Neural network-based tracked vehicle motion state control method and system, and medium

Abstract

The invention discloses a tracked vehicle motion state control method and system based on a neural network, and a medium. The method comprises the following steps: constructing a neural network prediction model based on multivariate data driving; performing model optimization on the neural network prediction model based on multivariate data; and constructing a motion state prediction model based on the optimized neural network prediction model, obtaining a reference trajectory of the tracked vehicle, calculating an optimal control parameter based on the reference trajectory and the motion state prediction model, and performing motion control on the tracked vehicle based on the optimal control parameter. The optimal control parameters are returned to the neural network prediction model for model parameter updating; according to the method, a neural network model based on data driving can be adopted as a prediction model, the problems that a traditional physical model is uncertain in parameter, poor in adaptability and difficult to analyze are solved, and real-time prediction of the motion state of the tracked vehicle and real-time dynamic adjustment of prediction model parameters are carried out based on an online learning mechanism.

Description

Technical Field

[0001] The present invention relates to the technical field of tracked vehicle prediction control, and in particular to a method, system and medium for controlling the motion state of a tracked vehicle based on a neural network. Background Art

[0002] As a highly adaptable and versatile mobile platform, unmanned tracked vehicles have demonstrated their irreplaceable advantages in multiple high-risk and complex environments such as military reconnaissance, engineering operations, emergency rescue, and geological exploration. Such vehicles can traverse complex terrains, including steep slopes, gullies, and muddy areas, providing strong support for various tasks. However, the driving stability and control accuracy of tracked vehicles in complex terrains have always been the focus and difficulty of research.

[0003] When a tracked vehicle is traveling on complex terrain, it will be significantly affected by the irregularities and instantaneous changes of the road surface. Especially under extreme conditions, such as steep slopes and gullies, if the driving state of the tracked vehicle cannot be accurately predicted, it will be difficult for the system to formulate targeted control strategies in advance based on the future motion state. This will cause the vehicle to have a dynamic response lag when dealing with changing terrain, increasing the instability of control. Especially when driving at high speeds or turning, if the track drive system cannot be controlled in time, the risk of vehicle slipping, rollover, etc. will increase, seriously threatening the safety and success rate of the mission.

[0004] At present, the research on tracked vehicle state prediction control method has made some progress, but there are still many shortcomings, as follows: (1) Chinese patent CN116861784A discloses a tracked vehicle motion control method combined with Gaussian regression model predictive control. This method decomposes the dynamics of the tracked vehicle into a modelable part and a difficult-to-model part, and predicts the difficult-to-model part through a Gaussian regression model. However, this type of method is more suitable for relatively simple and regular terrain conditions. In complex and changeable terrain, due to the neglect of the analysis of important parameters such as ground pressure, the dynamics are not fully considered, which may cause prediction deviation and reduce the control accuracy and stability of the vehicle.

[0005] (2) China's patent CN117784610A discloses a control method for a dual-motor-driven tracked vehicle based on a neural network prediction model. This method uses a multi-body dynamics model to generate the vehicle's motion state data under various working conditions, and combines it with an LSTM neural network model for prediction. However, this type of method mainly relies on simulation data rather than real environment data, so it is not well adapted to dynamic scenarios. At the same time, the LSTM model may not be able to respond in time to sudden changes in terrain or abrupt changes in vehicle status, resulting in a decrease in prediction accuracy.

[0006] (3) China's patent CN117864156A discloses a method and system for predicting the state of a tracked vehicle under extreme conditions. The solution constructs a comprehensive control model containing multiple dynamic models and uses a multiplexed rolling time domain estimation method to optimize the model parameters for prediction. However, this type of method relies on an idealized physical model, and the strong coupling of dynamics in actual tracked vehicle driving and the complexity of the relationship between the track and the ground make the constructed dynamic model inaccurate. In addition, the factors considered by the model cannot fully cover all actual situations. At the same time, due to the lack of online data collection and real-time update capabilities, the model is ultimately unable to cope with changes in terrain during actual driving.

[0007] In summary, the existing technical solutions have the following deficiencies in predicting and controlling the motion state of tracked vehicles: (1) The existing technical solutions do not consider the impact of ground pressure on the vehicle's motion state in complex terrain environments, resulting in large prediction errors and low control accuracy.

[0008] (2) Most existing technical solutions rely on static model parameters trained offline, which makes it difficult to cope with the dynamic changes of vehicle status and environmental conditions under complex working conditions.

[0009] (3) Existing technical solutions generally analyze the motion state of tracked vehicles based on physical models. However, in complex terrain, the nonlinearity and strong coupling of tracked vehicle dynamics make it difficult to effectively describe them with traditional physical models, resulting in poor prediction and control accuracy.

