Leveling and stability evaluation method and system for self-adaptive terrain of truck-mounted crane supporting leg

Abstract

The invention provides a terrain-adaptive leveling and stability evaluation method and system for truck-mounted crane supporting legs, and relates to the technical field of automatic control, and the method comprises the steps: building a ground rigidity-deformation mapping relation by using data collected by a sensor, constructing a nonlinear spring damping contact force model, and obtaining supporting leg load distribution data; obtaining a stability state matrix based on the supporting leg load distribution data; the stability state matrix is input into a model prediction controller, the model prediction controller adopts a reinforcement learning method to construct a terrain adaptability evaluation function, and a landing leg leveling control instruction is generated through an action value network; the landing leg leveling control instruction is input into a distributed cooperative control unit, the distributed cooperative control unit sets landing legs as network nodes and carries out information interaction, and a feedforward feedback composite controller with self-adaptive gain is adopted to control the displacement of the landing legs.

Description

Technical Field

[0001] The present invention relates to automatic control technology, and particularly to a method and system for leveling and stability evaluation of the outriggers of a truck-mounted crane adapting to terrain. Background Art

[0002] As an important lifting equipment, a truck-mounted crane is widely used in operations in various complex terrain environments. The leveling and stability of its outriggers are directly related to the safety and efficiency of the lifting operation. The defects and deficiencies of the existing technologies are mainly reflected in the following three aspects:

[0003] Firstly, the traditional outrigger leveling method lacks the adaptability to complex terrain. Facing uneven ground, the existing methods are difficult to accurately control the telescopic amount of the outriggers, resulting in a large deviation in the vehicle body attitude, affecting the operation stability, and even causing tipping accidents.

[0004] Secondly, the existing outrigger ground contact modeling is too simplified and difficult to accurately reflect the interaction force between the outrigger and the ground. Ignoring the non-linear relationship between ground stiffness and deformation, as well as the non-uniformity of the outrigger contact surface, leads to a deviation between the calculated outrigger load distribution and the actual situation, affecting the accuracy of the stability evaluation.

[0005] Finally, the existing stability evaluation methods are too simple and lack a comprehensive consideration of dynamic loads and stress distributions. Relying solely on simple geometric analysis and empirical formulas, it is difficult to effectively predict and warn of potential dangers, resulting in a high operation risk. Summary of the Invention

[0006] The embodiments of the present invention provide a method and system for leveling and stability evaluation of the outriggers of a truck-mounted crane adapting to terrain, which can solve the problems in the existing technologies.

[0007] In the first aspect of the embodiments of the present invention,

[0008] A method for leveling and stability evaluation of the outriggers of a truck-mounted crane adapting to terrain is provided, including:

[0009] Filtering and weight-fusing the terrain elevation data, outrigger pressure data, and vehicle body attitude data collected by sensors, and then inputting them into an outrigger ground contact modeling unit. The outrigger ground contact modeling unit classifies terrain features using a historical evolution predictor based on a deep neural network and establishes a ground stiffness-deformation mapping relationship, and constructs a non-linear spring-damping contact force model using an online learning contact parameter adaptor to obtain outrigger load distribution data;

[0010] Input the outrigger load distribution data into the stability evaluation unit of the outrigger. The stability evaluation unit includes a stability discriminator, a stress uniformity evaluator, and a dynamic load early warning device. The stability discriminator forms a support polygon based on the outrigger landing points and calculates the support stability margin in combination with the projection position of the vehicle body's center of gravity. The stress uniformity evaluator calculates the stress uniformity index based on the outrigger pressure variance. The dynamic load early warning device analyzes the dynamic characteristics of the outrigger pressure and outputs an early warning signal. Weightedly fuse the outputs of the stability discriminator, the stress uniformity evaluator, and the dynamic load early warning device through an adaptive weight coefficient that adjusts in real time according to the vehicle body attitude to obtain a stability state matrix.

[0011] Input the stability state matrix into the model predictive controller. The model predictive controller constructs a terrain adaptability evaluation function using the reinforcement learning method and generates an outrigger leveling control instruction through the action value network. Input the outrigger leveling control instruction into the distributed cooperative control unit. The distributed cooperative control unit sets the outriggers as network nodes for information interaction and controls the outrigger displacement using a feedforward-feedback composite controller with an adaptive gain.

[0012] In an optional implementation manner,

[0013] The outrigger ground contact modeling unit classifies terrain features using a historical evolution predictor based on a deep neural network and establishes a ground stiffness-deformation mapping relationship. The steps of obtaining the outrigger load distribution data by using an online learning contact parameter adaptor to construct a non-linear spring-damping contact force model include:

[0014] Classify the terrain features using a historical evolution predictor of a double-branch densely connected convolutional neural network. The double-branch densely connected convolutional neural network includes a high-frequency feature extraction branch for extracting local texture features and a low-frequency feature extraction branch for extracting regional morphological features. Each feature extraction branch is provided with an independent spatial pyramid pooling module, and the terrain classification result is output after adaptively fusing the double-branch features through a cascaded attention module.

[0015] Establish a hybrid ground mechanics model based on the terrain classification result, use a three-layer backpropagation neural network to establish a mapping relationship between the ground stiffness and the deformation amount to obtain the ground deformation characteristics, and input the ground deformation characteristics, the outrigger pressure sequence, the displacement sequence, and the terrain classification result into a double-layer long short-term memory network for soil mechanics parameter identification. The first layer of the double-layer long short-term memory network outputs a preliminary estimate of the soil parameters, and the second layer introduces historical data to perform time-series correction on the preliminary estimate of the soil parameters.

[0016] Construct a non-linear spring-damping contact force model for multi-physical field coupling according to soil mechanics parameters. The non-linear spring-damping contact force model includes a stiffness function related to the ground moisture content, a damping function related to temperature, and a friction coefficient function related to relative velocity and pressure. Use a parameter adaptor based on an action evaluation framework to evaluate the contact state between the outrigger and the ground and online optimize the parameters of the stiffness function, damping function, and friction coefficient function;

[0017] Establish a dynamic coupling equation for the outrigger group based on the non-linear spring-damping contact force model. Under the constraint of the dynamic coupling equation for the outrigger group, use a load distributor with a prediction-correction structure for optimal distribution. Extract the outrigger force signal characteristics through wavelet decomposition, and combine an adaptive notch filter and an integral sliding mode controller to compensate for high-frequency vibration and low-frequency drift, and output the outrigger load distribution data.

[0018] In an optional implementation manner,

[0019] The stress uniformity evaluator calculates the stress uniformity index based on the outrigger pressure variance, and the dynamic load early warning device outputs an early warning signal based on the dynamic characteristics analysis of the outrigger pressure; the steps of weighted fusion of the outputs of the stability discriminator, stress uniformity evaluator, and dynamic load early warning device through an adaptive weight coefficient that adjusts in real time with the vehicle body attitude to obtain the stability state matrix include:

[0020] Calculate the outrigger pressure variance based on the outrigger load distribution data, introduce a terrain correction coefficient related to the terrain slope and terrain undulation to correct the outrigger pressure variance, and combine the corrected outrigger pressure variance with the maximum pressure to calculate the stress uniformity index;

[0021] Extract multi-scale dynamic characteristics of the outrigger pressure sequence, extract dynamic characteristics through orthogonal complement decomposition and variational mode decomposition, construct a time-frequency joint characteristic index based on a coupling function, and trigger a dynamic load early warning signal when the time-frequency joint characteristic index exceeds an adaptive early warning threshold, where the adaptive early warning threshold dynamically adjusts with changes in the terrain slope and the slewing speed of the lifting mechanism;

[0022] Construct a weight mapping function based on the vehicle body attitude angle, height, and slewing speed of the lifting mechanism to perform initial weight allocation for the support stability margin, stress uniformity index, and dynamic load early warning signal. Calculate the change rates of the weights for the vehicle body attitude angle, height, and slewing speed of the lifting mechanism according to the weight mapping function to obtain a state sensitivity matrix. Dynamically update the weight coefficients based on the state sensitivity matrix, vehicle body state parameter increments, and newly calculated weight values, and perform weight optimization under the constraint conditions that the sum of the weight coefficients is one and the difference between the weight coefficients at adjacent times does not exceed a preset threshold. Weight the optimized weights with the support stability margin, stress uniformity index, and dynamic load early warning signal to obtain the stability state matrix.

[0023] In an alternative embodiment,

[0024] The steps of performing multi-scale dynamic feature extraction on the outrigger pressure sequence, extracting dynamic features through empirical mode decomposition and variational mode decomposition, constructing a time-frequency joint feature index based on a coupling function, and triggering a dynamic load warning signal when the time-frequency joint feature index exceeds an adaptive warning threshold, where the adaptive warning threshold is dynamically adjusted according to the changes in the terrain slope and the slewing speed of the lifting mechanism are as follows:

[0025] Obtain the outrigger pressure signal, outrigger deformation amount, and ground reaction force signal, construct an interaction dynamics feature matrix of the outrigger and the ground from the outrigger pressure signal, outrigger deformation amount, and ground reaction force signal, calculate transient components based on the interaction dynamics feature matrix using the empirical mode decomposition algorithm, and extract dynamic feature vectors based on the transient components;

[0026] Perform improved variational mode decomposition on the transient components, where an adaptive window function dynamically adjusted according to the local characteristics of the signal is introduced during the variational mode decomposition process to obtain multiple intrinsic mode components and their corresponding central frequencies;

[0027] Construct the joint distribution of the dynamic feature vectors, intrinsic mode components, and central frequencies based on the coupling function, where the generating function of the coupling function is dynamically selected according to feature correlation, and the time-frequency joint feature index is obtained through integral operation on the joint distribution;

[0028] Construct a multi-objective optimization function including the time-frequency joint feature index, terrain slope, and slewing speed of the lifting mechanism, perform mean processing on the time-frequency joint feature index and construct an accumulation sequence, establish a grey differential equation model based on the accumulation sequence, construct a background value sequence and establish a parameter estimation equation, solve the parameter estimation equation by the least squares method to obtain a time response function, perform cumulative reduction to obtain the predicted value of the time-frequency joint feature index, and perform online update on the grey differential equation model through a sliding time window; construct a fuzzy rule base based on the terrain slope, slewing speed of the lifting mechanism, and predicted time-frequency joint feature index, obtain a threshold adjustment amount through fuzzy inference, and adaptively update the warning threshold by dynamically adjusting the learning rate according to the prediction error; when the time-frequency joint feature index exceeds the warning threshold after adaptive update, trigger an outrigger dynamic load warning signal.

