment of the proportional parameter K P , integration parameter K I , differential parameter K D The specific operation is: First, initialize the parameter K P , K I , K D , use the real-time collected data to calculate the system error, which is the input of the PID algorithm controller, and then the new control output is calculated through the PID algorithm.
[0054] A shield machine automatic driving system of the present invention is applied in the above-mentioned shield machine automatic driving method combining deep learning and PID control, and the system includes: a parameter acquisition module, an MLP module, a PID module, a result display module, and a shield machine execution control module. The parameter acquisition module is used to obtain the parameters of the shield machine. The MLP module has a preset trained MLP neural network model, takes the 24 parameters of the shield machine as input data, and outputs 5 current prediction values and 3 longitude and latitude high prediction values. The PID module runs the PID control algorithm, and uses the 5 current prediction values and 3 longitude and latitude high prediction values obtained by the MLP module as process variables for PID control. The expected output of the PID algorithm controller is the thrust value. The result display module is used to display the calculation results of the PID module. The shield machine execution control module controls the shield machine according to the calculation results of the PID module.
[0055] A computer-readable storage medium of the present invention stores computer program instructions, which, when executed by a processor, implement the above-mentioned shield machine automatic driving method combining deep learning and PID control. The storage medium refers to a carrier storing the above-mentioned shield machine automatic driving method, such as a floppy disk, an optical disk, a DVD, a hard disk, a flash memory, a memory card, etc.
[0056] The following examples introduce in detail the shield machine automatic driving method combining deep learning and PID control. Example 1 is the model data established under the condition of relatively soft soil environment, and Example 2 is the model retrained under the condition of relatively hard soil environment, and the data are different. Figure 1 , the figure shows different types of construction soil layers.
[0057] <Example 1>
[0058] This embodiment provides a shield machine automatic driving method combining deep learning and PID control:
[0059] Step 1: The sensor equipment installed on the outside of the shield machine obtains the external environment data of the shield machine in real time, including the soil type, hardness, latitude and longitude, etc. The collected data is stored in a table in chronological order. At the same time, the shield machine's own parameter data is stored at the same time interval and at the same time, and stored in the same table, and stored in sequence as one piece of data at the same time.
[0060] Step 2: Combine the similar types of data in step 1 in pairs, fuzzify them into 7 levels, establish fuzzy rules, and finally fuzzify the 64 parameters into 32 parameters.
[0061] The specific implementation steps are as follows:
[0062] Step 2.1: Fuzzification. The parameters of the shield machine are combined in pairs and fuzzified separately. For example, the soil hardness parameter and the soil pressure parameter of the screw conveyor are fuzzified into 7 levels with a value range of [0,1] and combined into a new variable assuming that the value range of ψc is [0,1].
[0063] Step 2.2: Establish fuzzy rules as shown in the following table:
[0064]
[0065] Step 2.3: Defuzzification, convert the new variable ψc of fuzzy reasoning into the variable of actual execution. The decision type is defined as follows:
[0066] 0<ψc≤0.5 Small input value 0.5<ψc≤0.6 Moderate input value ψc>0.6 Large input value
[0067] Step 3: Store the shield machine data obtained in step 2 in Excel, and use Python to import the table into the program. Perform data initialization processing, such as filtering out erroneous data and matrix conversion processing, and finally divide the data into input data and labels.
[0068] The specific implementation steps are as follows:
[0069] Step 3.1: Convert the data into a matrix form that can be processed by the computer.
[0070] Step 3.2: Filter out possible erroneous data and retain correct and valid data.
[0071] Step 3.3: The shield machine parameters are input data, among which the current data and position parameters (latitude, longitude and altitude) of 5 positions are used as labels. In order to make the model more reliable and accurate, the output predicted current data is fed back to the model for further training to improve the accuracy. The output longitude, longitude and altitude data are operated with the actual longitude, longitude and altitude data to calculate the position information deviation and adjust the current.
[0072] Step 4: Input the data obtained in step 3 into the MLP algorithm for iterative training. Take the first 80% of the data as the training set and the last 20% of the data as the validation set. Input the data of the training set into the model for training. The input layer sets the first 24 data as 24 neurons and maps them to the hidden layer. The ReLU activation function is used. The MLP model used is the core operation of model training. Finally, the output layer reduces the dimension of the data. Use x′=V[:,:8]·x to reduce the dimension of 64 neurons to 8 neurons, which is the output data, namely 5 current magnitudes and 3 longitude and latitude parameters. Input 20% of the validation set into the trained model for testing to obtain the accuracy of the model. The relevant hyperparameter setting process is as follows:
[0073] 1) Number and size of hidden layers. This embodiment designs three hidden layers. The first layer receives data from 24 neurons in the input layer and has 64 neurons. The second layer has 128 neurons, the third layer has 64 neurons, and the output layer has 8 neurons.