[0010] Therefore, in order to overcome the shortcomings of the existing technology, it is urgent to develop a data-driven tracked vehicle motion state prediction control method that can adapt to complex terrain. Summary of the invention

[0011] The object of the present invention is to provide a method, system and medium for controlling the motion state of a tracked vehicle based on a neural network, thereby solving all or one of the above-mentioned problems existing in the prior art.

[0012] In order to solve the above technical problems, the specific technical solutions of the present invention are as follows: In one aspect, the present invention provides a method for controlling the motion state of a tracked vehicle using a neural network, comprising the following steps: Construct a neural network prediction model driven by multivariate data; Performing model optimization on the neural network prediction model based on multivariate data; A motion state prediction model is constructed based on the optimized neural network prediction model, a reference trajectory of the tracked vehicle is obtained, optimal control parameters are calculated based on the reference trajectory and the motion state prediction model, motion control of the tracked vehicle is performed based on the optimal control parameters, and the optimal control parameters are returned to the neural network prediction model to update the model parameters.

[0013] As an improved solution, the construction of a neural network prediction model driven by multivariate data includes: A three-layer BP neural network including a single hidden layer is used as the neural network prediction model; collecting the suspension pressure value of the tracked vehicle as a feedback state parameter set; collecting road surface type information of the tracked vehicle as a road surface type parameter set; Collecting the three-dimensional position, three-dimensional speed, body heading angle, body roll angle, body pitch angle and body yaw angular velocity of the tracked vehicle as a motion state parameter set; collecting driving torques on both sides of the tracked vehicle as a torque parameter set; The feedback state parameter set, the road surface type parameter set, the motion state parameter set and the torque parameter set are used as the multivariate data for training the neural network prediction model.

[0014] As an improved solution, the neural network prediction model is optimized based on multivariate data, including: performing normalization processing on the multivariate data; Initializing weight parameters and bias parameters of the neural network prediction model; A function that measures the deviation between the predicted value and the real data at each moment based on the mean square error is used as a loss function, the normalized multivariate data is input into the initialized neural network prediction model, and the neural network prediction model is trained based on the loss function; The gradient parameters of the neural network prediction model are calculated, and the model parameters of the neural network prediction model are updated based on the gradient parameters.

[0015] As an improved solution, the motion state prediction model is constructed based on the optimized neural network prediction model, including: Determine the state quantity prediction equation and the model output equation based on the prediction equation of the optimized neural network prediction model; The state quantity prediction equation and the model output equation are combined to obtain the motion state prediction model.

[0016] As an improved solution, the calculating of the optimal control parameters based on the reference trajectory and the motion state prediction model includes: The optimization objectives are as follows: the position tracking deviation in three directions should be small, the deviations of heading angle, roll angle and pitch angle should be small, and the change rate of input control signal should be small; Constructing a target optimization function based on the optimization target; Setting state quantity constraints for controlling the driving stability and compatibility of tracked vehicles with complex working conditions; Set control quantity constraints on control smoothness and boundedness of drive torque assignment and rate of change; Inputting the reference trajectory into the motion state prediction model, calling the motion state prediction model to calculate the optimal control sequence based on the target optimization function, the state quantity constraint condition and the control quantity constraint condition; Based on the optimal output torque increment sequence of the optimal control sequence, the optimal control amount of the tracked vehicle at the first moment is determined; and the optimal control amount at the first moment is used as the optimal control parameter of the tracked vehicle at the first moment.

[0017] As an improved solution, the step of obtaining a reference trajectory of the tracked vehicle includes: The three-way position, vehicle heading angle, vehicle roll angle and vehicle pitch angle of the tracked vehicle are selected from the motion state parameter set as the trajectory tracking target quantity of the reference trajectory.

[0018] As an improved solution, the construction of a neural network prediction model driven by multivariate data further includes: Preprocessing the multivariate data, and using the preprocessed multivariate data as input of the neural network prediction model; Determining the number of nodes in the output layer of the neural network prediction model based on the multivariate data; Setting tanh as the activation function between the input layer and the hidden layer of the neural network prediction model; Setting Purelin as the activation function between the hidden layer and the output layer of the neural network prediction model; An empirical formula is used to determine the number of hidden layer nodes of the neural network prediction model.

[0019] As an improved solution, the step of calculating the gradient parameters of the neural network prediction model includes: Calculate the output layer weight gradient, output layer threshold gradient, hidden layer weight gradient and hidden layer threshold gradient; The output layer weight gradient, the output layer threshold gradient, the hidden layer weight gradient and the hidden layer threshold gradient are used as the gradient parameters.

