Mining card steering control method and device, terminal, equipment, medium and chip

Abstract

The application provides a mine card steering control method and device, a terminal, equipment, a medium and a chip, and relates to the field of mine card control. The method comprises the following steps: based on the operation data of the unmanned mine card, a model is trained by using a machine learning method to obtain a steering RNN model; based on the steering RNN model, a quadratic optimization problem is constructed, wherein the constraint conditions of the quadratic optimization problem comprise a steering capability limit constraint and a safety limit constraint, the steering capability limit constraint is used to limit the generated target steering control instruction from exceeding the steering capability of the unmanned mine card, and the safety limit constraint is used to suppress the rollover problem of the unmanned mine card; the quadratic optimization problem is solved by using an iterative linear quadratic regulator to obtain a target steering control instruction; and the target steering control instruction is used to control the steer-by-wire actuator of the unmanned mine card.

Description

Technical Field

[0001] This application relates to the field of mining truck control, and in particular to a method, apparatus, terminal, equipment, medium, and chip for controlling the steering of mining trucks. Background Technology

[0002] Since large mining trucks all use hydraulic steering mechanisms, they suffer from large steering delays and poor steering ability, necessitating a lateral control scheme. Existing lateral control schemes present two problems: (1) The steering model used in the existing lateral control is a first-order inertial model, which cannot fully characterize the actual steering performance of mining trucks. In particular, the actual vehicle will experience a weakening and slowing of steering ability when making continuous turns, and the use of a first-order inertial model will produce incorrect predictions of the actual steering situation.

[0003] (2) The kinematic model used in the existing lateral control is a nonlinear model. Then, a linear model is finally obtained through approximation methods such as small angle theory. However, the approximation method has an inherent deviation from the real physical model, which will inevitably lead to model error.

[0004] (3) The existing lateral control does not take into account the rollover problem caused by excessive turning of the mining truck. Summary of the Invention

[0005] In view of this, this application provides a method, device, terminal, equipment, medium and chip for steering control of mining trucks, which solves the problem of inaccurate control caused by the use of a first-order inertial model and model approximation processing in the prior art.

[0006] In a first aspect, embodiments of this application provide a method for controlling the steering of a mining truck, including: Based on the operational data of unmanned mining trucks, a turning RNN model is obtained by training the model using machine learning methods. The turning RNN model includes an input layer, a gated recurrent unit, and an output layer. The input layer is used to acquire the operational data. The gated recurrent unit is used to dynamically process the operational data and historical hidden states through a gating mechanism to achieve long-term memory and obtain the final hidden state vector. The output layer is used to output the turning prediction value based on the operational data. Based on the aforementioned steering RNN model, a quadratic optimization problem is constructed, wherein the constraints of the quadratic optimization problem include steering capability constraint and safety constraint. The steering capability constraint is used to limit the generated target steering control command from exceeding the steering capability of the unmanned mining truck, and the safety constraint is used to suppress the rollover problem of the unmanned mining truck. The target steering control command is obtained by solving the quadratic optimization problem through an iterative linear quadratic regulator. The drive-by-wire steering actuator of the unmanned mining truck is controlled according to the target steering control command.

[0007] Secondly, embodiments of this application provide a mining truck steering control device, comprising: The first model processing module is used to train a turning RNN model based on the operating data of the unmanned mining truck using machine learning methods. The turning RNN model includes an input layer, a gated recurrent unit, and an output layer. The input layer is used to acquire the operating data. The gated recurrent unit is used to dynamically process the operating data and historical hidden states through a gating mechanism to achieve long-term memory and obtain the final hidden state vector. The output layer is used to output the turning prediction value based on the operating data. The second model processing module is used to construct a quadratic optimization problem based on the steering RNN model. The constraints of the quadratic optimization problem include steering capability constraints and safety constraints. The steering capability constraints are used to limit the generated target steering control commands from exceeding the steering capability of the unmanned mining truck. The safety constraints are used to suppress the rollover problem of the unmanned mining truck. The control command output module is used to solve the quadratic optimization problem through an iterative linear quadratic regulator to obtain the target steering control command; The steering control module is used to control the drive-by-wire steering actuator of the unmanned mining truck according to the target steering control command.

[0008] Thirdly, embodiments of this application provide an electronic device including a processor and a memory, the memory storing a program or instructions that can run on the processor, the program or instructions implementing the steps of the method as described in the first aspect when executed by the processor.

[0009] Fourthly, embodiments of this application provide a computer-readable storage medium on which a program or instructions are stored, which, when executed by a processor, implement the steps of the method as described in the first aspect.

[0010] Fifthly, embodiments of this application provide a chip including a processor and a communication interface, the communication interface being coupled to the processor, the processor being used to run programs or instructions to implement the method as described in the first aspect.

[0011] In a sixth aspect, embodiments of this application provide a computer program product stored in a storage medium, which is executed by at least one processor to implement the method as described in the first aspect.

[0012] The beneficial effects of this application are as follows: This application employs a steering RNN model and utilizes machine learning methods for offline training. This allows for more accurate steering predictions during lateral control, ensuring control accuracy. Furthermore, the steering RNN model incorporates a gated recurrent unit (GRU), which not only handles long-term dependencies, improving feature learning and prediction accuracy from runtime data, but also ensures computational efficiency and enhances real-time performance. Further, a quadratic optimization problem is constructed based on the steering RNN model, and the iLQR algorithm is used to calculate steering control commands. This reduces bias caused by linearization, further improving control accuracy. The quadratic optimization problem involves steering capability and safety constraints. By setting these two constraints, the final generated target steering control command is guaranteed not to exceed the steering capability of the unmanned mining truck, and the problem of truck rollover due to excessive steering is avoided, thus improving truck safety.