[0029] In an alternative embodiment,

[0030] The steps of constructing a terrain adaptability evaluation function for the model predictive controller using a reinforcement learning method and generating an outrigger leveling control command through an action value network include:

[0031] Construct a terrain adaptability evaluation function that includes outrigger height deviation, roll angle deviation, pitch angle deviation, ground contact force, and outrigger slip speed. The terrain adaptability evaluation function adopts an adaptive weight allocation strategy, which is dynamically adjusted based on terrain roughness, slope standard deviation, ground friction coefficient, and ground stress distribution standard deviation. During the dynamic adjustment of weights, a Sigmoid function with a periodic compensation term is used, and the ground stress distribution is corrected by a Gaussian kernel function;

[0032] Design an action value network structure that includes an evaluation network and a target network. The state spaces of the evaluation network and the target network include outrigger height vectors, vehicle body attitude angles, terrain feature vectors, and ground stress distribution features, and the action space is the outrigger speed command; The evaluation network uses a multi-head self-attention mechanism and a residual convolution structure for feature extraction, and calculates the value estimate of the state-action pair based on the output of the terrain adaptability evaluation function; In the model predictive controller, train and optimize the evaluation network, and softly update the parameters of the evaluation network to the target network according to the training error through an adaptive update rate;

[0033] Establish an optimization objective function that includes a short-term control cost term, a control quantity adjustment term, and a terminal penalty term. The short-term control cost term is provided by the evaluation network, the terminal penalty term is calculated from the long-term value estimate provided by the target network, and the control quantity adjustment term includes the square term of the control increment between adjacent moments;

[0034] Form an optimization problem by combining the optimization objective function with outrigger kinematic constraints, actuator limit constraints, and control increment constraints, and use the alternating direction multiplier method based on Nesterov acceleration to solve the optimization problem. The alternating direction multiplier method decomposes the original problem into a sequence of sub-problems through dual decomposition, and the conjugate gradient method is used to solve the sub-problems; Output the solution result of the optimization problem as the outrigger leveling control command.

[0035] In an alternative embodiment,

[0036] The step of training and optimizing the evaluation network in the model predictive controller and softly updating the parameters of the evaluation network to the target network according to the training error through an adaptive update rate includes:

[0037] Construct a distributed experience replay pool. The distributed experience replay pool adopts a hierarchical storage structure, calculates an environment complexity index based on terrain roughness and slope standard deviation, calculates a task difficulty index based on outrigger height error vectors, roll angle errors, and pitch angle errors, and stores state-action-reward value-next state transition samples hierarchically according to the weighted combination of the environment complexity index and the task difficulty index;

[0038] Design a hybrid priority sampling mechanism based on the distributed experience replay pool. The hybrid priority sampling mechanism includes priority calculation based on the Wasserstein distance and importance sampling weight calculation based on the ratio of control decision functions. Dynamically adjust the combined weight coefficient of the priority calculation and the importance sampling weight calculation according to the training stage, and adjust the sampling probability of samples with different difficulties according to the combined weight coefficient;

[0039] Construct a dual-gradient optimizer. The dual-gradient optimizer includes a control decision gradient optimizer and an evaluation network gradient optimizer. The control decision gradient optimizer uses the proximal policy optimization algorithm with trust region constraints to calculate the control decision update amount based on the ratio of the current control decision function and the historical control decision function. The evaluation network gradient optimizer uses the adaptive moment estimation algorithm with momentum term to construct a loss function based on the prediction error and gradient norm of the evaluation network;

[0040] Design a multi-scale adaptive learning mechanism based on the optimization results of the dual-gradient optimizer. Jointly adjust the base learning rate according to the temporal difference error, the variance of the control decision update amount, and the loss value of the evaluation network, and adaptively adjust the exploration noise amplitude based on the statistical distribution characteristics of the outrigger speed command output;

[0041] Adopt a hierarchical network update strategy, and design different update frequencies for the feature extraction layer and the control instruction output layer of the evaluation network respectively. The feature extraction layer updates its parameters according to a preset first update period, and the control instruction output layer updates its parameters according to a preset second update period, where the first update period is greater than the second update period. Soft-update the parameters of the evaluation network to the target network after weighted averaging according to the verification performance.

[0042] In an optional implementation manner,

[0043] The steps of the distributed cooperative control unit setting the outriggers as network nodes for information interaction and using a feedforward-feedback composite controller with adaptive gain to control the outrigger displacement include:

[0044] Construct a distributed cooperative control unit network. Each outrigger corresponds to a distributed cooperative control unit as a network node. The state vector of the network node includes outrigger height, roll angle, pitch angle, speed, and contact force information. Based on the state vector of the network node, construct the outrigger displacement error and the error change rate, and adaptively adjust the network communication radius, bandwidth parameters, and state component weights according to the outrigger displacement error and the error change rate to obtain the dynamic connection weight and the weighted state vector;

[0045] Based on the dynamic connection weight and the weighted state vector, an exponential function of the leg displacement error is calculated to generate an adaptive trigger function. When the leg displacement error is greater than the adaptive trigger function, information interaction is triggered. The leg displacement error is used to calculate the network node confidence. Based on the network node confidence, a time-varying information fusion weight is designed. A time-varying consistency protocol considering communication delay is constructed, and an adaptive update law of the protocol gain matrix is ​​designed to output a consistency control amount.

[0046] A feedforward feedback composite controller is constructed based on the consistency control amount, wherein the feedforward feedback composite controller includes a disturbance observer and a neural network compensator, wherein the disturbance observer estimates the disturbance amount in the outrigger motion process based on the feedforward control term and the outrigger displacement error, and the neural network compensator obtains the motion compensation control value by online learning the weighted state vector, and the disturbance amount, the motion compensation control value and the consistency control amount are combined to construct a composite control rate;

[0047] A predictive control and fault diagnosis mechanism is designed based on the composite control rate and the weighted state vector, a long short-term memory network is used to predict the state of a network node, a prediction error compensation term is designed based on the error between the predicted state of the network node and the actual state, a fault index is calculated by the residual between the weighted state vector and the predicted state of the network node, a fault is determined to have occurred when the fault index is greater than a preset threshold, and when a fault occurs, the control rate is recalculated based on the state vector of the neighboring network node and the control amount of the faulty leg is compensated.

[0048] According to a second aspect of the embodiments of the present invention,

[0049] Provides a self-adaptive terrain leveling and stability assessment system for the crane outriggers, including:

[0050] The first unit is used to filter and weight the terrain elevation data, outrigger pressure data and vehicle body posture data collected by the sensor and then input them into the outrigger ground contact modeling unit, wherein the outrigger ground contact modeling unit uses a historical evolution predictor based on a deep neural network to classify terrain features and establish a ground stiffness-deformation mapping relationship, and uses an online learning contact parameter adaptor to construct a nonlinear spring damping contact force model to obtain outrigger load distribution data;

[0051] The second unit is used to input the outrigger load distribution data into the stability evaluation unit of the outriggers. The stability evaluation unit includes a stability discriminator, a stress uniformity evaluator, and a dynamic load early warning device. The stability discriminator forms a support polygon based on the outrigger landing points and calculates the support stability margin in combination with the projection position of the vehicle body center of gravity. The stress uniformity evaluator calculates the stress uniformity index based on the outrigger pressure variance. The dynamic load early warning device analyzes the dynamic characteristics of the outrigger pressure and outputs an early warning signal. The outputs of the stability discriminator, the stress uniformity evaluator, and the dynamic load early warning device are weighted and fused through an adaptive weight coefficient that adjusts in real time according to the vehicle body attitude to obtain a stability state matrix.

[0052] The third unit is used to input the stability state matrix into a model predictive controller. The model predictive controller constructs a terrain adaptability evaluation function using a reinforcement learning method and generates an outrigger leveling control instruction through an action value network. The outrigger leveling control instruction is input into a distributed cooperative control unit. The distributed cooperative control unit sets the outriggers as network nodes for information interaction and controls the outrigger displacement using a feedforward-feedback composite controller with an adaptive gain.

[0053] In the third aspect of the embodiments of the present invention,

[0054] There is provided an electronic device, including:

[0055] A processor;

[0056] A memory for storing instructions executable by the processor;

[0057] Wherein, the processor is configured to call the instructions stored in the memory to execute the method described above.

[0058] In the fourth aspect of the embodiments of the present invention,

[0059] There is provided a computer-readable storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the method described above is implemented.

[0060] Through precise modeling of the terrain and in combination with outrigger load distribution and stability evaluation, the present invention can effectively avoid tipping or rollover accidents caused by uneven terrain and ensure operation safety.

