Agricultural tractor body stabilizing system

Abstract

The invention relates to the technical field of tractors, in particular to an agricultural tractor body stabilizing system which comprises a front axle hydraulic lifting system, a front axle is lifted and lowered through a hydraulic cylinder and a control valve, the system combines hydraulic pressure compensation and flow compensation technologies, and the lifting amplitude is automatically adjusted according to the load change of the front axle; and the pressure and flow compensation control system is used for adjusting hydraulic flow and pressure through a hydraulic control unit. By combining a hydraulic lifting system, pressure flow compensation control, intelligent feedback control and a high-precision sensor, an efficient and intelligent agricultural tractor body stabilizing system can be designed. The system can automatically adjust the height of the front axle, improve the grounding efficiency of the front wheels, reduce pitching and optimize traction distribution according to the road condition and load change, so that the driving speed and stability are improved.

Description

Technical Field

[0001] The invention relates to the technical field of tractors, in particular to a body stabilizing system of an agricultural tractor. Background Art

[0002] Existing tractor front axle. The front axle is mainly composed of left front terminal transmission assembly, front central transmission assembly, right front terminal transmission assembly, steering cylinder assembly, tie rod assembly, etc.

[0003] The front axle structure of tractors is widely used in the field of agricultural machinery. Its structural design and performance indicators have an important impact on agricultural production. The front axle structures of different types of tractors have different load-bearing capacity, steering performance, stability, etc. According to different operating requirements, the corresponding front axle structure should be selected for configuration. During driving, the driver can control the swing of the steering knuckle through the steering wheel, so that the tractor can turn left or right, and the driving force is transmitted to the tractor tires through the transmission of the front axle structure to make it rotate. At present, the front axle of domestic tractors is connected to the frame through a swing shaft, and the rear axle is rigidly connected to the frame. The tractor cannot realize the active suspension function and is not equipped with a suspension hydraulic system. Tractors driving on uneven roads have large vibrations at the driver, poor driving comfort, and low body stability.

[0004] Therefore, an agricultural tractor body stabilization system is proposed. Summary of the invention

[0005] The purpose of the present invention is to provide an agricultural tractor body stabilization system. By combining a hydraulic lifting system, pressure flow compensation control, intelligent feedback control and high-precision sensors, an efficient and intelligent agricultural tractor body stabilization system can be designed. The system can automatically adjust the height of the front axle according to road conditions and load changes, improve the front wheel ground contact efficiency, reduce pitch, optimize traction distribution, and thus improve driving speed and stability. Future optimization directions can focus on improving the intelligence level of the system, reducing energy consumption, increasing sensor accuracy, and improving the reliability and maintenance convenience of the overall system to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solution: an agricultural tractor body stabilization system, comprising:

[0007] The front axle hydraulic lifting system uses hydraulic cylinders and control valves to raise and lower the front axle. The system combines hydraulic pressure compensation and flow compensation technologies to automatically adjust the lifting range according to changes in the front axle load.

[0008] The pressure and flow compensation control system adjusts the hydraulic flow and pressure through the hydraulic control unit to respond to the load changes of the front axle in real time;

[0009] Sensor and feedback system, real-time feedback data of front axle lifting and lowering. By installing tilt sensors, load sensors and ground pressure sensors on the front axle, the system monitors the posture, load and ground contact of the front axle in real time;

[0010] The intelligent control unit integrates all sensor signals, determines the working status of the front axle through data fusion algorithm, and adjusts the hydraulic system. The control unit adopts fuzzy control, PID control or adaptive control algorithm to adjust the hydraulic system in real time according to different road conditions.

[0011] Preferably, the front axle hydraulic lifting system adopts a more efficient hydraulic pump and control valve combined with a control model to optimize the dynamic response of the hydraulic lifting system. The control model is:

[0012]

[0013] Where L is the inertia of the hydraulic system, A is the flow control area of ​​the hydraulic system, K p It is the compression stiffness of the hydraulic pump. When optimizing the hydraulic system, the ratio of A to L is reduced to increase K. p value, reducing the response time t r .

[0014] Preferably, the intelligent control unit introduces a deep neural network model combined with real-time terrain changes, load conditions, and vehicle speed data to dynamically adjust the lifting strategy;

[0015] The lifting amount h(t) of the front axle based on the deep neural network model, where the input features include vehicle speed v(t), load F(t), and ground inclination angle θ(t);

[0016] Based on the deep neural network model: h(t) = f(v(t), F(t), θ(t); W)

[0017] Among them, f is the forward propagation function of the neural network, W is the weight matrix obtained by training, and h(t) is the lifting amount of the front axle;

[0018] The training process can be completed by minimizing the loss function, and the loss function L is usually the mean squared error:

[0019]

[0020] Among them, h i (t) is the actual lifting amount, h pred (t) is the predicted rise or fall, and N is the total number of samples.