[0013] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0014] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This paper illustrates a schematic diagram of the working logic of the steering control scheme for an unmanned mining truck according to an embodiment of this application. Figure 2 A flowchart illustrating the mining truck steering control method according to an embodiment of this application is shown; Figure 3 A structural diagram of a gated loop unit according to an embodiment of this application is shown; Figure 4 A schematic diagram of the prediction results of the steering RNN model in an embodiment of this application is shown; Figure 5 A structural block diagram of a mining truck steering control device according to an embodiment of this application is shown; Figure 6 A schematic block diagram of the terminal structure according to an embodiment of this application is shown; Figure 7 A structural block diagram of a computer device according to an embodiment of this application is shown; Figure 8 A structural block diagram of a computer-readable storage medium according to an embodiment of this application is shown; Figure 9A structural block diagram of a chip according to an embodiment of this application is shown. Detailed Implementation

[0015] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0016] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0017] In the prior art, the use of a first-order inertial model in the steering model leads to incorrect predictions of actual steering conditions, and the use of approximation methods to handle nonlinear models results in model errors. Furthermore, existing technologies disclose vehicle trajectory tracking control schemes (e.g., CN120871885A), but these schemes have the following problems: (1) the steering delay of the steering mechanism is not considered in the model; (2) the vehicle speed in the model is always the current value during the solution process; and (3) the rollover problem caused by oversteering is not considered. Combining (1) and (2), the model of this scheme cannot accurately predict the vehicle state. Also, due to (3), the final output command may lead to safety risks. The mining truck steering control scheme provided in this application considers all these problems.

[0018] The following description, in conjunction with the accompanying drawings, details the mining truck steering control method, device, terminal, equipment, medium, and chip provided in this application through specific embodiments and application scenarios. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0019] This application provides a steering control scheme for an unmanned mining truck, such as... Figure 1As shown, it includes: an offline training part for the steering RNN model and an iLQR lateral control part. The offline training part trains the steering RNN model using offline data, which is then fed into the iLQR lateral control part. The iLQR lateral control part constructs a nonlinear kinematic error model based on the steering RNN model. Then, based on the nonlinear kinematic error model and real-time acquired motion data, it constructs a quadratic optimization problem to calculate the steering commands of the mining truck in real time, and outputs the target steering control command to the drive-by-wire steering actuator to achieve lateral control of the mining truck.

[0020] This application provides a method for controlling the steering of mining trucks, such as... Figure 1 and Figure 2 As shown, the method includes: Step 201: Based on the operational data of the unmanned mining truck, a model is trained using machine learning methods to obtain a steering RNN model.

[0021] The steering RNN model includes an input layer, a gated recurrent unit, and an output layer. The input layer is used to acquire the running data. The gated recurrent unit is used to dynamically process the running data and historical hidden states through a gating mechanism to achieve long-term memory and obtain the final hidden state vector. The output layer is used to output the steering prediction value based on the running data.

[0022] In this step, based on the operational data of the unmanned mining truck in actual working scenarios, including multi-source state parameters such as vehicle speed, steering commands, steering feedback, load, and gear position, machine learning methods are used to model and iteratively train the input-output relationship and dynamic characteristics of the steering system, constructing a steering RNN (Recurrent Neural Network) model suitable for the operating characteristics of mining trucks. Replacing the existing first-order inertial model with the steering RNN model can accurately characterize the nonlinear time delay and dynamic changes in the actual steering process, improving steering control accuracy and system robustness.

[0023] In one embodiment of this application, based on the operational data of unmanned mining trucks, a model is trained using machine learning to obtain a redirected RNN model, including: Collect operational data of unmanned mining trucks, including at least one of the following: vehicle speed, steering command, steering feedback, load, and gear. The Kalman filter algorithm is used to filter the running data, and the filtered running data is divided into a training set and a validation set. The basic RNN model is trained and validated based on the training and validation sets to obtain the steering RNN model. The basic RNN model is an RNN model with gated recurrent units.

[0024] In this embodiment, a large amount of operational data of the unmanned mining truck is first collected as training data for the steering RNN model. The collected operational data includes, but is not limited to, the following: vehicle speed, steering commands, steering feedback, load, gear, etc.

[0025] The Kalman filter algorithm is used to filter noise from the running data and remove outliers. Then, the running data is divided into a training set and a validation set according to a preset ratio, such as 4:1. The training set is used to train the model, and the validation set is used to verify the model's accuracy. Based on the training and validation sets, a basic RNN model is trained and validated to obtain the redirected RNN model.

[0026] This application trains a basic RNN model with gated recurrent units (GRUs). In some embodiments, other models, such as LSTM and Transformer, can also be used. Compared to other models, GRUs not only have the ability to handle long-term dependencies but also ensure computational efficiency, making them suitable for scenarios with high real-time requirements.

[0027] In one embodiment of this application, the steering RNN model or the basic RNN model includes an input layer, a gated recurrent unit, and an output layer. Wherein: (1) The input layer is used to obtain the running data and divide the running data into sample sequences according to the set time step t, which are used as the input of the input layer.