[0061] The adaptive terrain leveling function of the present invention can quickly adapt to various terrains, reducing the time and effort of manual outrigger adjustment and improving operation efficiency.

[0062] The control strategy based on deep neural networks and reinforcement learning adopted by the present invention enables the truck-mounted crane to maintain stable and reliable operation under various complex and changing terrain conditions, enhancing its adaptability and reliability. Brief Description of the Drawings

[0063] Figure 1 It is a schematic flowchart of the method for leveling and stability evaluation of the truck-mounted crane outriggers adapting to terrain in an embodiment of the present invention;

[0064] Figure 2 It is a schematic structural diagram of the system for leveling and stability evaluation of the truck-mounted crane outriggers adapting to terrain in an embodiment of the present invention. Detailed Embodiments

[0065] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are only a part rather than all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0066] The technical solutions of the present invention will be described in detail below with specific embodiments. These specific embodiments may be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments.

[0067] Figure 1 It is a schematic flowchart of the method for leveling and stability evaluation of the truck-mounted crane outriggers adapting to terrain in an embodiment of the present invention, as Figure 1 shown, the method includes:

[0068] S1. Filter and weight-fuse the terrain elevation data, outrigger pressure data, and vehicle body attitude data collected by sensors, and then input them into the outrigger ground contact modeling unit. The outrigger ground contact modeling unit classifies terrain features using a historical evolution predictor based on a deep neural network and establishes a ground stiffness-deformation mapping relationship, and constructs a non-linear spring-damping contact force model using an online learning contact parameter adaptor to obtain outrigger load distribution data;

[0069] S2. Input the outrigger load distribution data into the stability evaluation unit of the outrigger. The stability evaluation unit includes a stability discriminator, a stress uniformity evaluator, and a dynamic load early warning device. The stability discriminator forms a support polygon based on the outrigger landing points and calculates the support stability margin in combination with the projection position of the vehicle body center of gravity. The stress uniformity evaluator calculates the stress uniformity index based on the outrigger pressure variance. The dynamic load early warning device analyzes the dynamic characteristics of the outrigger pressure and outputs an early warning signal; the outputs of the stability discriminator, the stress uniformity evaluator, and the dynamic load early warning device are weighted and fused through an adaptive weight coefficient that adjusts in real time with the vehicle body attitude to obtain a stability state matrix;

[0070] S3. Input the stability state matrix into the model predictive controller. The model predictive controller constructs a terrain adaptability evaluation function using the reinforcement learning method and generates a leg leveling control instruction through the action value network. Input the leg leveling control instruction into the distributed cooperative control unit. The distributed cooperative control unit sets the legs as network nodes for information interaction and controls the leg displacement using a feedforward-feedback composite controller with an adaptive gain.

[0071] In an alternative embodiment,

[0072] The leg-ground contact modeling unit classifies terrain features using a historical evolution predictor based on a deep neural network and establishes a ground stiffness-deformation mapping relationship. The steps of constructing a non-linear spring-damping contact force model using an online learning contact parameter adaptor to obtain leg load distribution data include:

[0073] Classify the terrain features using a historical evolution predictor of a double-branch densely connected convolutional neural network. The double-branch densely connected convolutional neural network includes a high-frequency feature extraction branch for extracting local texture features and a low-frequency feature extraction branch for extracting regional morphological features. Each feature extraction branch is provided with an independent spatial pyramid pooling module. The terrain classification result is output after adaptively fusing the double-branch features through a cascaded attention module.

[0074] Based on the terrain classification result, establish a hybrid ground mechanics model. Use a three-layer backpropagation neural network to establish a mapping relationship between ground stiffness and deformation to obtain ground deformation characteristics. Input the ground deformation characteristics, combined with the leg pressure sequence, displacement sequence, and terrain classification result, into a double-layer long short-term memory network for soil mechanics parameter identification. The first layer of the double-layer long short-term memory network outputs a preliminary estimate of the soil parameters, and the second layer introduces historical data to perform a temporal correction on the preliminary estimate of the soil parameters.

[0075] Construct a non-linear spring-damping contact force model that couples multiple physical fields according to the soil mechanics parameters. The non-linear spring-damping contact force model includes a stiffness function related to the ground moisture content, a damping function related to the temperature, and a friction coefficient function related to the relative velocity and pressure. Use a parameter adaptor based on an action evaluation framework to evaluate the leg-ground contact state and perform online optimization on the parameters of the stiffness function, damping function, and friction coefficient function.

[0076] Based on the non-linear spring-damping contact force model, establish a dynamic coupling equation for the leg group. Under the constraint of the dynamic coupling equation of the leg group, use a load distributor with a prediction-correction structure for optimal distribution. Extract the leg force signal features through wavelet decomposition, and compensate for high-frequency vibration and low-frequency drift by combining an adaptive notch filter and an integral sliding mode controller, and output the leg load distribution data.

[0077] Exemplarily, first, a dual-branch densely connected convolutional neural network is used to classify the features of the terrain. The high-frequency feature extraction branch is responsible for extracting the local texture features of the terrain, such as fine cracks, particle size, etc.; the low-frequency feature extraction branch is responsible for extracting the regional morphological features, such as the overall slope, undulation degree, etc. Each branch is equipped with an independent spatial pyramid pooling module to integrate features of different scales. The features extracted by the two branches are adaptively fused through a concatenated attention module, and finally the terrain classification results are output, such as different types of terrain like sandy soil, clay, gravel, etc.

[0078] To illustrate more clearly, assume that we use a set of image data containing different terrain types such as sandy soil, clay, gravel, etc. for training. The training data includes the images themselves and the corresponding terrain type labels. After the network learns the features of different terrain types, it can accurately classify new terrain images. For example, when a new image is input into the network and the network determines that the image is sandy soil, it outputs the label of the sandy soil type.

[0079] Based on the terrain classification results, a hybrid ground mechanics model is established. A three-layer backpropagation neural network is used to establish the mapping relationship between the ground stiffness and the deformation amount. This neural network takes the terrain type as one of the input features and combines the deformation amount to predict the ground stiffness. This part of the training data includes the deformation amounts and the corresponding stiffness values under different terrain types. After the network learns the relationship between the deformation amount and the stiffness value under different terrain types, it can predict the stiffness according to the deformation amount. For example, when the sandy soil type and a deformation amount value are input, the network can predict the corresponding stiffness value of the sandy soil.

[0080] The ground deformation characteristics, the leg pressure sequence, the displacement sequence, and the terrain classification results are input into a two-layer long short-term memory network (LSTM) for soil mechanics parameter identification. The first-layer LSTM network outputs the preliminary estimated values of the soil parameters, such as the Young's modulus, Poisson's ratio, etc. of the soil. The second-layer LSTM network introduces historical data to perform temporal correction on the preliminary estimated values of the soil parameters output by the first layer, improving the accuracy of parameter estimation.

[0081] Assume that the leg pressure sequence, the displacement sequence, and the terrain classification results for a period of time have been obtained. These data are used as inputs and input into the first-layer LSTM network, and the network outputs the preliminary estimated values of the soil parameters. Then, these preliminary estimated values and the historical data are input into the second-layer LSTM network together, and the network outputs the corrected soil parameter values. The corrected soil parameter values more accurately reflect the mechanical properties of the soil.

[0082] Based on the identified soil mechanical parameters, a non-linear spring-damping contact force model for multi-physical field coupling is constructed. This model includes a stiffness function related to the ground water content, a damping function related to the temperature, and a friction coefficient function related to the relative velocity and pressure. A parameter adaptor based on the action evaluation framework is used to evaluate the contact state between the outrigger and the ground, and online optimize the parameters of the stiffness function, damping function, and friction coefficient function. For example, if the outrigger sliding is detected, the parameters of the friction coefficient function are adjusted to more accurately simulate the sliding process.

[0083] Assume an initial soil parameter is set into the contact force model. During the contact process between the outrigger and the ground, the parameter adaptor adjusts the parameters in the model, such as stiffness, damping, and friction coefficient, according to the real-time monitored contact state, such as the displacement, velocity, pressure, etc. of the outrigger. If a change in the contact state is detected, such as from static contact to sliding, the adaptor adjusts the parameters so that the model can more accurately reflect the actual situation.

[0084] Based on the non-linear spring-damping contact force model, the dynamic coupling equations of the outrigger group are established. Under the constraints of this equation, a load distributor with a prediction-correction structure is used for load optimization distribution. The outrigger force signal features are extracted by wavelet decomposition, and combined with an adaptive notch filter and an integral sliding mode controller to compensate for high-frequency vibration and low-frequency drift, and finally the outrigger load distribution data is output.

[0085] Assume there are four outriggers. According to the established dynamic coupling equations, the load distributor calculates the load that each outrigger should bear. By wavelet decomposition, the influence of high-frequency vibration and low-frequency drift can be removed, improving the accuracy of the results. Finally, the load magnitude and direction of each outrigger are output.

[0086] The present invention can more accurately predict terrain features, establish a ground stiffness-deformation mapping relationship, and construct an accurate contact force model through a deep learning model and a parameter adaptive algorithm, ultimately improving the calculation accuracy of the outrigger load distribution; the online learning contact parameter adaptor can adjust the model parameters according to the real-time contact state, enhancing the adaptability and robustness of the model, and improving the applicability of the model under different terrains and working conditions; the use of an efficient deep learning algorithm and a load distributor with a prediction-correction structure can reduce the calculation complexity while ensuring the calculation accuracy, improving the calculation efficiency.