[0074] 2) The activation function uses the ReLu activation function.
[0075] 3) The learning rate determines the update progress of the model parameters in each iteration and is set to 0.01.
[0076] 4) Regularization is a method to control model overfitting, using L2 regularization.
[0077] 5) The number of iterations is set to 5000.
[0078] Step 5: The output result after step 4 training is fed back to the step 3 data, and step 4 training is performed again. The training is terminated when the number of iterations is reached;
[0079] Step 6: The trained model is verified using all the data, and 5 current prediction values and 3 longitude and latitude high prediction values are output;
[0080] Step 7: The position error can be calculated using the latitude and longitude results output by the model. If there is an error, the predicted current in step 6 can be adjusted.
[0081] Step 8: Use the predicted value obtained in step 7 as the process variable of PID control, and set the expected output to the adjusted thrust size.
[0082] Step 9: Use PID algorithm controller to adjust parameter value K online in real time P ,K I ,K D To achieve the optimal value and ultimately control the shield machine to excavate autonomously.
[0083] The specific implementation steps are as follows:
[0084] Step 9.1: Initialize the predicted value and bring the predicted value into the preset thrust solution model.
[0085] Step 9.2: Using the error between the actual value and the predicted value, calculate the proportional term, integral term, and differential term. These terms will be used to adjust the PID parameters online.
[0086] Step 9.3: Based on the error calculation results, use the gradient descent method to adjust the proportional gain K online P , integration time K I and differential time K D .
[0087] Step 9.4: Update the adjusted PID parameters to the controller in real time to ensure that the output of the controller can more accurately adjust the excavation behavior of the shield machine.
[0088] Step 10: The shield machine performs excavation work.
[0089] <Example 2>
[0090] Steps 1 to 3 in Example 2 are the same as those in Example 1 above.
[0091] Step 4: Input the data obtained in step 3 into the MLP algorithm for iterative training. Take the first 80% of the data as the training set and the last 20% of the data as the validation set. Input the data of the training set into the model for training. The input layer sets the first 24 data as 24 neurons and maps them to the hidden layer. The ReLU activation function is used. The MLP model used is the core operation of model training. Finally, the output layer reduces the dimension of the data to obtain 8 neurons, namely 5 current magnitudes and 3 longitude and latitude parameters. Input 20% of the validation set into the trained model for testing to obtain the accuracy of the model. The relevant hyperparameter setting process is as follows:
[0092] 1) Number and size of hidden layers. This embodiment designs three hidden layers. The first layer receives data from 24 neurons in the input layer and has 128 neurons. The second layer has 128 neurons, the third layer has 128 neurons, and the output layer has 8 neurons.
[0093] 2) The activation function uses the ReLu activation function.
[0094] 3) The learning rate determines the update progress of the model parameters in each iteration and is set to 0.05.
[0095] 4) Regularization is a method to control model overfitting, using L2 regularization.
[0096] 5) The number of iterations is set to 2000.
[0097] Steps 5 to 10 in Example 2 are the same as those in Example 1 above.
[0098] The present invention combines deep learning and PID control of the shield machine automatic driving method, through the shield machine's own parameters, external parameters, geological data and the driver's operation of the shield machine, these parameters are substituted into different soil conditions for training and learning, giving full play to the prediction method of MLP deep learning, maintaining the robustness, real-time and interpretability of the algorithm, and improving the accuracy by combining the PID control algorithm, so as to realize the shield machine in the case of unmanned operation through the trained model can autonomously forward dig tunnel behavior. The algorithm has a high accuracy rate and a low shield boundary error, and it is concluded that the combination of MLP and PID control algorithms can accurately dig under different soil conditions.
[0099] The above embodiments are only preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the scope of protection of the present invention.