[0020] On the other hand, the present invention also provides a neural network tracked vehicle motion state control system, comprising: Initialization module, used to: build a neural network prediction model driven by multivariate data; A training module, used to: optimize the neural network prediction model based on multivariate data; The model deployment module is used to: construct a motion state prediction model based on the optimized neural network prediction model, obtain a reference trajectory of the tracked vehicle, calculate optimal control parameters based on the reference trajectory and the motion state prediction model, perform motion control on the tracked vehicle based on the optimal control parameters, and return the optimal control parameters to the neural network prediction model to update the model parameters.

[0021] On the other hand, the present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method for controlling the motion state of a tracked vehicle using the neural network are implemented.

[0022] The beneficial effects of the technical solution of the present invention are: 1. The neural network tracked vehicle motion state control method described in the present invention can consider the ground pressure as a key parameter affecting the motion stability of the tracked vehicle on different terrains, and use the pressure value measured by the suspension unit pressure sensor as the input feature of the neural network, so that the model can more comprehensively obtain the impact of complex terrain changes on the vehicle motion state.

[0023] 2. The neural network tracked vehicle motion state control method described in the present invention can use a data-driven neural network model as a prediction model to solve the characteristics of traditional physical models such as uncertain parameters, poor adaptability and difficult analysis.

[0024] 3. The neural network tracked vehicle motion state control method described in the present invention can perform real-time prediction of the tracked vehicle motion state based on the BP neural network and online learning mechanism, and support real-time dynamic adjustment of the prediction model parameters, so that the prediction model can adapt to complex terrain and road surface changes, and realize stable trajectory tracking control of the tracked vehicle.

[0025] 4. The neural network tracked vehicle motion state control system described in the present invention can realize the neural network tracked vehicle motion state control method described in the present invention through the mutual cooperation of system modules.

[0026] 5. The computer-readable storage medium described in the present invention can cooperate with the guidance system module to realize the tracked vehicle motion state control method of the neural network described in the present invention, and the computer-readable storage medium described in the present invention also effectively improves the operability of the tracked vehicle motion state control method of the neural network. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0028] Figure 1 It is a flow chart of the method for controlling the motion state of a tracked vehicle using a neural network according to Embodiment 1 of the present invention; Figure 2 is a schematic diagram of the vehicle structure of the tracked vehicle in the method for controlling the motion state of a tracked vehicle using a neural network according to Embodiment 1 of the present invention; Figure 3 is a schematic diagram of the contact between the track of the tracked vehicle and the ground in the method for controlling the motion state of a tracked vehicle using a neural network according to Embodiment 1 of the present invention; Figure 4 It is a schematic diagram of the network structure of the BP neural network in the method for controlling the motion state of a tracked vehicle using a neural network in Example 1 of the present invention; Figure 5 It is a schematic diagram of the logic architecture of the method for controlling the motion state of a tracked vehicle using a neural network according to Embodiment 1 of the present invention; Figure 6 It is a schematic diagram of the architecture of the tracked vehicle motion state control system of the neural network described in Example 2 of the present invention; The markings in the accompanying drawings are as follows: 1. Inertial navigation; 2. GPS; 3. LiDAR; 4. Camera; 5. Suspension unit; 6. Pressure sensor. DETAILED DESCRIPTION

[0029] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby making a clearer and more definite definition of the protection scope of the present invention.

[0030] In the description of the present invention, it should be noted that the embodiments described in the present invention are only part of the embodiments of the present invention, rather than all of the embodiments; based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present invention.