[0087] In an alternative embodiment,

[0088] The stress uniformity evaluator calculates the stress uniformity index based on the variance of the outrigger pressures, and the dynamic load early warning device outputs an early warning signal based on the analysis of the dynamic characteristics of the outrigger pressures. The steps of weighted fusion of the outputs of the stability discriminator, the stress uniformity evaluator, and the dynamic load early warning device through adaptive weight coefficients that are adjusted in real time according to the vehicle body attitude to obtain the stability state matrix include:

[0089] Calculate the variance of the outrigger pressures based on the outrigger load distribution data, introduce a terrain correction coefficient related to the terrain slope and terrain undulation to correct the variance of the outrigger pressures, and combine the corrected variance of the outrigger pressures with the maximum pressure to calculate the stress uniformity index;

[0090] Extract multi-scale dynamic characteristics of the outrigger pressure sequence, extract dynamic characteristics through orthogonal complement decomposition and variational mode decomposition, construct a time-frequency joint characteristic index based on the coupling function, and trigger a dynamic load early warning signal when the time-frequency joint characteristic index exceeds the adaptive early warning threshold, where the adaptive early warning threshold is dynamically adjusted according to the changes in the terrain slope and the slewing speed of the hoisting mechanism;

[0091] Construct a weight mapping function based on the vehicle body attitude angle, height, and slewing speed of the hoisting mechanism to perform initial weight allocation for the support stability margin, the stress uniformity index, and the dynamic load early warning signal, calculate the rate of change of the weight with respect to the vehicle body attitude angle, height, and slewing speed of the hoisting mechanism according to the weight mapping function to obtain the state sensitivity matrix, dynamically update the weight coefficients based on the state sensitivity matrix, the increment of the vehicle body state parameters, and the newly calculated weight values, and perform weight optimization under the constraint conditions that the sum of the weight coefficients is 1 and the difference between the weight coefficients at adjacent times does not exceed a preset threshold, and perform weighted fusion of the optimized weights with the support stability margin, the stress uniformity index, and the dynamic load early warning signal to obtain the stability state matrix.

[0092] Exemplarily, first, collect the load data of the crane outriggers, including the pressure values of each outrigger. Then, calculate the stability margin of each outrigger according to the structural parameters and load distribution of the crane. The stability margin refers to the difference between the actual load borne by the outrigger and its ultimate load-bearing capacity, reflecting the stability degree of the outrigger. The calculation of the stability margin takes into account the influence of factors such as the inclination angle of the crane body, height, and the extension length of the boom. For example, if the pressures of the four outriggers are 10 tons, 12 tons, 11 tons, and 13 tons respectively, and the ultimate load-bearing capacity of each outrigger is 20 tons, then the stability margins of each outrigger are 10 tons, 8 tons, 9 tons, and 7 tons respectively. These data can be stored in a database for use in subsequent steps.

[0093] Collect the pressure data of the crane outriggers, calculate the variance of the outrigger pressure, and the variance value reflects the uniformity of the outrigger pressure distribution. The smaller the variance value, the more uniform the outrigger pressure distribution. To more accurately reflect the actual situation, a terrain correction coefficient is introduced, which is related to the terrain slope and terrain undulation. The greater the terrain slope and the greater the terrain undulation, the greater the terrain correction coefficient and the greater the correction amplitude for the outrigger pressure variance. Suppose the calculated outrigger pressure variance is 2 and the terrain correction coefficient is 1.2, then the corrected outrigger pressure variance is 2.4. Finally, combine the corrected outrigger pressure variance and the maximum outrigger pressure to calculate the stress uniformity index. The closer the stress uniformity index is to 1, the more uniform the outrigger pressure distribution and the more stable the crane. For example, if the corrected outrigger pressure variance is 2.4 and the maximum pressure is 13 tons, then the corresponding stress uniformity index, such as 0.8, can be calculated according to a pre-set formula.

[0094] Extract the multi-scale dynamic features of the outrigger pressure sequence, and use the orthogonal complement decomposition and variational mode decomposition methods to extract dynamic features, such as frequency, amplitude and other information. Then, based on these features, construct a time-frequency joint feature index. When this index exceeds the preset adaptive warning threshold, a dynamic load warning signal is triggered. The adaptive warning threshold will be dynamically adjusted according to the terrain slope and the slewing speed of the crane mechanism. For example, if the terrain slope is large and the slewing speed of the crane mechanism is fast, then the adaptive warning threshold will be increased accordingly. The processed data in this part is also stored in the database.

[0095] Construct a weight mapping function based on the vehicle body attitude angle, height and slewing speed of the crane mechanism, and perform initial weight allocation for the support stability margin, stress uniformity index and dynamic load warning signal. For example, in flat terrain and at low speed operation, the weight of the support stability margin can be set relatively high, while the weights of the other two indicators are relatively low. Then, calculate the change rates of the weights with respect to the vehicle body attitude angle, height and slewing speed of the crane mechanism according to the weight mapping function to obtain the state sensitivity matrix. Dynamically update the weight coefficients based on the state sensitivity matrix, the increment of the vehicle body state parameters and the newly calculated weight values. During the weight update process, it is necessary to ensure that the sum of all weight coefficients is 1, and the difference between the weight coefficients at adjacent times does not exceed the preset threshold to ensure the smoothness of the weight adjustment.

[0096] Perform weighted fusion of the optimized weights with the support stability margin, stress uniformity index and dynamic load warning signal to obtain the final stability state matrix. This matrix reflects the current stability state of the crane and can be used for real-time monitoring and warning. For example, if the value in the stability state matrix is lower than the preset threshold, the system issues an alarm signal to remind the operator to pay attention to safety.

[0097] The present invention comprehensively evaluates the stability of a crane by integrating multiple indicators, effectively reduces the probability of lifting accidents, monitors the crane status in real time, gives early warnings in a timely manner, avoids operation interruptions caused by stability problems, and improves work efficiency; dynamically adjusts the weight coefficients to adapt to different working conditions and improves the reliability of stability evaluation.

[0098] In an alternative embodiment,

[0099] For multi-scale dynamic feature extraction of the outrigger pressure sequence, dynamic features are extracted through empirical mode decomposition and variational mode decomposition, and a time-frequency joint feature index is constructed based on a coupling function. When the time-frequency joint feature index exceeds an adaptive warning threshold, a dynamic load warning signal is triggered. The steps of dynamically adjusting the adaptive warning threshold according to the terrain slope and the slewing speed of the lifting mechanism include:

[0100] Obtain the outrigger pressure signal, outrigger deformation amount, and ground reaction force signal, construct an interaction dynamics feature matrix of the outrigger and the ground based on the outrigger pressure signal, outrigger deformation amount, and ground reaction force signal, calculate the transient component based on the interaction dynamics feature matrix using the empirical mode decomposition algorithm, and extract the dynamic feature vector based on the transient component;

[0101] Perform improved variational mode decomposition on the transient component, where an adaptive window function dynamically adjusted according to the local characteristics of the signal is introduced during the variational mode decomposition process to obtain multiple intrinsic mode components and their corresponding central frequencies;

[0102] Construct the joint distribution of the dynamic feature vector, intrinsic mode components, and central frequencies based on the coupling function. The generating function of the coupling function is dynamically selected according to feature correlation, and the time-frequency joint feature index is obtained through integral operation on the joint distribution;

[0103] Construct a multi-objective optimization function including the time-frequency joint feature index, terrain slope, and slewing speed of the lifting mechanism, perform mean processing on the time-frequency joint feature index and construct an accumulation sequence, establish a grey differential equation model based on the accumulation sequence, construct a background value sequence and establish a parameter estimation equation, solve the parameter estimation equation by the least squares method to obtain the time response function, use cumulative reduction to obtain the predicted value of the time-frequency joint feature index, and perform online update on the grey differential equation model through a sliding time window; construct a fuzzy rule base based on the terrain slope, slewing speed of the lifting mechanism, and predicted time-frequency joint feature index, obtain the threshold adjustment amount through fuzzy inference, and adaptively update the warning threshold by dynamically adjusting the learning rate according to the prediction error; when the time-frequency joint feature index exceeds the warning threshold after adaptive update, trigger the outrigger dynamic load warning signal.

[0104] Exemplarily, first, sensors are used to obtain the outrigger pressure signal, outrigger deformation signal, and ground reaction force signal during the operation of the crane. These signals are usually obtained through sensors installed on the outriggers and the ground, such as pressure sensors, displacement sensors, etc. The sampling frequency of the sensors needs to be determined according to the actual situation to ensure that the rapid changes in the dynamic load of the outriggers can be captured. For example, the sampling frequency can be set to 1000 Hz. The collected original signals may contain noise and need to be preprocessed, such as filtering, etc., to remove the influence of noise. Wavelet filtering or mean filtering, etc., can be selected as the preprocessing method.

[0105] Next, the collected outrigger pressure signal, outrigger deformation signal, and ground reaction force signal are integrated. The values of these signals at the same time point are combined into a vector, and these vectors are arranged in chronological order to construct the interactive dynamic characteristic matrix of the outrigger and the ground.

[0106] The interactive dynamic characteristic matrix is decomposed using the positive interactive complementary decomposition algorithm, and the matrix is decomposed into a trend component and a transient component. The positive interactive complementary decomposition algorithm is a non-linear signal processing method that can effectively separate different components in the signal. The transient component represents the rapidly changing part of the signal and contains important information about the dynamic load of the outrigger. Focus on the transient component because it is directly related to the dynamic load.

[0107] Then, the dynamic feature vector is extracted from the extracted transient component. The dynamic feature vector can include multiple features, such as mean, variance, peak value, kurtosis, waveform factor, etc., and these features can reflect the characteristics of the dynamic load of the outrigger. The feature extraction method can be selected according to the actual needs. For example, the energy, spectral entropy, etc., of the transient component can be calculated. Finally, a dynamic feature vector containing multiple feature values is obtained, and these feature values represent certain characteristics of the dynamic load of the outrigger.