[0028] (2) The gated loop unit is used to dynamically process the information flow of running data and historical hidden states through the gated mechanism, realize long-term memory and obtain the final hidden state vector. The final hidden state vector represents the temporal characteristics and key information of the running data.

[0029] In one embodiment of this application, the core structure of the gated loop unit includes two gating mechanisms and a candidate hidden state, achieving long-term memory by dynamically adjusting the information flow. Figure 3 As shown, the gated loop unit includes a reset gate, an update gate, and candidate hidden states.

[0030] Given time step t, reset the gate and Update Gate The calculation formula is: ; ; in, , , , These are weight parameters. , It is a bias parameter. It is the sigmoid activation function. It is a small batch of sample sequences of running data. It is the hidden state at time step t-1.

[0031] Candidate hidden states at time step t The calculation formula is: ; in, , These are weight parameters. It is a bias parameter. It is the symbol for element-wise multiplication.

[0032] The final hidden state vector at time step t The calculation formula is: .

[0033] (3) The output layer is used to output the steering prediction value based on the running data.

[0034] The output of the gated recurrent unit is directly mapped to the final steering prediction value through a linear layer: ; in, For the predicted value of the shift, For weight parameters, This is the bias parameter.

[0035] In one embodiment, mean squared error is used as the loss function for model training, and the loss function L is:

[0036] Where M is the number of training samples, Let be the predicted turning value of the i-th sample output by the model. It is the actual feedback value of the i-th sample.

[0037] The model was trained offline using the training set, and its prediction accuracy was validated using the validation set, resulting in a redirected RNN model. The prediction results of the redirected RNN model are as follows: Figure 4 As shown, by Figure 4 It can be seen that the steering value predicted by using the steering RNN model is more accurate.

[0038] It is worth noting that existing technologies disclose mining truck simulation schemes based on log data (such as CN120745434A), but these schemes only train the LSTM model of the mining truck for simulation testing and do not provide effective engineering applications for the model. Therefore, they differ from the purpose of this application—real-time control of unmanned mining trucks. Furthermore, the mining truck model in this scheme includes all state values ​​in both the horizontal and vertical directions, making training difficult and achieving high accuracy challenging. This application only trains the steering model, which is less difficult, easier to train, and easily achieves higher accuracy. The scheme is simple and easy to implement, and can be well applied in engineering.

[0039] Step 202: Based on the steering RNN model, construct a quadratic optimization problem. The constraints of the quadratic optimization problem include steering capability constraints and safety constraints. The steering capability constraints are used to limit the generated target steering control commands from exceeding the steering capability of the unmanned mining truck, and the safety constraints are used to suppress the rollover problem of the unmanned mining truck.

[0040] In one embodiment of this application, a quadratic optimization problem is constructed based on the aforementioned steering RNN model, including: Based on the steering RNN model, a nonlinear kinematic error model for the unmanned mining truck is constructed; based on the nonlinear kinematic error model, the quadratic optimization problem is constructed.

[0041] In this embodiment, a nonlinear kinematic error model is constructed based on a steering RNN model. This embodiment employs a nonlinear kinematic error model; however, other embodiments may also use a dynamic model.

[0042] The nonlinear kinematic error model in this application is a nonlinear discrete model, and the nonlinear kinematic error model is set as follows:

[0043]

[0044] in, It is the vehicle state value at time k, including lateral error. Heading error and steering feedback , It is the vehicle state value at time k+1, including lateral error. Heading error and steering feedback f represents the nonlinear state transition function. It is the steering control command at time k. Let k be the vehicle speed at time k. For discrete periods, Let k be the reference path curvature. This refers to the vehicle's wheelbase. This refers to the steering RNN model.

[0045] A nonlinear kinematic error model is constructed based on the trained steering RNN model, thereby more accurately describing the error between actual and theoretical motion and improving the accuracy of pose estimation and motion control.

[0046] In one embodiment, the above-mentioned nonlinear kinematic error model includes the vehicle speed at time k. Where, the vehicle speed at time k Based on the target vehicle speed at time k, the estimated speed is obtained by combining the first-order inertial model of the speed, which enables the prediction of the vehicle state to achieve high accuracy in the subsequent optimization solution.

[0047] In one embodiment, the vehicle speed at time k for:

[0048] in, The inertial parameter values ​​for the first-order inertial model of velocity are obtained by calculating using the least squares method with offline running data. , Let k be the vehicle speed at time k-1. Let k be the target vehicle speed at time k.

[0049] In this embodiment of the application, a nonlinear model is designed in conjunction with a steering RNN model for lateral control, wherein a first-order inertial model is used to predict the velocity, which further improves the accuracy of the model.

[0050] Furthermore, a quadratic optimization problem is constructed based on a nonlinear kinematic error model. In this embodiment, a quadratic optimization problem is defined, incorporating steering capability and safety constraints to more closely reflect the actual steering performance of mining trucks while ensuring safety. The quadratic optimization problem is:

[0051]

[0052]

[0053]

[0054]

[0055]

[0056] in, The total control cost function, For the terminal cost function, Let k be the cost function. The initial state value, For steering control sequence, To predict the time domain, This is the terminal state value. Let k be the vehicle state value at time k. It is the steering control command at time k. Let k+1 be the vehicle state value. The state weight matrix is... To control the weight matrix, This is the final state weight matrix.