[0031] The terms "first", "second", etc. in the specification and claims of this article and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of this article described here can be implemented in an order other than those illustrated or described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, device, product or equipment that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0032] Embodiment 1, this embodiment provides a method for controlling the motion state of a tracked vehicle using a neural network, such as Figure 1~Figure 5 As shown, the following steps are included: S100, build a neural network prediction model driven by multivariate data, including: S101, constructing a feedback state parameter set of the crawler vehicle, as follows: In this embodiment, a tracked vehicle with four road wheels on each side is selected as a specific example for explanation; a pressure sensor is set in the suspension device of each road wheel; the load of the suspension unit is measured by these sensors, and the load characteristics of each road wheel at each moment are accurately captured, thereby indirectly mapping the force between the road wheel and the ground; since there is an approximately linear relationship between the compression of the suspension system and the force borne by the road wheel, the suspension pressure data collected by the sensor can intuitively reflect the ground pressure distribution of the tracked vehicle at each road wheel position, and finally, the 8 suspension pressure values ​​are expressed as: ; It should be noted that the above pressure values ​​are integrated into the feedback state parameter set P, which is then input into the neural network model to predict the dynamic response of the vehicle under various working conditions, thereby providing strong data support for optimizing the vehicle control strategy; in view of the fact that the ground pressure of the tracked vehicle under complex terrain is affected by a variety of factors, it is difficult to accurately obtain the actual measured value, and considering the significant influence of the load distribution of each road wheel on the vehicle dynamic characteristics when the vehicle is driving, the above data is obtained for substitution. Although the suspension pressure and the road wheel ground pressure are not completely consistent in value, the conversion relationship between them is linear enough, and this relationship will be naturally captured and represented in the learning process of the neural network; It should also be noted that the tracked vehicle structure in this application is shown in the following figure: Figure 2 As shown, the laser radar is generally installed on the upper front side of the tracked vehicle and is used to provide point cloud information of the terrain in front; the camera is installed in front of the tracked vehicle and is used to collect road surface information as input for the subsequent neural network; the positioning system is installed at the center of mass of the tracked vehicle and is used to provide the spatial coordinate information of the vehicle at the current moment, including the three-dimensional position, three-dimensional turning angle and yaw angular velocity; the pressure sensor is installed on the suspension arm of the road wheel and is used to detect the load distribution information of the tracked vehicle's road wheel; in addition, the on-board sensor is also used to provide the motion state information of the tracked vehicle at each moment.

[0033] S102, constructing a pavement type parameter set for a tracked vehicle, as follows: In this embodiment, the camera equipped on the tracked vehicle is used to capture the road surface image in front of the vehicle in real time; at the same time, the multi-source information fusion technology is used to comprehensively analyze the collected information in combination with the data collected in the state detection unit and the suspension unit; in order to ensure the consistency of image data and dynamic data in time and space, the system uses timestamp synchronization technology to align data; the processed data is input into the optimized GCN (graph convolutional neural network); the network can deeply analyze the sensor data to estimate the key physical properties of the road surface such as flatness, bumpiness and slipperiness; these characteristics provide the network with sufficient information to identify more accurate road surface types; the road surface type recognition result is the output of the deep learning network, which provides valuable reference information for the intelligent control and navigation of the tracked vehicle. Finally, the road surface type parameter set is defined as follows: ; It should be noted that, for the above-mentioned set N of 10 road feature parameters, the discrete classification features therein are converted into binary vector form through One-hot encoding technology, and in subsequent steps are input into the BP neural network model together with other numerical features to further predict the motion state of the tracked vehicle; in view of the significant difference in resistance under different road conditions, in order to more accurately evaluate the kinematic performance of the tracked vehicle on complex terrain, the above steps are performed to obtain accurate identification results of the road type; furthermore, the high-resolution camera equipped in the present invention can clearly capture closer road detail information due to its excellent focusing ability and adaptability.

[0034] S103, constructing a motion state parameter set of the crawler vehicle, specifically as follows: In this embodiment, the three-dimensional position (x, y, z) and three-dimensional speed of the tracked vehicle are selected. , vehicle heading angle , Roll Angle , Pitch angle and the yaw rate , etc., together constitute the motion state parameter set X; these parameters are collected in real time by high-precision sensors, providing a solid data foundation for accurately predicting the motion state of tracked vehicles, which can be expressed as: ; In this embodiment, the driving torque on both sides of the tracked vehicle is used as the control amount of the trajectory tracking control rolling optimization and is recorded as: ; It should be noted that the motion state of tracked vehicles is complex and changeable, covering a variety of working conditions such as acceleration and deceleration, turning, climbing, and obstacle crossing. Especially when facing extremely complex terrains such as trenches, steep slopes, side slopes, and vertical walls, it is necessary to select parameters that can fully reflect the motion characteristics of tracked vehicles in various scenarios. Therefore, the vehicle's current motion state is also used as an important reference value for prediction.

[0035] S104, parameter set integration, as follows: In this embodiment, since the prediction of the motion state of the tracked vehicle is a process affected by multiple factors, P, N, and X determined in the above steps constitute the key elements of the prediction; in order to accurately predict the motion state at the next moment, P, N, and X are combined with the tracked vehicle driving torque U optimized moment by moment in the predictive control model, and together serve as the input information of the neural network.