[0021] Preferably, the sensor and feedback system adopts multi-sensor fusion technology to improve the perception accuracy of the system, and fuses the signals of multiple sensors through a Kalman filter;

[0022] The state equation and observation equation of the Kalman filter are as follows:

[0023] State equation: xk=Axk-1+Buk+wk

[0024] Among them, xk is the current state, such as the position and speed of the front axle, A is the system state transfer matrix, B is the control matrix, uk is the control input, and wk is the process noise;

[0025] Observation equation: zk=Hxk+vk

[0026] Where zk is the observation value, the reading from the sensor, H is the observation matrix, and vk is the observation noise;

[0027] During the Kalman filtering process, the filter optimizes the state estimation by minimizing the covariance of the observation noise vk and the process noise wk, and finally gives the optimal front axle lift estimate.

[0028] Preferably, the deep neural network model also includes a front axle lifting optimization control strategy based on road surface characteristics, by defining the network input as the dynamic information of the vehicle, including: vehicle speed v(t), load F(t), ground inclination angle θ(t) and road surface type characteristics r(t), and the road surface type characteristics are collected and identified by ground pressure sensors;

[0029] The neural network model can be expressed as: h(t) = f(W·x(t) + b)

[0030] Where f is the forward propagation function of the neural network, W is the trained weight matrix, h(t) is the lift amount of the front axle, and x(t) = [v(t)F(t), θ(t)r(t)] is the input feature vector.

[0031] Preferably, the deep neural network model also includes reinforcement learning optimization, fuzzy control and adaptive control. Through reinforcement learning, the tractor adaptively learns the best front axle lifting strategy in a dynamic environment. In reinforcement learning, the tractor control task is modeled as a Markov decision process, where: state space S: includes environmental states, such as load, vehicle speed, front axle tilt angle, and road type;

[0032] Action space A: Control actions for lifting and lowering the front axle, such as raising, lowering or maintaining the current state;

[0033] Reward function R(s,a): rewards are given based on the current state and the selected action, with the goal of maximizing the long-term cumulative reward; if the front axle is successfully raised and lowered to maintain a horizontal state, a high reward is given; if an overreaction occurs or the front axle becomes unstable, a negative reward is given;

[0034] The value function Q(s,a) of reinforcement learning represents the expected return given state s and action a, which is updated by the Bellman equation:

[0035] Q(s,a)←Q(s,a)+α[R(s,a)+γamaxQ(s′,a)-Q(s,a)]

[0036] Where α is the learning rate, γ is the discount factor, and s′ is the next state after the current action is executed;

[0037] Fuzzy control and adaptive control:

[0038] The relationship between input and output is modeled through fuzzy rules. If the input of the system is the vehicle speed v(t) and the load F(t), and the output is the lifting amount of the front axle h(t), the control amount of the lifting of the front axle can be obtained through fuzzy reasoning and defuzzification process.

[0039] Preferably, a hybrid control method is adopted, combining deep learning, reinforcement learning and fuzzy control, switching different control strategies under different working conditions, using fuzzy control for smooth response under low speed and low load conditions, and using reinforcement learning for adaptive adjustment in complex dynamic environments;

[0040] Combination of adaptive fuzzy control and deep learning: When the environment is relatively stable, fuzzy control is used to avoid the computational overhead of deep neural networks; when the environment changes greatly, deep learning is used for strategy optimization. Fuzzy control provides smooth control outputs, while deep learning extracts deep dynamic patterns from large amounts of data.

[0041] Preferably, a decision layer is set in the hybrid control to dynamically select the control algorithm;

[0042] If the current state of the system is S(t), by analyzing S(t), we can get the complexity index C(t) of the current environment and decide which control algorithm to use:

[0043] C(t)=w1·load(t)+w2·speed(t)+w3·terrain(t)

[0044] Among them, C(t) represents the complexity of the current environment, w1, w2, w3 are the weights of each environmental factor, and are dynamically adjusted according to environmental changes;

[0045] If C(t) is large, it means that the system faces a more complex operating environment, and deep reinforcement learning is preferred for adaptive control;

[0046] Control selection strategy:

[0047]

[0048] Where T1 is the complexity threshold.

[0049] Preferably, the sensor and feedback system adopts multimodal sensor fusion to integrate information from different types of sensors such as vision, laser radar, and pressure sensor, and adopts extended Kalman filtering or particle filtering to fuse signals from multiple sensors to optimize the prediction accuracy of the lifting and lowering of the front axle;

[0050] If the state vector of the system is x(t), including the front axle position, speed, and tilt angle, and the sensor observation value is z(t), then the state update and observation update formulas of the system are as follows:

[0051] Status prediction:

[0052]

[0053] Where F is the state transfer matrix, B is the control matrix, and u(t) is the control input;

[0054] Observation Update:

[0055] K(t)=P(t|t-1)H T (HP(t|t-1)H T +R) -1

[0056]

[0057] Where K(t) is the Kalman gain, P(t|t-1) is the covariance of the state prediction, RRR is the observation noise covariance, and H is the observation matrix.