[0057] In addition, to ensure the usability and security of the output results, the following inequality constraints need to be added: (1) Steering capability limitation constraints, including steering limit value constraints and steering increment constraints, as shown below:

[0058]

[0059]

[0060] in, For the maximum turning angle, This represents the maximum single-step increment of control commands during adjacent control cycles. The control command from the previous control cycle. The maximum single-step increment of control commands during adjacent forecast weeks. It is the steering control command at time k. It is the steering control command at time k+1.

[0061] By setting steering capability limit constraints, it is ensured that the final generated target steering control command does not exceed the steering capability of the unmanned mining truck.

[0062] (2) Safety Constraints. Due to the high center of gravity of unmanned mining trucks and the fact that the operating surfaces are often uneven and sloping, vehicles are prone to rollover due to lateral load transfer. Therefore, this application provides safety constraints applicable to unmanned mining trucks, as shown in the following formula:

[0063]

[0064]

[0065] in, This is the upper limit of lateral acceleration. This is the lower limit of lateral acceleration. The estimated coefficients for quasi-static rollover. These are the estimation coefficients for the transient rollover threshold. It is the acceleration due to gravity. The road surface tilt angle. This refers to the distance between the left and right tires of the mining truck. The height of the center of mass from the ground. Let k be the vehicle speed at time k. For the steering feedback at time k, This refers to the vehicle's wheelbase.

[0066] By setting safety limits and constraints, the problem of mining trucks tipping over due to excessive turning can be avoided.

[0067] Step 203: Solve the quadratic optimization problem by iterating a linear quadratic regulator to obtain the target steering control command.

[0068] In this step, based on the established quadratic optimization problem, and combined with the Iterative Linear Quadratic Regulator (iLQR) algorithm, optimal constraints and rolling optimizations are performed on the system state and control variables. By iteratively solving for the optimal control increment, the target steering control command that satisfies the driving stability and tracking accuracy of the unmanned mining truck is finally calculated. iLQR is an iterative optimization algorithm for optimal control of nonlinear systems, combining dynamic programming and local linearization techniques.

[0069] This application employs the iLQR optimization algorithm; other examples may also use nonlinear solvers such as IPOPT. iLQR is an efficient iterative numerical algorithm for solving optimal control problems of nonlinear systems. At each time step, it linearizes the system state equations using a first-order Taylor expansion and performs a second-order approximation of the cost function, then uses a linear quadratic regulator (LQR) to solve this approximation problem. By iteratively adjusting the strategy, it gradually approximates the optimal strategy. In one embodiment, based on the iterative linear quadratic regulator, a quadratic optimization problem is solved to obtain the target steering control command, such as... Figure 1 As shown, it includes: Step 1, Initialization and Forward Simulation: Determine the initial state values Initial steering control sequence ( and convergence threshold Based on the initial state value Forward simulation is performed using the initial steering control sequence and the nonlinear kinematic error model to generate the initial state sequence. ), and calculate the initial control cost. .

[0070] Step 2, Backward Propagation: Recursively tracing from the end of the prediction time domain to the initial time, at each sampling time, the nonlinear kinematic error model is expanded using Taylor at the corresponding state value and control quantity to obtain a linearized model. The linear quadratic adjustment subproblem corresponding to the linearized model, i.e., the LQ subproblem, is solved to obtain the control gain, the gradient of the cost function, and the Hessian matrix, providing direction for the optimization and updating of the control sequence.

[0071] Step 3, Forward Propagation: Based on the control gain obtained from backward propagation, update the steering control sequence and state sequence, recalculate the control cost, and verify whether the control cost has been reduced (i.e., whether the control performance has been optimized), thus completing one iteration update.

[0072] Step 4, Iterative Convergence Judgment: Compare the difference between the control cost of the current iteration and the previous iteration; if the difference is greater than the convergence threshold... If the difference is less than the convergence threshold, then repeat steps 2 and 3; The algorithm converges and outputs the optimal state sequence. and optimal steering control sequence .

[0073] Step 5: Determine the target steering control command based on the commands in the optimal steering control sequence. For example, take the first command in the optimal steering control sequence. As a target steering control command.

[0074] Step 204: Control the wire-controlled steering actuator of the unmanned mining truck according to the target steering control command.

[0075] In this step, the target steering control command is output to the drive-by-wire steering actuator to achieve lateral control of the mining truck.

[0076] Compared to the first-order inertial steering model in existing technologies, this application employs a steering RNN model and utilizes machine learning methods for offline training of the steering RNN model. This allows for more accurate steering prediction when using this model for lateral control, ensuring control accuracy. Furthermore, the steering RNN model incorporates a gated recurrent unit (GRU), which not only handles long-term dependencies, improving feature learning and prediction accuracy of runtime data, but also ensures computational efficiency and enhances real-time performance. Further, a quadratic optimization problem is constructed based on the steering RNN model, and the iLQR algorithm is used to calculate steering control commands. Compared to the linearization methods used in existing technologies that approximate linear models, this reduces the bias introduced by linearization, further improving control accuracy. Moreover, the quadratic optimization problem involves steering capability constraints and safety constraints. By setting these two constraints, the final generated target steering control command is guaranteed not to exceed the steering capability of the unmanned mining truck, and the problem of the mining truck tipping over due to excessive steering is avoided, thus improving the safety of the mining truck.