[0108] The obtained transient component is subjected to improved variational mode decomposition. The improvement lies in introducing an adaptive window function during the variational mode decomposition process. The size of the adaptive window function is dynamically adjusted according to the local characteristics of the signal, and it can better adapt to the non-stationary characteristics of the signal. For example, when the signal changes violently, a smaller window is used; when the signal changes gently, a larger window is used, so that the signal can be decomposed more precisely to obtain multiple intrinsic mode components and their corresponding central frequencies.

[0109] Using a connection function, the extracted dynamic feature vectors are combined with the obtained intrinsic mode components and their central frequencies. The choice of the connection function depends on the correlation between the features. If some features are highly correlated, a connection function that can reflect this correlation can be selected. For example, an appropriate connection function can be selected according to the correlation coefficient between the features. For example, a simple weighted average or a more complex non-linear function can be used. Integrating the joint distribution yields the final time-frequency joint feature index.

[0110] Construct a multi-objective optimization function that includes the time-frequency joint feature index, terrain slope, and slewing speed of the hoisting mechanism. The time-frequency joint feature index is averaged, and an accumulated sequence is constructed. Based on this accumulated sequence, a grey differential equation model is established to predict the future time-frequency joint feature index. The time response function is obtained by solving the parameter estimation equation using the least squares method. The predicted value of the time-frequency joint feature index is obtained by inverse accumulation reduction, and the grey differential equation model is updated online through a sliding time window.

[0111] A fuzzy rule base is constructed using the terrain slope, slewing speed of the hoisting mechanism, and the predicted time-frequency joint feature index. The threshold adjustment amount is obtained through fuzzy inference, and the learning rate is dynamically adjusted according to the prediction error to adaptively update the warning threshold. For example, if the terrain slope is large, the warning threshold should be increased accordingly; if the slewing speed of the hoisting mechanism is fast, the warning threshold should also be increased accordingly.

[0112] When the calculated time-frequency joint feature index exceeds the warning threshold after adaptive update, a warning signal for the outrigger dynamic load is triggered.

[0113] The multi-scale dynamic feature extraction and adaptive warning threshold adjustment adopted in the present invention improve the accuracy of the warning, reduce the possibility of false alarms and missed alarms; the online updated grey differential equation model and adaptive warning threshold mechanism ensure the real-time nature of the warning, providing timely and effective warning information for the operator; it can adapt to different terrain slopes and slewing speeds of the hoisting mechanism, enhancing the robustness of the system and improving the reliability and stability of the system.

[0114] In an alternative embodiment,

[0115] The steps for the model predictive controller to construct a terrain adaptability evaluation function using the reinforcement learning method and generate outrigger leveling control commands through the action value network include:

[0116] Construct a terrain adaptability evaluation function that includes outrigger height deviation, roll angle deviation, pitch angle deviation, ground contact force, and outrigger slip speed. The terrain adaptability evaluation function adopts an adaptive weight allocation strategy, and the adaptive weight allocation strategy is dynamically adjusted based on terrain roughness, slope standard deviation, ground friction coefficient, and ground stress distribution standard deviation. During the dynamic adjustment of the weights, a Sigmoid function with a periodic compensation term is used, and the ground stress distribution is corrected by a Gaussian kernel function;

[0117] Design an action value network structure that includes an evaluation network and a target network. The state spaces of the evaluation network and the target network include outrigger height vectors, vehicle body attitude angles, terrain feature vectors, and ground stress distribution features, and the action space is the outrigger speed command; The evaluation network uses a multi-head self-attention mechanism and a residual convolution structure for feature extraction, and calculates the value estimate of the state-action pair based on the output of the terrain adaptability evaluation function; In the model predictive controller, the evaluation network is trained and optimized, and the parameters of the evaluation network are softly updated to the target network according to the training error through an adaptive update rate;

[0118] Establish an optimization objective function that includes a short-term control cost term, a control quantity adjustment term, and a terminal penalty term. The short-term control cost term is provided by the evaluation network, the terminal penalty term is calculated from the long-term value estimate provided by the target network, and the control quantity adjustment term includes the square term of the control increment between adjacent moments;

[0119] Formulate the optimization objective function, the outrigger kinematic constraints, the actuator limit constraints, and the control increment constraints into an optimization problem, and use the alternating direction multiplier method based on Nesterov acceleration to solve the optimization problem. The alternating direction multiplier method decomposes the original problem into a sequence of sub-problems through dual decomposition, and the conjugate gradient method is used to solve the sub-problems; The solution result of the optimization problem is output as the outrigger leveling control command.

[0120] Exemplarily, construct an evaluation function to evaluate the terrain adaptability of the vehicle. This function considers multiple key factors: outrigger height deviation, roll angle deviation, pitch angle deviation, ground contact force, and outrigger slip speed. To enable the evaluation function to adapt to different terrain conditions, an adaptive weight allocation strategy is adopted. This strategy dynamically adjusts the weights of each factor according to real-time terrain information. The terrain information includes terrain roughness, slope standard deviation, ground friction coefficient, and ground stress distribution standard deviation.

[0121] Specifically, first, vehicle sensor data is obtained to calculate the above-mentioned terrain information and vehicle state parameters. Then, an improved Sigmoid function (with a periodic compensation term added to handle periodically changing terrain) is used to dynamically adjust each weight. The input of this Sigmoid function is the terrain information, and the output is the weight value, ranging from 0 to 1. The introduction of the periodic compensation term is to better handle periodically changing terrain, such as wavy road surfaces. Finally, the Gaussian kernel function is used to smooth the ground stress distribution to reduce the influence of measurement noise.

[0122] For example, assume the initial weights are: leg height deviation 0.5, roll angle deviation 0.2, pitch angle deviation 0.2, ground contact force 0.05, leg slip speed 0.05. When encountering a relatively flat road surface (low terrain roughness, low slope standard deviation, high friction coefficient, low ground stress distribution standard deviation), the weight of the leg height deviation may slightly decrease, while the weight of the ground contact force may slightly increase to ensure the stability of the vehicle. Conversely, when encountering a rough road surface (high terrain roughness, high slope standard deviation, low friction coefficient, high ground stress distribution standard deviation), the weight of the leg height deviation may increase to ensure that the vehicle can adapt to the uneven terrain.

[0123] Design an action value network consisting of an evaluation network and a target network. The state spaces of the evaluation network and the target network include: leg height vector (height of each leg), vehicle body attitude angles (roll angle and pitch angle), terrain feature vector (terrain roughness, slope standard deviation, ground friction coefficient, ground stress distribution standard deviation), and ground stress distribution characteristics (e.g., mean and variance of the ground stress distribution). The action space is the leg speed command (speed command for each leg).

[0124] The evaluation network uses the multi-head self-attention mechanism to capture the complex relationships between various features in the state space, and uses the residual convolution structure to extract features. Finally, combined with the output of the terrain adaptability evaluation function, the value estimate of the state-action pair is calculated. The structure of the target network is the same as that of the evaluation network, and its parameters are updated from the evaluation network through a soft update mechanism. The soft update mechanism uses an adaptive update rate to adjust the update amplitude according to the training error, improving the training efficiency and stability.

[0125] Construct an optimization objective function, which includes a short-term control cost term, a control quantity adjustment term, and a terminal penalty term. The short-term control cost term is provided by the evaluation network, representing the cost of taking a specific action in the current state. The terminal penalty term is provided by the target network, representing the long-term value estimate, which is used to guide the controller to select actions that can bring long-term benefits. The control quantity adjustment term penalizes the change in the control increment at adjacent moments to ensure the smoothness of the control instruction and prevent the control from jittering violently.

[0126] The optimization objective function is combined with the outrigger kinematic constraints, such as outrigger stroke limits, actuator limit constraints, such as motor speed limits, and control increment constraints, such as control increment limits, to form a constrained optimization problem. The alternating direction method of multipliers (ADMM) based on Nesterov acceleration is used to solve this optimization problem. The ADMM method decomposes the original problem into multiple sub-problems through dual decomposition, and then uses the conjugate gradient method to solve each sub-problem. The final solution result is the outrigger leveling control command. The obtained outrigger leveling control command is sent to the actuator to control the movement of the outrigger and achieve the leveling of the vehicle.

[0127] Through the adaptive weight allocation strategy and the multi-factor evaluation function, the present invention can effectively cope with various terrain conditions; by optimizing the objective function and constraint conditions, it can ensure the stability and smoothness of the vehicle during the leveling process, avoid dangerous situations such as severe jitter or overturning, and improve the safety of the vehicle; based on the action value network of reinforcement learning and the efficient optimization algorithm, it can quickly and accurately calculate the outrigger leveling control command, thereby realizing precise leveling control and improving the control efficiency and accuracy.