[0058] The agricultural tractor body stabilizer comprises a front axle body, a central transmission assembly is arranged in the middle of the front axle body, and the two ends of the front axle body are respectively connected to the left front axle and the right front axle, two groups of steering cylinders connected to the left front axle and the right front axle are symmetrically distributed and fixed on the front side of the front axle body, a frame plate body is arranged on the top of the front axle body, and connecting plates are fixed on the front and rear sides of the frame plate body, and the frame plate body is connected and fixed to the front axle body through the connecting plates.

[0059] Compared with the prior art, the present invention has the following beneficial effects:

[0060] By combining the hydraulic lifting system, pressure flow compensation control, intelligent feedback control and high-precision sensors, an efficient and intelligent agricultural tractor body stabilization system can be designed. The system can automatically adjust the front axle height according to road conditions and load changes, improve the front wheel ground contact efficiency, reduce pitch, optimize traction distribution, and thus improve driving speed and stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0062] Figure 1 A block diagram of the agricultural tractor body stabilization system of the present invention;

[0063] Figure 2 This is an overall structural view of the agricultural tractor body stabilizing device of the present invention;

[0064] Figure 3 A bottom view of the agricultural tractor body stabilizer of the present invention;

[0065] Figure 4 It is a front view of the agricultural tractor body stabilizer of the present invention.

[0066] Description of reference numerals:

[0067] 1. Frame plate body; 2. Left front axle; 3. Right front axle; 4. Central transmission assembly; 5. Steering cylinder; 6. Front axle body; 7. Connecting plate. DETAILED DESCRIPTION

[0068] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0069] See also Figures 1 to 4 , the present invention provides a technical solution:

[0070] Agricultural tractor body stabilization system, including:

[0071] The front axle hydraulic lifting system uses hydraulic cylinders and control valves to raise and lower the front axle. The system combines hydraulic pressure compensation and flow compensation technologies to automatically adjust the lifting range according to changes in the front axle load.

[0072] Hydraulic cylinder: used to raise and lower the front axle, quickly respond to weight changes, and raise or lower the front axle position.

[0073] Flow and pressure compensation valve: By adjusting the flow and pressure of the hydraulic oil, it ensures the smoothness and accuracy of the front axle lifting and lowering. Flow compensation can ensure that the system maintains a constant lifting rate under different loads and speeds, while pressure compensation can adjust the lifting force according to weight changes.

[0074] The pressure and flow compensation control system adjusts the hydraulic flow and pressure through the hydraulic control unit to respond to the load changes of the front axle in real time;

[0075] The system uses a hydraulic control unit to adjust the hydraulic flow and pressure, responding to changes in the load on the front axle in real time. The pressure compensation valve can adjust the pressure according to the change in the force required by the hydraulic cylinder, while the flow compensation valve ensures smooth lifting and lowering movements without jitter.

[0076] Pressure compensation: When the load on the tractor's front axle increases, the pressure compensation system automatically increases the hydraulic pressure to ensure that the front axle can lift and lower smoothly and withstand larger loads.

[0077] Flow compensation: The flow compensation system ensures that the hydraulic system can maintain a constant flow rate when the load fluctuates, making the lifting and lowering of the front axle smoother and avoiding violent up and down fluctuations.

[0078] Sensor and feedback system, real-time feedback data of front axle lifting and lowering. By installing tilt sensors, load sensors and ground pressure sensors on the front axle, the system monitors the posture, load and ground contact of the front axle in real time;

[0079] Load sensor: monitors the weight changes of the front axle in real time to determine whether the hydraulic system needs to be adjusted to adapt to the load changes.

[0080] Tilt sensor: measures the pitch angle of the tractor and ensures that the tractor is kept level by controlling the lifting and lowering of the front axle.

[0081] Ground pressure sensor: detects the contact pressure between the front wheel and the ground, helping the system determine whether the front wheel has sufficient ground contact force and then adjust the front axle position

[0082] The intelligent control unit integrates all sensor signals, determines the working status of the front axle through data fusion algorithm, and adjusts the hydraulic system. The control unit adopts fuzzy control, PID control or adaptive control algorithm to adjust the hydraulic system in real time according to different road conditions.