[0036] S105, determine the BP neural network prediction model, as follows: In this embodiment, a BP neural network is used to construct a prediction model; specifically, a three-layer BP neural network including a single hidden layer (where the input layer is regarded as the 0th layer, the hidden layer is the 1st layer, and the output layer is the 2nd layer) can effectively approximate any nonlinear function when the number of hidden layer nodes is large enough; therefore, in this embodiment, the above three-layer BP neural network is selected, and its prediction model is expressed as: = ; It should be noted that, considering the high-frequency sampling rate of the sensor, the motion characteristics of the tracked vehicle and the computing power of the model, 0.1 seconds is set as the prediction time step, which means that the BP neural network model will use the state variables at the current moment as input to predict the motion state in the next 0.1 seconds; this design is intended to ensure that the prediction is both forward-looking and can make full use of real-time data to improve the accuracy and practicality of the prediction.

[0037] S106, neural network pre-configuration, specifically as follows: (1) The above multidimensional variable data is preprocessed and the preprocessed data is used as the input of the BP neural network, which is expressed as follows: ; Among them, since there are 21 input variables in the input layer, it is determined that the input layer has 21 nodes; (2) The motion state vector of the tracked vehicle at the next moment output by the neural network prediction model is expressed as: ; Based on this, the number of nodes in the output layer is set to 10, expressed as: ; Among them, the 10 nodes are the horizontal positions at time t+1 , vertical position , vertical position , heading angle α, roll angle β, pitch angle , speed in three directions and the yaw rate ; (3) In the connection from the input layer to the hidden layer, tanh is set as the activation function to perform a smooth nonlinear mapping on the input, specifically expressed as tanh( )= ; Among them, tanh( ) is the output of the tanh activation function, which is recorded as is the input value of the hidden layer; (4) A linear activation function Purelin is set between the hidden layer and the output layer. During the driving process of the tracked vehicle, the state prediction value of its output often shows an approximate linear change trend in a short period of time. The use of the linear activation function Purelin can accurately capture and maintain this linear relationship, ensuring the accuracy of the output prediction results. It is specifically expressed as purelin ( )= ; Among them, purelin ( ) is the output of the purelin activation function, which is recorded as is the input value of the output layer; (5) An empirical formula is used to determine the number of hidden layer nodes of the BP neural network. This method fully considers the impact of the network structure on the prediction performance and aims to find a hidden layer node number configuration that can effectively capture data characteristics and avoid overfitting or underfitting. The specific empirical formula is: λ; where m is the number of input layer nodes, n is the number of output layer nodes, is the empirical adjustment constant, ∈[1,10]; (6) Given that the real-time driving environment of the tracked vehicle is a dynamic online learning scenario and lacks a comprehensive and static data set, 8 hidden layer nodes are selected as the starting point in the initial construction of the model. After subsequent experimental verification, this number of nodes has been proven to ensure that the model achieves the optimal state in terms of prediction performance and generalization ability. The final training parameter table is as follows: layer Output size Parameters Input Layer 21 0 Hidden Layer 8 176 Output Layer 10 90 S200, optimizing / training the neural network prediction model based on multivariate data, including: S201, data normalization processing, specifically as follows: In this embodiment, an experience-based maximum and minimum value setting strategy is used to normalize the input and output, wherein an initial maximum and minimum value are set for each input feature according to the physical properties of the crawler vehicle state parameters, the actual operating range, and the accumulated experience data. The formula is expressed as follows: =2 ; Among them, x is the original data, is the normalized data, is the minimum value, is the maximum value; It should be noted that since the present invention runs in online learning mode, it is impossible to rely on a pre-set data set to directly calculate the values ​​required for normalization, so an experience-based maximum and minimum value setting strategy is adopted; this strategy is in line with the actual operation of the tracked vehicle and can ensure the smooth progress of the normalization process, laying a solid foundation for subsequent model training and prediction.

[0038] S202, parameter initialization, as follows: ; in, is the weight of the l-th layer neural network from the h-th neuron to the k-th neuron; , are random variables with values ​​ranging from [-1,1]; H is the number of neurons in the l-1 layer; K is the number of neurons in the l layer; is the bias of the kth neuron in the lth layer; G is the scaling factor for adjusting the weight and bias range; Among them, after initialization, the weight from the input layer to the hidden layer and bias And the weights from the hidden layer to the output layer and bias , respectively as follows: ; ; Among them, the hidden layer input is: ; The hidden layer output is: ; Among them, the output layer input is: ; The output of the output layer is: .

[0039] S203, setting the loss function, as follows: In this embodiment, a function based on mean square error to measure the deviation between the predicted value and the real data at each moment is used as the loss function, and the function formula is: ; Where n is the number of output variables; is the predicted value of the cth output; is the true value of the cth output.

[0040] S204, gradient calculation, specifically as follows: (1) Calculate the output layer weight gradient: First, ; Since there is only the jth output About So for other outputs: ; so the original form ;in, .