[0128] In an alternative embodiment,

[0129] The step of training and optimizing the evaluation network in the model predictive controller and softly updating the parameters of the evaluation network to the target network according to the training error through an adaptive update rate includes:

[0130] Construct a distributed experience replay pool. The distributed experience replay pool adopts a hierarchical storage structure, calculates the environmental complexity index based on the terrain roughness and slope standard deviation, calculates the task difficulty index based on the outrigger height error vector, roll angle error, and pitch angle error, and stores the state-action-reward value-next state transition samples hierarchically according to the weighted combination of the environmental complexity index and the task difficulty index;

[0131] Design a hybrid priority sampling mechanism based on the distributed experience replay pool. The hybrid priority sampling mechanism includes priority calculation based on the Wasserstein distance and importance sampling weight calculation based on the ratio of control decision functions. Dynamically adjust the combined weight coefficient of the priority calculation and the importance sampling weight calculation during the training phase, and adjust the sampling probability of different difficulty samples according to the combined weight coefficient;

[0132] Construct a dual-gradient optimizer, which includes a control decision gradient optimizer and an evaluation network gradient optimizer. The control decision gradient optimizer uses the proximal policy optimization algorithm with trust region constraints to calculate the control decision update amount based on the ratio of the current control decision function and the historical control decision function. The evaluation network gradient optimizer uses the adaptive moment estimation algorithm with momentum term to construct a loss function based on the prediction error and gradient norm of the evaluation network;

[0133] Design a multi-scale adaptive learning mechanism based on the optimization results of the dual-gradient optimizer. Jointly adjust the base learning rate according to the temporal difference error, the variance of the control decision update amount, and the loss value of the evaluation network, and adaptively adjust the exploration noise amplitude based on the statistical distribution characteristics of the leg speed command output;

[0134] Adopt a hierarchical network update strategy, and design different update frequencies for the feature extraction layer and the control instruction output layer of the evaluation network respectively. The feature extraction layer updates its parameters according to a preset first update period, and the control instruction output layer updates its parameters according to a preset second update period, where the first update period is greater than the second update period. Soft update the parameters of the evaluation network to the target network after weighted averaging according to the verification performance.

[0135] Exemplarily, construct an experience replay pool with hierarchical storage to store state-action-reward-next state transition samples. First, define the environmental complexity index and the task difficulty index. The environmental complexity index is calculated based on terrain data. For example, a value is calculated through terrain roughness and slope standard deviation. The higher the value, the more complex the environment. The task difficulty index is calculated based on the robot control error. For example, a value is calculated according to the leg height error, roll angle error, and pitch angle error. The higher the value, the more difficult the task. Then, according to the weighted combination of the environmental complexity index and the task difficulty index, the samples are stored in different levels of the replay pool in a hierarchical manner. For example, weights of 0.6 and 0.4 can be assigned to the environmental complexity index and the task difficulty index respectively, and the weighted sum is calculated as the basis for grading. The weights can be adjusted according to the actual situation.

[0136] Based on the constructed distributed experience replay pool, a hybrid priority sampling mechanism is designed. This mechanism combines priority calculation based on the Wasserstein distance and importance sampling weight calculation based on the ratio of control decision functions. The Wasserstein distance is used to measure the similarity between samples. The smaller the distance, the higher the priority. The ratio of control decision functions reflects the difference between the current policy and the historical policy. The larger the ratio, the higher the importance sampling weight. In the initial stage of training, more reliance is placed on the Wasserstein distance for sampling to avoid being dominated by a small number of high-reward samples prematurely. In the later stage of training, more reliance is placed on the importance sampling weight to improve the sampling efficiency. The combined weight coefficient is dynamically adjusted according to the training stage. For example, in the initial stage, the Wasserstein distance weight is 0.8 and the importance sampling weight is 0.2; in the later stage, it is the opposite.

[0137] Construct a dual-gradient optimizer that includes a control decision gradient optimizer and an evaluation network gradient optimizer. The control decision gradient optimizer uses the proximal policy optimization algorithm with trust region constraints to update the control decision by calculating the ratio of the current control decision function and the historical control decision function. The evaluation network gradient optimizer uses the adaptive moment estimation algorithm with momentum terms to construct a loss function based on the prediction error and gradient norm of the evaluation network. The step size of each update is restricted by the trust region constraint to ensure the stability of the optimization process.

[0138] Dynamically adjust the base learning rate according to the temporal difference error, the variance of the control decision update amount, and the loss value of the evaluation network. The larger the temporal difference error, the smaller the learning rate; the larger the variance of the control decision update amount, the smaller the learning rate; the larger the loss value of the evaluation network, the smaller the learning rate. At the same time, adaptively adjust the exploration noise amplitude based on the statistical distribution characteristics of the outrigger speed command output. If the command output distribution is too concentrated, increase the exploration noise amplitude; otherwise, decrease it.

[0139] For example, if the temporal difference errors of multiple consecutive training steps are large, reduce the base learning rate to 50% of the original; if the loss value of the evaluation network continues to decrease, consider increasing the learning rate.

[0140] The feature extraction layer and the control command output layer of the evaluation network adopt different update frequencies. The feature extraction layer updates the parameters at a lower frequency, for example, once every 100 iterations, and the control command output layer updates the parameters at a higher frequency, for example, at each iteration. Finally, the parameters of the evaluation network are softly updated to the target network by weighted averaging according to the validation performance. For example, if the parameter weights with better performance on the validation set are larger.

[0141] Through hierarchical storage of samples, hybrid priority sampling, and dual gradient optimizers, the controller can better learn control strategies under different complexities and difficulties, thereby improving control accuracy and stability; the multi-scale adaptive learning mechanism and hierarchical network update strategy make the training process more efficient, fully utilize sample information, and avoid resource waste; adaptively adjusting the exploration noise amplitude and learning rate enables the controller to better adapt to environmental changes and disturbances, enhancing robustness.

[0142] In an alternative embodiment,

[0143] The steps of the distributed cooperative control unit setting the outriggers as network nodes for information interaction and using a feedforward-feedback composite controller with adaptive gain to control the outrigger displacement include:

[0144] Construct a distributed cooperative control unit network, where each outrigger corresponds to a distributed cooperative control unit as a network node. The state vector of the network node includes outrigger height, roll angle, pitch angle, speed, and contact force information. Based on the state vector of the network node, construct the outrigger displacement error and error change rate, and adaptively adjust the network communication radius, bandwidth parameter, and state component weight according to the outrigger displacement error and error change rate to obtain the dynamic connection weight and weighted state vector;

[0145] Based on the dynamic connection weight and weighted state vector, calculate the exponential function of the outrigger displacement error to generate an adaptive trigger function. When the outrigger displacement error is greater than the adaptive trigger function, trigger information interaction. Calculate the network node confidence using the outrigger displacement error, design a time-varying information fusion weight based on the network node confidence, construct a time-varying consensus protocol considering communication delay, and design an adaptive update law for the protocol gain matrix to output a consensus control quantity;

[0146] Construct a feedforward-feedback composite controller based on the consensus control quantity. The feedforward-feedback composite controller includes a disturbance observer and a neural network compensator. The disturbance observer estimates the disturbance quantity during the outrigger movement based on the feedforward control term and the outrigger displacement error. The neural network compensator obtains the motion compensation control value by online learning the weighted state vector, and combines the disturbance quantity, motion compensation control value, and the consensus control quantity to construct a composite control law;

[0147] Design a predictive control and fault diagnosis mechanism based on the composite control rate and the weighted state vector. Use a long short-term memory network to predict the state of network nodes. Design a prediction error compensation term based on the error between the predicted network node state and the actual state. Calculate a fault index through the weighted state vector and the residual of the predicted network node state. When the fault index is greater than a preset threshold, it is determined that a fault has occurred. When a fault occurs, recalculate the control rate based on the state vectors of neighboring network nodes and compensate for the control amount of the faulty leg.

[0148] Exemplarily, first, construct a distributed cooperative control unit network. Each leg corresponds to a control unit, which serves as a node in the network. Each node needs to collect and update its own state information in real time, including data such as the current height, roll angle, pitch angle, vertical speed, and contact force with the ground of the leg. These data constitute the state vector of each node.

[0149] Based on the state vector of each node, calculate the leg displacement error. This error is the difference between the current displacement of the leg and the target displacement. At the same time, calculate the rate of change of the error, that is, the speed at which the error changes over time. According to the calculated displacement error and the rate of change of the error, the system will adaptively adjust the connection strength between nodes in the network, that is, the communication radius and bandwidth, as well as the importance of each state component in control, that is, the state component weight. For example, if the error is large and the rate of change is fast, increase the connection strength and importance; otherwise, decrease them. These adjusted parameters constitute the dynamic connection weights, which are multiplied by the state vector to obtain the weighted state vector. For example, if the weight of the height component is set to 0.5, the weighted height information is 1.2×0.5 = 0.6.

[0150] Use the leg displacement error and the dynamic connection weights to calculate an adaptive trigger function. Only when the leg displacement error is greater than this trigger function will the node send its state information to its neighboring nodes, avoiding unnecessary communication. At the same time, calculate the confidence of each node according to the leg displacement error. The higher the confidence, the more reliable the state information of the node. For example, the smaller the error, the higher the confidence.

[0151] Based on the confidence of the nodes, design a time-varying information fusion weight to weighted average the state information from neighboring nodes. This part will consider the influence of communication delay. At the same time, design a time-varying consensus protocol to ensure that all nodes finally reach a consistent control goal. To optimize the performance of the protocol, an adaptive update law needs to be designed to adjust the protocol gain matrix. For example, if it is found that the control effect is not good, increase the gain; otherwise, decrease the gain. Finally, output the consensus control amount.

[0152] Based on the consistency control quantity, a feedforward-feedback composite controller is constructed. This controller includes a disturbance observer and a neural network compensator. The disturbance observer estimates various disturbances during the movement of the outrigger according to the feedforward control term and the outrigger displacement error, such as uneven ground or wind influence. The neural network compensator learns and compensates for the non-linear characteristics in the outrigger movement by online learning the weighted state vector, so as to obtain the motion compensation control value. Finally, the disturbance quantity obtained by the disturbance observer, the motion compensation control value obtained by the neural network compensator and the consistency control quantity are combined to construct the final composite control rate.