[0077] As a specific implementation of the above-mentioned mining truck steering control method, this application provides a mining truck steering control device. For example... Figure 5 As shown, the mining truck steering control device 500 includes: a first model processing module 501, a second model processing module 502, a control command output module 503, and a steering control module 504.

[0078] The first model processing module 501 is used to train a turning RNN model based on the operating data of the unmanned mining truck using machine learning methods. The turning RNN model includes an input layer, a gated recurrent unit, and an output layer. The input layer is used to acquire the operating data. The gated recurrent unit is used to dynamically process the operating data and historical hidden states through a gating mechanism to achieve long-term memory and obtain the final hidden state vector. The output layer is used to output the turning prediction value based on the operating data. The second model processing module 502 is used to construct a quadratic optimization problem based on the steering RNN model. The constraints of the quadratic optimization problem include steering capability constraint and safety constraint. The steering capability constraint is used to limit the generated target steering control command from exceeding the steering capability of the unmanned mining truck. The safety constraint is used to suppress the rollover problem of the unmanned mining truck. The control command output module 503 is used to solve the quadratic optimization problem through an iterative linear quadratic regulator to obtain the target steering control command; Steering control module 504 is used to control the drive-by-wire steering actuator of the unmanned mining truck according to the target steering control command.

[0079] Furthermore, based on the aforementioned turning RNN model, a quadratic optimization problem is constructed, including: Based on the steering RNN model, a nonlinear kinematic error model for the unmanned mining truck is constructed; the nonlinear kinematic error model includes the vehicle speed at time k, wherein the vehicle speed at time k is estimated based on the target vehicle speed at time k and combined with the first-order inertial speed model. Based on the aforementioned nonlinear kinematic error model, the quadratic optimization problem is constructed.

[0080] Furthermore, the nonlinear kinematic error model is as follows:

[0081]

[0082] in, It is the vehicle state value at time k, including lateral error. Heading error and steering feedback , It is the vehicle state value at time k+1, including lateral error. Heading error and steering feedback f represents the nonlinear state transition function. It is the steering control command at time k. Let k be the vehicle speed at time k. For discrete periods, Let k be the reference path curvature. This refers to the vehicle's wheelbase. For the aforementioned steering RNN model; The vehicle speed at time k for:

[0083] in, The inertial parameter values ​​for the first-order inertial model of velocity are obtained by calculating using the least squares method with offline running data. , Let k be the vehicle speed at time k-1. Let k be the target vehicle speed at time k.

[0084] Furthermore, the quadratic optimization problem is:

[0085]

[0086]

[0087]

[0088]

[0089]

[0090] in, The total control cost function, For the terminal cost function, Let k be the cost function. The initial state value, For steering control sequence, To predict the time domain, This is the terminal state value. Let k be the vehicle state value at time k. It is the steering control command at time k. Let k+1 be the vehicle state value. The state weight matrix is... To control the weight matrix, This is the final state weight matrix; The steering capability limitation constraint is as follows:

[0091]

[0092]

[0093] in, For the maximum turning angle, This represents the maximum single-step increment of control commands during adjacent control cycles. The control command from the previous control cycle. The maximum single-step increment of control commands during adjacent forecast weeks. It is the steering control command at time k. It is the steering control command at time k+1; The security constraints are as follows:

[0094]

[0095]

[0096] in, This is the upper limit of lateral acceleration. This is the lower limit of lateral acceleration. The estimated coefficients for quasi-static rollover. These are the estimation coefficients for the transient rollover threshold. It is the acceleration due to gravity. The road surface tilt angle. This refers to the distance between the left and right tires of the mining truck. The height of the center of mass from the ground. Let k be the vehicle speed at time k. For the steering feedback at time k, This refers to the vehicle's wheelbase.

[0097] Furthermore, the first model processing module 501 is specifically used for: The operation data of the unmanned mining truck is collected, and the operation data includes at least one of the following: vehicle speed, steering command, steering feedback, load, and gear. The Kalman filter algorithm is used to filter the running data, and the filtered running data is divided into a training set and a validation set. The basic RNN model is trained and validated based on the training set and the validation set to obtain the steering RNN model. The basic RNN model is an RNN model with gated recurrent units.

[0098] Furthermore, the gated loop unit includes a reset gate, an update gate, and candidate hidden states; Reset gate for time step t and Update Gate The calculation formula is: ; ; in, , , , These are weight parameters. , It is a bias parameter. It is the sigmoid activation function. It is a small batch sample sequence of the running data. It is the hidden state at time step t-1; Candidate hidden states at time step t The calculation formula is: ; in, , These are weight parameters. It is a bias parameter. It uses the sign of element-wise multiplication; The final hidden state vector at time step t The calculation formula is: ; The steering prediction value output by the output layer is: ; in, The predicted steering value, For weight parameters, These are bias parameters; The loss function for model training is:

[0099] Where M is the number of training samples, Let be the predicted turning value of the i-th sample output by the model. It is the actual feedback value of the i-th sample.