[0041] (2) Calculate the output layer threshold gradient: For the threshold (jth output node threshold), according to the above formula Make the following changes: .

[0042] (3) Calculate the hidden layer weight gradient: ; Since there is only the jth About function, so the above formula is transformed into: ; Based on the derivative of the composite function, the original formula is = Formula ②; Among them, formula ②= ;Depend on have to: .

[0043] (4) Calculate the hidden layer threshold gradient: Based on the hidden layer weight gradient solution logic, and according to the formula have:

[0044] .

[0045] S205: Update the model parameters based on the above gradients, as follows: ; in, The learning rate is fixed at 0.01; is the old value of the weight from neuron j to i in the weight matrix at the previous moment; is the old value of the bias of neuron i; , are the gradient values ​​of the aforementioned loss functions with respect to each parameter; It should be noted that every time the weights and biases of the BP neural network are updated, the BP neural network immediately uses the latest parameters to predict the state at the next moment; in this process, each collection and processing of new data will activate a model prediction and self-update; as long as the tracked vehicle is in a state of continuous motion, the system will continuously receive new state data, thereby driving the model parameters to be continuously adjusted; ultimately, the BP neural network forms a closed-loop prediction process: "real-time data collection → state prediction → error calculation → parameter update"; this process ensures that the model can keep up with the dynamic changes of the tracked vehicle and achieve accurate real-time predictions.

[0046] S300, constructing a motion state prediction model based on the optimized neural network prediction model, obtaining a reference trajectory of a tracked vehicle, calculating optimal control parameters based on the reference trajectory and the motion state prediction model, performing motion control on the tracked vehicle based on the optimal control parameters, and returning the optimal control parameters to the neural network prediction model to update model parameters, including: S301, receiving a reference trajectory of a navigation system, as follows: In this embodiment, considering the trajectory tracking control requirements of the crawler vehicle traveling on complex terrains such as trenches and ramps, x, y, z, α, β, and γ are selected from the motion state variable set X as the target quantities for trajectory tracking, which are expressed as: .

[0047] S302: Construct a prediction model, as follows: (1) Based on the aforementioned data-driven neural network model as a prediction model, the state prediction of multiple time steps k is considered at time t. The prediction equation in the predictive control model is: ; (2) Define state quantity , control amount , approximate linearization of the formula; (3) In order to achieve smooth processing of the control signal, the change in the control input needs to be Optimize and take into account the cumulative impact of all control input changes on the system state in the future prediction time domain, and finally derive the state quantity prediction equation as follows: ; Where A is the state transfer matrix, B is the control input matrix, H is the sum of external disturbance and constant term, To control the time domain; (4) The model output equation of the predictive control model is: ; (5) The state quantity prediction equation and the output equation of the control model are combined to obtain the final prediction model, which is expressed as: ; (6) Based on the recursive relationship between the system state and system output in the future time domain of the above prediction model, the system prediction output model is obtained through comprehensive analysis as follows: ;in, is the coefficient matrix of the state quantity, is the coefficient matrix of the input quantity; then, the output quantity at time t is ; Change of control quantity at time t ; Reference value in each prediction time domain ; S303, setting a multi-objective optimization function, as follows: In this embodiment, in order to enable the tracked vehicle to stably track the reference trajectory, three optimization indicators and optimization directions are set, specifically: (1) The three-way position tracking deviation should be small; (2) The deviations of heading, roll and pitch angles should be small; (3) The change rate of the input control signal should be small; In this embodiment, the objective optimization function is: ;in, is the prediction time domain of the system, To track the weight coefficient, is the relaxation factor, , are the weight matrices, is the output of the predictive control model, is the reference trajectory output by the controller, To control the increment.

[0048] S304: Set constraints, as follows: In this embodiment, considering that the tracked vehicle needs to travel in complex terrain and variable working conditions, and at the same time needs to meet physical limitations and ensure control stability, the following constraints are set for the state quantity: ; In this embodiment, based on the smoothness of control and the bounded characteristics of the amplitude and change rate of the driving torque, the constraints of the control amount are set as follows: .

[0049] S305, calculate the optimal control amount, as follows: In this embodiment, the above multi-objective optimization function and various constraints are combined to calculate the optimal control value, and the optimal control sequence is obtained as follows: ; In this embodiment, the optimal output torque increment sequence at time t is calculated, and the first item thereof is added to the control amount at time t-1 to obtain the optimal control amount at time t: ; In this embodiment, the optimal control quantity at time t is passed to the tracked vehicle and used as the input of the neural network model at the current moment for rolling prediction of the motion state in the next prediction time domain; at this point, the control flow of a motion control cycle of the tracked vehicle ends, and the tracked vehicle moves under the control. After that, the method returns to step S301 to start the control process of the next cycle, and finally performs cyclic control based on the above process until the tracked vehicle successfully completes the tracking task.