[0153] Based on the composite control rate and the weighted state vector, a predictive control and fault diagnosis mechanism is designed. The long short-term memory network (LSTM) is used to predict the future state of each node. The predicted state is compared with the actual state, the predicted error compensation term is calculated and added to the control rate. At the same time, the residual between the weighted state vector and the predicted state is calculated as the fault index. When the fault index exceeds the preset threshold, it is determined that a fault has occurred. Once a fault occurs, the system recalculates the control rate based on the state vectors of neighboring nodes and compensates the control quantity of the faulty outrigger.

[0154] Through the combination of distributed cooperative control and feedforward-feedback composite control, the present invention effectively suppresses the influence of disturbances and non-linear factors, and significantly improves the accuracy and stability of outrigger displacement control; the introduction of adaptive adjustment of network connections, confidence weighting, and fault diagnosis and compensation mechanisms enhances the robustness and fault tolerance of the system. Even if some outriggers fail, the overall stable operation of the system can be guaranteed; by controlling the information interaction frequency through an adaptive trigger function, the network communication burden is reduced, the computational complexity is lowered, and the real-time performance of the system is improved.

[0155] Figure 2 It is a schematic structural diagram of the outrigger self-adaptive terrain leveling and stability evaluation system according to the embodiment of the present invention, as Figure 2 shown, the system includes:

[0156] The first unit is used to filter and weight-fuse the terrain elevation data, outrigger pressure data and vehicle body attitude data collected by the sensor and then input them into the outrigger ground contact modeling unit. The outrigger ground contact modeling unit classifies the terrain features by using a historical evolution predictor based on a deep neural network and establishes a ground stiffness-deformation mapping relationship, and constructs a non-linear spring-damping contact force model by using an online learning contact parameter adaptor to obtain the outrigger load distribution data;

[0157] A second unit for inputting the outrigger load distribution data into a stability evaluation unit of the outriggers. The stability evaluation unit includes a stability discriminator, a stress uniformity evaluator, and a dynamic load early warning device. The stability discriminator forms a support polygon based on the outrigger landing points and calculates the support stability margin in combination with the projection position of the vehicle body's center of gravity. The stress uniformity evaluator calculates the stress uniformity index based on the outrigger pressure variance. The dynamic load early warning device analyzes the dynamic characteristics of the outrigger pressure and outputs an early warning signal. The outputs of the stability discriminator, the stress uniformity evaluator, and the dynamic load early warning device are weighted and fused through an adaptive weight coefficient that is adjusted in real time according to the vehicle body attitude to obtain a stability state matrix.

[0158] A third unit for inputting the stability state matrix into a model predictive controller. The model predictive controller constructs a terrain adaptability evaluation function using a reinforcement learning method and generates an outrigger leveling control instruction through an action value network. The outrigger leveling control instruction is input into a distributed cooperative control unit. The distributed cooperative control unit sets the outriggers as network nodes for information interaction and controls the outrigger displacement using a feedforward-feedback composite controller with an adaptive gain.

[0159] In the third aspect of the embodiments of the present invention,

[0160] There is provided an electronic device, including:

[0161] A processor;

[0162] A memory for storing instructions executable by the processor;

[0163] Wherein, the processor is configured to call the instructions stored in the memory to execute the method described above.

[0164] In the fourth aspect of the embodiments of the present invention,

[0165] There is provided a computer-readable storage medium, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the method described above is implemented.

[0166] The present invention can be a method, a device, a system, and / or a computer program product. The computer program product can include a computer-readable storage medium, on which computer-readable program instructions for executing various aspects of the present invention are uploaded.

[0167] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for evaluating the leveling and stability of a truck-mounted crane outrigger in an adaptive terrain manner, characterized in that: include: The terrain elevation data, outrigger pressure data and vehicle body posture data collected by the sensor are filtered and weighted and then input into the outrigger ground contact modeling unit. The outrigger ground contact modeling unit uses a historical evolution predictor based on a deep neural network to classify terrain features and establish a ground stiffness-deformation mapping relationship, and uses an online learning contact parameter adaptor to build a nonlinear spring damping contact force model to obtain outrigger load distribution data; The load distribution data of the outrigger is input into a stability evaluation unit of the outrigger, wherein the stability evaluation unit includes a stability discriminator, a stress uniformity evaluator and a dynamic load early warning device, wherein the stability discriminator forms a support polygon based on the outrigger landing point and calculates the support stability margin in combination with the projection position of the center of gravity of the vehicle body, the stress uniformity evaluator calculates the stress uniformity index based on the outrigger pressure variance, and the dynamic load early warning device outputs an early warning signal based on the dynamic characteristic analysis of the outrigger pressure; the outputs of the stability discriminator, the stress uniformity evaluator and the dynamic load early warning device are weightedly fused through an adaptive weight coefficient adjusted in real time with the vehicle body posture to obtain a stability state matrix; The stability state matrix is ​​input into a model predictive controller, which uses a reinforcement learning method to construct a terrain adaptability evaluation function and generates a leg leveling control instruction through an action value network; the leg leveling control instruction is input into a distributed collaborative control unit, which sets the legs as network nodes and performs information exchange, and uses a feedforward feedback composite controller with adaptive gain to control the leg displacement.

2. The method according to claim 1, characterized in that The outrigger ground contact modeling unit uses a historical evolution predictor based on a deep neural network to classify terrain features and establish a ground stiffness-deformation mapping relationship, and uses an online learning contact parameter adaptor to build a nonlinear spring damping contact force model. The steps of obtaining outrigger load distribution data include: A historical evolution predictor of a dual-branch densely connected convolutional neural network is used to perform feature classification on the terrain. The dual-branch densely connected convolutional neural network includes a high-frequency feature extraction branch for extracting local texture features and a low-frequency feature extraction branch for extracting regional morphological features. Each feature extraction branch is provided with an independent spatial pyramid pooling module. The dual-branch features are adaptively fused through a series of attention modules to output the terrain classification result. A hybrid ground mechanics model is established based on the terrain classification result, a three-layer back propagation neural network is used to establish a mapping relationship between ground stiffness and deformation variable to obtain ground deformation characteristics, and the ground deformation characteristics are combined with the outrigger pressure sequence, displacement sequence and terrain classification results and input into a double-layer long short-term memory network for soil mechanics parameter identification, the first layer of the double-layer long short-term memory network outputs a preliminary estimated value of the soil parameter, and the second layer introduces historical data to perform time series correction on the preliminary estimated value of the soil parameter; A multi-physics field coupled nonlinear spring-damping contact force model is constructed according to soil mechanical parameters. The nonlinear spring-damping contact force model includes a stiffness function related to ground moisture content, a damping function related to temperature, and a friction coefficient function related to relative velocity and pressure. A parameter adaptor based on an action evaluation framework is used to evaluate the contact state between the outrigger and the ground and to perform online optimization on the parameters of the stiffness function, the damping function, and the friction coefficient function. The dynamic coupling equation of the outrigger group is established based on the nonlinear spring-damping contact force model. Under the constraint of the dynamic coupling equation of the outrigger group, a load distributor with a prediction-correction structure is used to optimize the distribution. The characteristics of the outrigger force signal are extracted by wavelet decomposition. The high-frequency vibration and low-frequency drift are compensated in combination with an adaptive notch filter and an integral sliding mode controller, and the outrigger load distribution data is output.

3. The method according to claim 1, characterized in that The stress uniformity evaluator calculates the stress uniformity index based on the outrigger pressure variance, and the dynamic load warning device outputs a warning signal based on the dynamic characteristic analysis of the outrigger pressure; the steps of weighted fusion of the outputs of the stability discriminator, the stress uniformity evaluator and the dynamic load warning device through an adaptive weight coefficient adjusted in real time with the vehicle body posture to obtain a stability state matrix include: The outrigger pressure variance is calculated based on the outrigger load distribution data, and the terrain correction coefficient related to the terrain slope and terrain undulation is introduced to correct the outrigger pressure variance. The stress uniformity index is calculated by combining the corrected outrigger pressure variance with the maximum pressure. The outrigger pressure sequence is subjected to multi-scale dynamic feature extraction, and the dynamic features are extracted by orthogonal complementary decomposition and variational mode decomposition. A time-frequency joint feature index is constructed based on the link function. When the time-frequency joint feature index exceeds an adaptive warning threshold, a dynamic load warning signal is triggered, wherein the adaptive warning threshold is dynamically adjusted with changes in the terrain slope and the rotation speed of the lifting mechanism. A weight mapping function based on the vehicle body attitude angle, height and lifting mechanism rotation speed is constructed to perform initial weight allocation on the support stability margin, stress uniformity index and dynamic load warning signal. The state sensitivity matrix is ​​obtained by calculating the change rate of the weight to the vehicle body attitude angle, height and lifting mechanism rotation speed according to the weight mapping function. The weight coefficient is dynamically updated based on the state sensitivity matrix, the vehicle body state parameter increment and the newly calculated weight value, and the weight optimization is performed under the constraint that the sum of the weight coefficients is one and the difference of the weight coefficients at adjacent moments does not exceed the preset threshold. The optimized weights are weightedly fused with the support stability margin, stress uniformity index and dynamic load warning signal to obtain the stability state matrix.