[0100] Furthermore, the control command output module 503 is specifically used for: Step 1, Initialization and Forward Simulation: Determine the initial state values Initial steering control sequence and convergence threshold Based on the initial state value The initial steering control sequence and the nonlinear kinematic error model are used for forward simulation to generate the initial state sequence and calculate the initial control cost. ; Step 2, reverse propagation: recursively from the end of the prediction time domain to the initial time. At each sampling time, the nonlinear kinematic error model is Taylor expanded at the corresponding state value and control quantity to obtain a linearized model. The linear quadratic adjustment subproblem corresponding to the linearized model is solved to obtain the control gain, the gradient of the cost function, and the Hessian matrix. Step 3, Forward pass: Based on the control gain obtained from the backward pass, update the steering control sequence and state sequence, recalculate the control cost, and verify whether the control cost has decreased, thus completing one iteration update; Step 4, Iterative Convergence Judgment: Compare the difference between the control cost of the current iteration and the previous iteration; if the difference is greater than the convergence threshold... Then repeat steps 2 and 3; if the difference is less than the convergence threshold Then the optimal state sequence and the optimal steering control sequence will be output; Step 5: Determine the target steering control command based on the commands in the optimal steering control sequence.

[0101] The mining truck steering control device 500 in this application embodiment can be a computer device or a component within a computer device, such as an integrated circuit or a chip. The mining truck steering control device 500 provided in this application embodiment can achieve... Figure 1 and Figure 2The various processes implemented in the mining truck steering control method embodiment will not be described again here to avoid repetition.

[0102] This application also provides a terminal, such as... Figure 6 As shown, the terminal 600 includes the aforementioned mining truck steering control device 500.

[0103] The terminal 600 described above can execute the mining truck steering control method described in the above embodiments through the mining truck steering control device 500. It is understood that the way the terminal 600 controls the mining truck steering control device 500 can be set according to the actual application scenario, and this application embodiment does not make specific limitations.

[0104] The aforementioned terminal 600 includes, but is not limited to, vehicles, vehicle-mounted terminals, vehicle-mounted controllers, vehicle-mounted modules, vehicle-mounted components, vehicle-mounted chips, vehicle-mounted units, vehicle-mounted radar, or vehicle-mounted cameras, and other sensors. Vehicles can implement the methods provided in this application through these vehicle-mounted terminals, controllers, modules, components, chips, units, radar, or cameras. Vehicles in this application include passenger cars and commercial vehicles. Common commercial vehicle models include, but are not limited to, pickup trucks, mini-trucks, light trucks, mini-vans, dump trucks, cargo trucks, tractors, trailers, special-purpose vehicles, and mining vehicles. Mining vehicles include, but are not limited to, mining trucks, wide-body trucks, articulated trucks, excavators, electric shovels, and bulldozers. This application does not further limit the type of intelligent vehicle; any type of vehicle is within the scope of protection of this application.

[0105] This application also provides a computer device, such as... Figure 7 As shown, the computer device 700 includes a first processor 701 and a first memory 702. The first memory 702 stores programs or instructions that can run on the first processor 701. When the program or instructions are executed by the first processor 701, they implement the various steps of the above-described mining truck steering control method embodiment and achieve the same technical effect. To avoid repetition, they will not be described again here.

[0106] The first memory 702 can be used to store software programs and various data. The first memory 702 may primarily include a first storage area for storing programs or instructions and a second storage area for storing data. The first storage area may store the operating system, application programs or instructions required for at least one function (such as sound playback function, image playback function, etc.). Furthermore, the first memory 702 may include volatile memory or non-volatile memory, or both. The non-volatile memory may be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory can be random access memory (RAM), static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct memory bus RAM (DRRAM). The first memory 702 in the embodiments of this application includes, but is not limited to, these and any other suitable types of memory.

[0107] The first processor 701 may include one or more processing units; optionally, the first processor 701 integrates an application processor and a modem processor, wherein the application processor mainly handles operations involving the operating system, user interface, and applications, and the modem processor mainly handles wireless communication signals, such as a baseband processor. It is understood that the aforementioned modem processor may also not be integrated into the first processor 701.

[0108] This application also provides a computer-readable storage medium, such as... Figure 8 As shown, a program or instruction 801 is stored on the computer-readable storage medium 800. When the program or instruction 801 is executed by the processor, it implements the various processes of the above-described mining truck steering control method embodiment and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0109] The methods described in the above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any combination thereof. The computer-readable storage medium 800 may include computer storage media and communication media, and may also include any medium capable of transferring a computer program from one place to another. The storage medium can be any target medium accessible by a computer.

[0110] As one possible design, computer-readable storage medium 800 may include a compact optical disc read-only memory (CD-ROM), RAM, ROM, EEPROM, or other optical disc storage; computer-readable storage medium 800 may also include a disk storage device or other disk storage device. Furthermore, any connecting cable may also be appropriately referred to as a computer-readable storage medium. For example, if software is transmitted from a website, server, or other remote source using coaxial cable, fiber optic cable, twisted pair, DSL (Digital Subscriber Line), or wireless technologies such as infrared, radio, and microwave, then coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of medium. As used herein, disks and optical discs include optical discs (CD), laser discs, optical discs, digital versatile discs (DVD), floppy disks, and Blu-ray discs, where disks typically reproduce data magnetically, while optical discs optically reproduce data using lasers.

[0111] This application also provides a chip, such as... Figure 9 As shown, the chip 900 includes at least one second processor 901 and a communication interface 902. The communication interface 902 is coupled to the second processor 901. The second processor 901 is used to run programs or instructions to implement the various processes of the above-described mining truck steering control method embodiment and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0112] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0113] Preferably, the chip 900 further includes a memory, such as a second memory 903, which stores executable modules or data structures, or subsets thereof, or extended sets thereof.