[0050] In summary, the present application integrates an online learning mechanism, by inputting the multi-dimensional state information of the tracked vehicle at time k into the BP neural network, and based on the forward propagation characteristics of the neural network, predicting the motion state variables of the tracked vehicle at time k+1; when the vehicle travels to time k+1, the error between the state information fed back by the sensor in real time and the previously predicted state information is calculated and compared based on the present method; the weights and biases of the neural network are dynamically adjusted through the gradient descent algorithm, so as to give the model the ability to adapt to complex and changeable terrain; then, the tracked vehicle is precisely controlled according to the motion state predicted by the present method, and a neural network is used to replace the traditional dynamic equation as a prediction model. In each prediction time domain, according to the predicted state of the tracked vehicle, the objective function is constructed by comprehensively considering the reference trajectory and the physical constraints of the system, and the optimal control sequence is solved, and the driving torque of the tracked vehicle is continuously optimized and adjusted to achieve stable trajectory tracking control of the tracked vehicle in complex terrain.

[0051] It should be noted that the above examples are only for explaining the present invention and cannot limit the protection scope of the present invention.

[0052] Embodiment 2, based on the same inventive concept as the method for controlling the motion state of a tracked vehicle using a neural network described in Embodiment 1, this embodiment provides a control system for the motion state of a tracked vehicle using a neural network, such as Figure 6 As shown, including: Initialization module, used to: build a neural network prediction model driven by multivariate data; A training module, used to: optimize the neural network prediction model based on multivariate data; The model deployment module is used to: construct a motion state prediction model based on the optimized neural network prediction model, obtain a reference trajectory of the tracked vehicle, calculate optimal control parameters based on the reference trajectory and the motion state prediction model, perform motion control on the tracked vehicle based on the optimal control parameters, and return the optimal control parameters to the neural network prediction model to update the model parameters.

[0053] Embodiment 3: This embodiment provides a computer-readable storage medium, including: The storage medium is used to store computer software instructions used to implement the neural network tracked vehicle motion state control method described in the above-mentioned embodiment 1, which includes a program for executing the above-mentioned program set for the neural network tracked vehicle motion state control method; specifically, the executable program can be built into the neural network tracked vehicle motion state control system described in embodiment 2, so that the neural network tracked vehicle motion state control system can implement the neural network tracked vehicle motion state control method described in embodiment 1 by executing the built-in executable program.

[0054] In addition, the computer-readable storage medium of this embodiment may adopt any combination of one or more computer-readable storage media, wherein the computer-readable storage medium includes electrical, optical, electromagnetic, infrared or semiconductor systems, devices or components, or any combination thereof.

[0055] Different from the prior art, the method, system and medium for controlling the motion state of a tracked vehicle using a neural network of the present application can use a data-driven neural network model as a prediction model to solve the characteristics of uncertainty in parameters, poor adaptability and difficulty in analysis of traditional physical models. Based on an online learning mechanism, real-time prediction of the motion state of the tracked vehicle and real-time dynamic adjustment of the prediction model parameters are performed, so that the prediction model can adapt to complex terrain and road surface changes, thereby achieving stable trajectory tracking control of the tracked vehicle.

[0056] It should be understood that in the various embodiments of this document, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this document.

[0057] It should also be understood that in the embodiments of this article, the term "and / or" is only a description of the association relationship of the associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0058] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this article.

[0059] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0060] In the several embodiments provided herein, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, or can be electrical, mechanical or other forms of connection.

[0061] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiments of this article.

[0062] In addition, each functional unit in each embodiment of this invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above integrated unit may be implemented in the form of hardware or in the form of software functional unit.

[0063] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this article is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of this article. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0064] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for controlling the motion state of a tracked vehicle based on a neural network, characterized in that: The following steps are involved: Construct a neural network prediction model driven by multivariate data; Performing model optimization on the neural network prediction model based on multivariate data; A motion state prediction model is constructed based on the optimized neural network prediction model, a reference trajectory of the tracked vehicle is obtained, optimal control parameters are calculated based on the reference trajectory and the motion state prediction model, motion control of the tracked vehicle is performed based on the optimal control parameters, and the optimal control parameters are returned to the neural network prediction model to update the model parameters.