Claims
1. A shield machine automatic driving method combining deep learning and PID control, characterized in that: The following steps are involved: Step S1: Obtaining parameters of the shield machine; Step S2: fuzzy process all the parameters obtained in step S1 to obtain a set of new data; Step S3: Process the data obtained in step S2 and divide the data into input data and labels; Step S4: input the data obtained in step S3 into the neural network model for iterative training; Step S5: Use the output result of the model trained in step S4 as the test set data to determine whether the test set data meets the iteration number requirement. If not, feed the test set data back to step S3, and then perform the operation of step S4 again until the training is terminated after the iteration number is reached to obtain a trained model. Step S6: Using the trained model obtained in step S5 as a prediction model, and outputting a prediction result of the prediction model; Step S7: input the prediction result obtained in step S6 into the PID algorithm controller as a process variable of PID control; Step S8: Use PID algorithm controller to adjust the proportional parameter K online in real time P , integration parameter K I , differential parameter K D Make it reach the optimal value, and then use the optimal value to ultimately control the excavation work of the shield machine.
2. The shield machine automatic driving method combining deep learning and PID control as claimed in claim 1, characterized in that: in, The parameters acquired in step S1 include the shield machine's own parameters, external environment parameters, and latitude and longitude parameters. The self parameters include current data in five directions: upper current, right current, lower current, left current, and middle current. The latitude, longitude and altitude parameters are longitude, latitude and altitude respectively.
3. The shield machine automatic driving method combining deep learning and PID control as claimed in claim 2, characterized in that: in, The specific process of step S3 is as follows: Step S3-1: first convert the data matrix; Step S3-2: filter out the wrong data and keep the correct data; Step S3-3: The current data of the five directional positions and the latitude and longitude parameters are used as labels, and the remaining parameters are used as input data.
4. The shield machine automatic driving method combining deep learning and PID control as claimed in claim 3, characterized in that: in, Step S4 takes the first 80% of the data obtained in step S3 as a training set, and the last 20% of the data as a validation set, and inputs the training set into the MLP neural network model for iterative training. The input layer sets the 24 input data as 24 neurons, maps them to the hidden layer, and uses the ReLU activation function. Finally, the output layer reduces the data to 8 neurons, namely, the current data of 5 directional positions and 3 longitude and latitude parameters.
5. The shield machine automatic driving method combining deep learning and PID control as claimed in claim 4, characterized in that: in, The MLP neural network model is provided with 3 hidden layers, the first layer has 64 neurons, the second layer has 128 neurons, and the third layer has 64 neurons.
6. The shield machine automatic driving method combining deep learning and PID control as claimed in claim 4, characterized in that: in, The MLP neural network model is provided with 3 hidden layers, the first layer has 128 neurons, the second layer has 128 neurons, and the third layer has 128 neurons.
7. The shield machine automatic driving method combining deep learning and PID control as claimed in claim 4, characterized in that: in, The prediction results obtained in step S6 are 5 current prediction values and 3 longitude and latitude high prediction values. The 5 current prediction values and 3 longitude and latitude high prediction values are used as process variables for PID control. The expected output of the PID algorithm controller is the thrust value, which is the target value to be adjusted for the shield machine.
8. The shield machine automatic driving method combining deep learning and PID control as claimed in claim 1, characterized in that: in, Step S8: online real-time adjustment of the proportional parameter K P , integration parameter K I , differential parameter K D The specific operation is: First, initialize the parameter K P , K I , K D , use the real-time collected data to calculate the system error, which is the input of the PID algorithm controller, and then the new control output is calculated through the PID algorithm.
9. A shield machine automatic driving system, applied in the shield machine automatic driving method combining deep learning and PID control as claimed in any one of claims 1 to 8, characterized in that: include: Parameter acquisition module, used to obtain the parameters of the shield machine; The MLP module has a preset trained MLP neural network model, which takes the 24 parameters of the shield machine as input data and outputs 5 current prediction values and 3 longitude and latitude prediction values; The PID module runs the PID control algorithm and uses the five current prediction values and three latitude and longitude prediction values obtained by the MLP module as the process variables of the PID control. The expected output of the PID algorithm controller is the thrust value. Result display module, used to display the calculation results of PID module; The shield machine execution control module controls the shield machine according to the calculation results of the PID module.
10. A computer-readable storage medium having computer program instructions stored thereon, wherein the computer program instructions, when executed by a processor, implement the shield machine automatic driving method combining deep learning and PID control as described in any one of claims 1 to 8.
Citation Information
Cited By
Laser frequency locking system multistage PID parameter regulation and control method based on deep learning
CN120871580A