[0114] In this embodiment, the second memory 903 may include read-only memory and random access memory, and provides instructions and data to the second processor 901. A portion of the second memory 903 may also include non-volatile random access memory (NVRAM).

[0115] In this embodiment, the second processor 901, the communication interface 902, and the second memory 903 are coupled together via a bus system 904. The bus system 904 may include a data bus, a power bus, a control bus, and a status signal bus, in addition to the data bus. For ease of description, in... Figure 9 The general labeled all buses as Bus System 904.

[0116] The mining truck steering control method described in the embodiments of this application can be applied to, or implemented by, the second processor 901. The second processor 901 may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware or by instructions in software within the second processor 901. The second processor 901 can be a general-purpose processor (e.g., a microprocessor or conventional processor), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gates, transistor logic devices, or discrete hardware components. The second processor 901 can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention.

[0117] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0118] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A method for controlling the steering of a mining truck, characterized in that, include: Based on the operational data of unmanned mining trucks, a turning RNN model is obtained by training the model using machine learning methods. The turning RNN model includes an input layer, a gated recurrent unit, and an output layer. The input layer is used to acquire the operational data. The gated recurrent unit is used to dynamically process the operational data and historical hidden states through a gating mechanism to achieve long-term memory and obtain the final hidden state vector. The output layer is used to output the turning prediction value based on the operational data. Based on the aforementioned steering RNN model, a quadratic optimization problem is constructed, wherein the constraints of the quadratic optimization problem include steering capability constraint and safety constraint. The steering capability constraint is used to limit the generated target steering control command from exceeding the steering capability of the unmanned mining truck, and the safety constraint is used to suppress the rollover problem of the unmanned mining truck. The target steering control command is obtained by solving the quadratic optimization problem through an iterative linear quadratic regulator. The wire-controlled steering actuator of the unmanned mining truck is controlled according to the target steering control command; The construction of a quadratic optimization problem based on the aforementioned steering RNN model includes: Based on the steering RNN model, a nonlinear kinematic error model for the unmanned mining truck is constructed; the nonlinear kinematic error model includes the vehicle speed at time k, wherein the vehicle speed at time k is estimated based on the target vehicle speed at time k and combined with the first-order inertial speed model. Based on the aforementioned nonlinear kinematic error model, the quadratic optimization problem is constructed. The nonlinear kinematic error model is as follows: in, It is the vehicle state value at time k, including lateral error. Heading error and steering feedback , It is the vehicle state value at time k+1, including lateral error. Heading error and steering feedback f represents the nonlinear state transition function. It is the steering control command at time k. Let k be the vehicle speed at time k. For discrete periods, Let k be the reference path curvature. This refers to the vehicle's wheelbase. For the aforementioned steering RNN model; The vehicle speed at time k for: in, The inertial parameter values ​​for the first-order inertial model of velocity are obtained by calculating using the least squares method with offline running data. , Let k be the vehicle speed at time k-1. Let k be the target vehicle speed at time k; The quadratic optimization problem is: in, The total control cost function, For the terminal cost function, Let k be the cost function. The initial state value, For steering control sequence, To predict the time domain, This is the terminal state value. Let k be the vehicle state value at time k. It is the steering control command at time k. Let k+1 be the vehicle state value. The state weight matrix is... To control the weight matrix, This is the final state weight matrix; The steering capability limitation constraint is as follows: in, For the maximum turning angle, This represents the maximum single-step increment of control commands during adjacent control cycles. The control command from the previous control cycle. The maximum single-step increment of control commands during adjacent forecast weeks. It is the steering control command at time k. It is the steering control command at time k+1; The security constraints are as follows: in, This is the upper limit of lateral acceleration. This is the lower limit of lateral acceleration. The estimated coefficients for quasi-static rollover. These are the estimation coefficients for the transient rollover threshold. It is the acceleration due to gravity. The road surface tilt angle. This refers to the distance between the left and right tires of the mining truck. The height of the center of mass from the ground. Let k be the vehicle speed at time k. For the steering feedback at time k, This refers to the vehicle's wheelbase.

2. The mining truck steering control method according to claim 1, characterized in that, The operational data based on the unmanned mining trucks is used to train a model using machine learning methods, resulting in a redirected RNN model, including: The operation data of the unmanned mining truck is collected, and the operation data includes at least one of the following: vehicle speed, steering command, steering feedback, load, and gear. The Kalman filter algorithm is used to filter the running data, and the filtered running data is divided into a training set and a validation set. The basic RNN model is trained and validated based on the training set and the validation set to obtain the steering RNN model. The basic RNN model is an RNN model with gated recurrent units.

3. The mining truck steering control method according to claim 1, characterized in that, The gated loop unit includes a reset gate, an update gate, and candidate hidden states; Reset gate for time step t and Update Gate The calculation formula is: ； ； in, , , , These are weight parameters. , It is a bias parameter. It is the sigmoid activation function. It is a small batch sample sequence of the running data. It is the hidden state at time step t-1; Candidate hidden states at time step t The calculation formula is: ； in, , These are weight parameters. It is a bias parameter. It uses the sign of element-wise multiplication; The final hidden state vector at time step t The calculation formula is: ； The steering prediction value output by the output layer is: ； in, The predicted steering value, For weight parameters, These are bias parameters; The loss function for model training is: Where M is the number of training samples, Let be the predicted turning value of the i-th sample output by the model. It is the actual feedback value of the i-th sample.