2. The method for controlling the motion state of a tracked vehicle based on a neural network according to claim 1, characterized in that: The construction of a neural network prediction model driven by multivariate data includes: A three-layer BP neural network including a single hidden layer is used as the neural network prediction model; collecting the suspension pressure value of the tracked vehicle as a feedback state parameter set; collecting road surface type information of the tracked vehicle as a road surface type parameter set; Collecting the three-dimensional position, three-dimensional speed, body heading angle, body roll angle, body pitch angle and body yaw angular velocity of the tracked vehicle as a motion state parameter set; collecting driving torques on both sides of the tracked vehicle as a torque parameter set; The feedback state parameter set, the road surface type parameter set, the motion state parameter set and the torque parameter set are used as the multivariate data for training the neural network prediction model.

3. The method for controlling the motion state of a tracked vehicle based on a neural network according to claim 1, characterized in that: The method of optimizing the neural network prediction model based on multivariate data includes: performing normalization processing on the multivariate data; Initializing weight parameters and bias parameters of the neural network prediction model; A function that measures the deviation between the predicted value and the real data at each moment based on the mean square error is used as a loss function, the normalized multivariate data is input into the initialized neural network prediction model, and the neural network prediction model is trained based on the loss function; The gradient parameters of the neural network prediction model are calculated, and the model parameters of the neural network prediction model are updated based on the gradient parameters.

4. The method for controlling the motion state of a tracked vehicle based on a neural network according to claim 1, characterized in that: The step of constructing a motion state prediction model based on the optimized neural network prediction model includes: Determine the state quantity prediction equation and the model output equation based on the prediction equation of the optimized neural network prediction model; The state quantity prediction equation and the model output equation are combined to obtain the motion state prediction model.

5. The method for controlling the motion state of a tracked vehicle based on a neural network according to claim 1, characterized in that: The calculating of the optimal control parameters based on the reference trajectory and the motion state prediction model comprises: The optimization objectives are as follows: the position tracking deviation in three directions should be small, the deviations of heading angle, roll angle and pitch angle should be small, and the change rate of input control signal should be small; Constructing a target optimization function based on the optimization target; Setting state quantity constraints for controlling the driving stability and compatibility of tracked vehicles with complex working conditions; Set control quantity constraints on control smoothness and boundedness of drive torque assignment and rate of change; Inputting the reference trajectory into the motion state prediction model, calling the motion state prediction model to calculate the optimal control sequence based on the target optimization function, the state quantity constraint condition and the control quantity constraint condition; Based on the optimal output torque increment sequence of the optimal control sequence, the optimal control amount of the tracked vehicle at the first moment is determined; and the optimal control amount at the first moment is used as the optimal control parameter of the tracked vehicle at the first moment.

6. The method for controlling the motion state of a tracked vehicle based on a neural network according to claim 2, characterized in that: The step of obtaining a reference trajectory of the tracked vehicle comprises: The three-way position, vehicle heading angle, vehicle roll angle and vehicle pitch angle of the tracked vehicle are selected from the motion state parameter set as the trajectory tracking target quantity of the reference trajectory.

7. The method for controlling the motion state of a tracked vehicle based on a neural network according to claim 2, characterized in that: The construction of a neural network prediction model driven by multivariate data further includes: Preprocessing the multivariate data, and using the preprocessed multivariate data as input of the neural network prediction model; Determining the number of nodes in the output layer of the neural network prediction model based on the multivariate data; Setting tanh as the activation function between the input layer and the hidden layer of the neural network prediction model; Setting Purelin as the activation function between the hidden layer and the output layer of the neural network prediction model; An empirical formula is used to determine the number of hidden layer nodes of the neural network prediction model.

8. The method for controlling the motion state of a tracked vehicle based on a neural network according to claim 3, characterized in that: The step of calculating the gradient parameters of the neural network prediction model comprises: Calculate the output layer weight gradient, output layer threshold gradient, hidden layer weight gradient and hidden layer threshold gradient; The output layer weight gradient, the output layer threshold gradient, the hidden layer weight gradient and the hidden layer threshold gradient are used as the gradient parameters.

9. A tracked vehicle motion state control system based on a neural network, characterized in that: include: Initialization module, used to: build a neural network prediction model driven by multivariate data; A training module, used to: optimize the neural network prediction model based on multivariate data; The model deployment module is used to: construct a motion state prediction model based on the optimized neural network prediction model, obtain a reference trajectory of the tracked vehicle, calculate optimal control parameters based on the reference trajectory and the motion state prediction model, perform motion control on the tracked vehicle based on the optimal control parameters, and return the optimal control parameters to the neural network prediction model to update the model parameters.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the neural network-based tracked vehicle motion state control method described in any one of claims 1 to 8 are implemented.

Images