4. The method according to claim 3, characterized in that The outrigger pressure sequence is subjected to multi-scale dynamic feature extraction, and the dynamic features are extracted by orthogonal complementary decomposition and variational mode decomposition. A time-frequency joint feature index is constructed based on the link function. When the time-frequency joint feature index exceeds an adaptive warning threshold, a dynamic load warning signal is triggered. The adaptive warning threshold is dynamically adjusted with changes in the terrain slope and the rotation speed of the lifting mechanism, and the steps include: Acquire the outrigger pressure signal, outrigger deformation and ground reaction force signal, construct the interactive dynamic characteristic matrix between the outrigger and the ground with the outrigger pressure signal, outrigger deformation and ground reaction force signal, calculate the transient component based on the interactive dynamic characteristic matrix by using the orthogonal complementary decomposition algorithm, and extract the dynamic characteristic vector based on the transient component; Performing improved variational mode decomposition on the transient component, wherein an adaptive window function dynamically adjusted according to local characteristics of the signal is introduced in the variational mode decomposition process to obtain multiple eigenmode components and their corresponding center frequencies; Based on the link function, a joint distribution of the dynamic feature vector, the intrinsic mode component and the center frequency is constructed, a generating function of the link function is dynamically selected according to feature correlation, and a time-frequency joint feature index is obtained by integrating the joint distribution; A multi-objective optimization function including the time-frequency joint characteristic index, terrain slope and lifting mechanism rotation speed is constructed, the time-frequency joint characteristic index is averaged and a cumulative sequence is constructed, a grey differential equation model is established based on the cumulative sequence, a background value sequence is constructed and a parameter estimation equation is established, the parameter estimation equation is solved by the least squares method to obtain the time response function, the predicted value of the time-frequency joint characteristic index is obtained by cumulative reduction, and the grey differential equation model is updated online through a sliding time window; a fuzzy rule base is constructed based on the terrain slope, the lifting mechanism rotation speed and the predicted time-frequency joint characteristic index, the threshold adjustment amount is obtained by fuzzy reasoning, and the warning threshold is adaptively updated by dynamically adjusting the learning rate according to the prediction error; when the time-frequency joint characteristic index exceeds the warning threshold after adaptive update, the outrigger dynamic load warning signal is triggered.

5. The method according to claim 1, characterized in that The model predictive controller uses a reinforcement learning method to construct a terrain adaptability evaluation function, and the steps of generating a leg leveling control instruction through an action value network include: A terrain adaptability evaluation function including outrigger height deviation, roll angle deviation, pitch angle deviation, ground contact force and outrigger sliding speed is constructed. The terrain adaptability evaluation function adopts an adaptive weight allocation strategy. The adaptive weight allocation strategy is dynamically adjusted based on terrain roughness, slope standard deviation, ground friction coefficient and outrigger ground stress distribution standard deviation. In the process of dynamic weight adjustment, a Sigmoid function with a periodic compensation term is used to correct the outrigger ground stress distribution through a Gaussian kernel function. Design an action value network structure including an evaluation network and a target network, wherein the state space of the evaluation network and the target network includes an outrigger height vector, a vehicle body attitude angle, a terrain feature vector and an outrigger stress distribution feature, and the action space is an outrigger speed instruction; the evaluation network uses a multi-head self-attention mechanism and a residual convolution structure to extract features, and calculates the value estimate of the state-action pair based on the output of the terrain adaptability evaluation function; the evaluation network is trained and optimized in the model predictive controller, and the parameters of the evaluation network are soft-updated to the target network through an adaptive update rate according to the training error; Establishing an optimization objective function including a short-term control cost term, a control amount adjustment term and a terminal penalty term, wherein the short-term control cost term is provided by the evaluation network, the terminal penalty term is calculated by the long-term value estimate provided by the target network, and the control amount adjustment term includes a square term of the control increment at adjacent moments; The optimization objective function and the outrigger kinematic constraints, actuator limit constraints and control increment constraints constitute an optimization problem, and the optimization problem is solved by the alternating direction multiplier method based on Nesterov acceleration. The alternating direction multiplier method decomposes the original problem into a sequence of sub-problems through dual decomposition, and the sub-problems are solved by the conjugate gradient method; the solution result of the optimization problem is output as an outrigger leveling control instruction.

6. The method according to claim 5, characterized in that The step of training and optimizing the evaluation network in the model predictive controller and soft-updating the parameters of the evaluation network to the target network through an adaptive update rate according to the training error comprises: Constructing a distributed experience replay pool, wherein the distributed experience replay pool adopts a hierarchical storage structure, calculates an environmental complexity index based on terrain roughness and slope standard deviation, calculates a task difficulty index based on leg height error vector, roll angle error, and pitch angle error, and hierarchically stores state-action-reward value-next state transition samples according to a weighted combination of the environmental complexity index and the task difficulty index; A hybrid priority sampling mechanism is designed based on the distributed experience replay pool, wherein the hybrid priority sampling mechanism includes a priority calculation based on the Wasserstein distance and an importance sampling weight calculation based on the control decision function ratio, and the combined weight coefficient of the priority calculation and the importance sampling weight calculation is dynamically adjusted according to the training stage, and the sampling probability of samples of different difficulty levels is adjusted according to the combined weight coefficient; Constructing a dual gradient optimizer, the dual gradient optimizer comprising a control decision gradient optimizer and an evaluation network gradient optimizer, the control decision gradient optimizer adopting a proximal strategy optimization algorithm with a trust region constraint, calculating a control decision update amount based on a ratio of a current control decision function to a historical control decision function, and the evaluation network gradient optimizer adopting an adaptive moment estimation algorithm with a momentum term, constructing a loss function based on a prediction error and a gradient norm of the evaluation network; Based on the optimization results of the dual gradient optimizer, a multi-scale adaptive learning mechanism is designed, the basic learning rate is adjusted according to the temporal difference error, the variance of the control decision update amount and the loss value of the evaluation network, and the exploration noise amplitude is adaptively adjusted based on the statistical distribution characteristics of the outrigger speed command output; A hierarchical network update strategy is adopted, and differentiated update frequencies are designed for the feature extraction layer and the control instruction output layer of the evaluation network respectively. The feature extraction layer updates parameters according to a preset first update period, and the control instruction output layer updates parameters according to a preset second update period, wherein the first update period is greater than the second update period. The parameters of the evaluation network are soft-updated to the target network after weighted averaging according to the verification performance.

7. The method according to claim 1, characterized in that The steps of setting the outriggers as network nodes and performing information exchange by the distributed collaborative control unit and using a feedforward feedback composite controller with adaptive gain to control the displacement of the outriggers include: Construct a distributed collaborative control unit network, where each leg corresponds to a distributed collaborative control unit as a network node, and the state vector of the network node includes leg height, roll angle, pitch angle, speed and contact force information. The leg displacement error and error change rate are constructed based on the state vector of the network node. The network communication radius and bandwidth parameters and state component weights are adaptively adjusted according to the leg displacement error and error change rate to obtain a dynamic connection weight and a weighted state vector. Based on the dynamic connection weight and the weighted state vector, an exponential function of the leg displacement error is calculated to generate an adaptive trigger function. When the leg displacement error is greater than the adaptive trigger function, information interaction is triggered. The leg displacement error is used to calculate the network node confidence. Based on the network node confidence, a time-varying information fusion weight is designed. A time-varying consistency protocol considering communication delay is constructed, and an adaptive update law of the protocol gain matrix is ​​designed to output a consistency control amount. A feedforward feedback composite controller is constructed based on the consistency control amount, wherein the feedforward feedback composite controller includes a disturbance observer and a neural network compensator, wherein the disturbance observer estimates the disturbance amount in the outrigger motion process based on the feedforward control term and the outrigger displacement error, and the neural network compensator obtains the motion compensation control value by online learning the weighted state vector, and the disturbance amount, the motion compensation control value and the consistency control amount are combined to construct a composite control rate; A predictive control and fault diagnosis mechanism is designed based on the composite control rate and the weighted state vector, a long short-term memory network is used to predict the state of a network node, a prediction error compensation term is designed based on the error between the predicted state of the network node and the actual state, a fault index is calculated by the residual between the weighted state vector and the predicted state of the network node, a fault is determined to have occurred when the fault index is greater than a preset threshold, and when a fault occurs, the control rate is recalculated based on the state vector of the neighboring network node and the control amount of the faulty leg is compensated.

8. A self-adaptive terrain leveling and stability assessment system for a truck-mounted crane outrigger, used to implement the method described in any one of claims 1 to 7, characterized in that: include: The first unit is used to filter and weight the terrain elevation data, outrigger pressure data and vehicle body posture data collected by the sensor and then input them into the outrigger ground contact modeling unit, wherein the outrigger ground contact modeling unit uses a historical evolution predictor based on a deep neural network to classify terrain features and establish a ground stiffness-deformation mapping relationship, and uses an online learning contact parameter adaptor to construct a nonlinear spring damping contact force model to obtain outrigger load distribution data; The second unit is used to input the outrigger load distribution data into the stability evaluation unit of the outrigger, the stability evaluation unit includes a stability discriminator, a stress uniformity evaluator and a dynamic load early warning device, the stability discriminator forms a support polygon based on the outrigger landing point and calculates the support stability margin in combination with the projection position of the vehicle body's center of gravity, the stress uniformity evaluator calculates the stress uniformity index based on the outrigger pressure variance, and the dynamic load early warning device outputs an early warning signal based on the dynamic characteristic analysis of the outrigger pressure; the outputs of the stability discriminator, the stress uniformity evaluator and the dynamic load early warning device are weightedly fused through an adaptive weight coefficient adjusted in real time with the vehicle body posture to obtain a stability state matrix; The third unit is used to input the stability state matrix into a model predictive controller, the model predictive controller uses a reinforcement learning method to construct a terrain adaptability evaluation function, and generates a leg leveling control instruction through an action value network; the leg leveling control instruction is input into a distributed collaborative control unit, the distributed collaborative control unit sets the leg as a network node and performs information exchange, and uses a feedforward feedback composite controller with adaptive gain to control the leg displacement.

9. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method described in any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images