4. The mining truck steering control method according to claim 1, characterized in that, The step of solving the quadratic optimization problem through an iterative linear quadratic regulator to obtain the target steering control command includes: Step 1, Initialization and Forward Simulation: Determine the initial state values Initial steering control sequence and convergence threshold Based on the initial state value The initial steering control sequence and the nonlinear kinematic error model are used for forward simulation to generate the initial state sequence and calculate the initial control cost. ; Step 2, reverse propagation: recursively from the end of the prediction time domain to the initial time. At each sampling time, the nonlinear kinematic error model is Taylor expanded at the corresponding state value and control quantity to obtain a linearized model. The linear quadratic adjustment subproblem corresponding to the linearized model is solved to obtain the control gain, the gradient of the cost function, and the Hessian matrix. Step 3, Forward pass: Based on the control gain obtained from the backward pass, update the steering control sequence and state sequence, recalculate the control cost, and verify whether the control cost has decreased, thus completing one iteration update; Step 4, Iterative Convergence Judgment: Compare the difference between the control cost of the current iteration and the previous iteration; if the difference is greater than the convergence threshold... Then repeat steps 2 and 3; if the difference is less than the convergence threshold Then the optimal state sequence and the optimal steering control sequence will be output; Step 5: Determine the target steering control command based on the commands in the optimal steering control sequence.

5. A mining truck steering control device, characterized in that, include: The first model processing module is used to train a turning RNN model based on the operating data of the unmanned mining truck using machine learning methods. The turning RNN model includes an input layer, a gated recurrent unit, and an output layer. The input layer is used to acquire the operating data. The gated recurrent unit is used to dynamically process the operating data and historical hidden states through a gating mechanism to achieve long-term memory and obtain the final hidden state vector. The output layer is used to output the turning prediction value based on the operating data. The second model processing module is used to construct a quadratic optimization problem based on the steering RNN model. The constraints of the quadratic optimization problem include steering capability constraints and safety constraints. The steering capability constraints are used to limit the generated target steering control commands from exceeding the steering capability of the unmanned mining truck. The safety constraints are used to suppress the rollover problem of the unmanned mining truck. The control command output module is used to solve the quadratic optimization problem through an iterative linear quadratic regulator to obtain the target steering control command; The steering control module is used to control the drive-by-wire steering actuator of the unmanned mining truck according to the target steering control command; The second model processing module is specifically used for: Based on the steering RNN model, a nonlinear kinematic error model for the unmanned mining truck is constructed; the nonlinear kinematic error model includes the vehicle speed at time k, wherein the vehicle speed at time k is estimated based on the target vehicle speed at time k and combined with the first-order inertial speed model. Based on the aforementioned nonlinear kinematic error model, the quadratic optimization problem is constructed. The nonlinear kinematic error model is as follows: in, It is the vehicle state value at time k, including lateral error. Heading error and steering feedback , It is the vehicle state value at time k+1, including lateral error. Heading error and steering feedback f represents the nonlinear state transition function. It is the steering control command at time k. Let k be the vehicle speed at time k. For discrete periods, Let k be the reference path curvature. This refers to the vehicle's wheelbase. For the aforementioned steering RNN model; The vehicle speed at time k for: in, The inertial parameter values ​​for the first-order inertial model of velocity are obtained by calculating using the least squares method with offline running data. , Let k be the vehicle speed at time k-1. Let k be the target vehicle speed at time k; The quadratic optimization problem is: in, The total control cost function, For the terminal cost function, Let k be the cost function. The initial state value, For steering control sequence, To predict the time domain, This is the terminal state value. Let k be the vehicle state value at time k. It is the steering control command at time k. Let k+1 be the vehicle state value. The state weight matrix is... To control the weight matrix, This is the final state weight matrix; The steering capability limitation constraint is as follows: in, For the maximum turning angle, This represents the maximum single-step increment of control commands during adjacent control cycles. The control command from the previous control cycle. The maximum single-step increment of control commands during adjacent forecast weeks. It is the steering control command at time k. It is the steering control command at time k+1; The security constraints are as follows: in, This is the upper limit of lateral acceleration. This is the lower limit of lateral acceleration. The estimated coefficients for quasi-static rollover. These are the estimation coefficients for the transient rollover threshold. It is the acceleration due to gravity. The road surface tilt angle. This refers to the distance between the left and right tires of the mining truck. The height of the center of mass from the ground. Let k be the vehicle speed at time k. For the steering feedback at time k, This refers to the vehicle's wheelbase.

6. A terminal, characterized in that, The terminal includes the mining truck steering control device as described in claim 5.

7. A computer device, characterized in that, It includes a first processor and a first memory, the first memory storing a program or instructions that run on the first processor, the program or instructions being executed by the first processor to implement the steps of the mining truck steering control method as described in any one of claims 1 to 4.

8. A computer-readable storage medium having a program or instructions stored thereon, characterized in that, When the program or instructions are executed by the processor, they implement the steps of the mining truck steering control method as described in any one of claims 1 to 4.

9. A chip, characterized in that, The chip includes at least one second processor and a communication interface, the communication interface being coupled to the at least one second processor, the at least one second processor being used to run programs or instructions to implement the steps of the mining truck steering control method as described in any one of claims 1 to 4.

Images