[0083] Adaptive algorithm: Dynamically adjusts the system's response speed and lifting range according to the tractor's operating conditions (such as vehicle speed, load, terrain, etc.). For example, when driving at high speed, the control unit will reduce the frequency of lifting and lowering the front axle to avoid overreaction; while when driving at low speed or in complex terrain, it will increase the response speed to maintain a more stable ground contact.

[0084] Working principle:

[0085] Perception stage: The system monitors load changes, front axle tilt angle, and contact status between the front wheels and the ground in real time through sensors on the front axle.

[0086] Data analysis and control phase:

[0087] Load change: When the tractor load changes (such as passing an obstacle, tilling, etc.), the load sensor sends a signal to the control unit.

[0088] Road impact: When the system detects that the front axle is impacted by the road (such as going over a pothole or an obstacle), the tilt sensor and ground pressure sensor will provide feedback information.

[0089] Feedback control: Based on the above information, the intelligent control unit determines whether it is necessary to adjust the pressure and flow of the hydraulic system, and then adjusts the height of the front axle to ensure smooth and timely lifting and lowering of the front axle.

[0090] Adjustment stage: The hydraulic control unit adjusts the lifting and lowering of the front axle according to the instructions of the intelligent control unit to keep the tractor in a stable and horizontal state and reduce pitch.

[0091] Dynamic stability: The system always maintains good contact between the front wheels and the ground through continuous feedback and adjustment, improving traction and reducing power loss caused by loss of contact between the front axle.

[0092] Specifically, the front axle hydraulic lifting system adopts a more efficient hydraulic pump and control valve combined with a control model to optimize the dynamic response of the hydraulic lifting system. The control model is:

[0093]

[0094] Where L is the inertia of the hydraulic system, A is the flow control area of ​​the hydraulic system, K p It is the compression stiffness of the hydraulic pump. When optimizing the hydraulic system, the ratio of A to L is reduced to increase K. p value, reducing the response time t r .

[0095] Assuming a conventional hydraulic system, the response time is tr = 0.5tr = 0.5 seconds, and the inertia in the system is L = 0.2L = 0.2kg·m 2 , control valve flow adjustment area A = 0.02A = 0.02m 2 , hydraulic pump stiffness Kp = 1000Kp = 1000N·m·s / rad. After using high-efficiency pumps and flow control valves, after optimization, the response time is shortened to tr = 0.3tr = 0.3 seconds, the hydraulic pump stiffness is increased to Kp = 1500Kp = 1500N·m·s / rad, and the inertia is reduced to L = 0.1L = 0.1kg·m2

[0096] Specifically, the intelligent control unit introduces a deep neural network model combined with real-time terrain changes, load conditions, and vehicle speed data to dynamically adjust the lifting strategy;

[0097] The lifting amount h(t) of the front axle based on the deep neural network model, where the input features include vehicle speed v(t), load F(t), and ground inclination angle θ(t);

[0098] Based on the deep neural network model: h(t) = f(v(t), F(t), θ(t); W)

[0099] Among them, f is the forward propagation function of the neural network, W is the weight matrix obtained by training, and h(t) is the lifting amount of the front axle;

[0100] The training process can be completed by minimizing the loss function, and the loss function L is usually the mean squared error:

[0101]

[0102] Among them, h i (t) is the actual lifting amount, h pred (t) is the predicted rise or fall, and N is the total number of samples.

[0103] By using deep neural networks to train load and vehicle speed characteristics of different terrains (such as flat land, slopes, and muddy land), the results show that the optimized control strategy can better adjust the lifting and lowering of the front axle and reduce over-adjustment or slow response compared to the traditional PID control strategy. Through experiments, the DNN control model can reduce the pitch angle deviation of the front axle by about 30%.

[0104] Specifically, the sensor and feedback system adopts multi-sensor fusion technology to improve the perception accuracy of the system, and fuses the signals of multiple sensors through the Kalman filter;

[0105] The state equation and observation equation of the Kalman filter are as follows:

[0106] State equation: xk=Axk-1+Buk+wk

[0107] Among them, xk is the current state, such as the position and speed of the front axle, A is the system state transfer matrix, B is the control matrix, uk is the control input, and wk is the process noise;

[0108] Observation equation: zk=Hxk+vk

[0109] Where zk is the observation value, the reading from the sensor, H is the observation matrix, and vk is the observation noise;

[0110] During the Kalman filtering process, the filter optimizes the state estimation by minimizing the covariance of the observation noise vk and the process noise wk, and finally gives the optimal front axle lift estimate.

[0111] After using the Kalman filter for sensor data fusion, the control accuracy of the front axle lifting and lowering has increased by 20%. For example, the original front axle height estimation error was ±0.05m, but after the Kalman filter, the error was reduced to ±0.04m, significantly improving the system stability.

[0112] Specifically, the deep neural network model also includes a front axle lifting optimization control strategy based on road surface characteristics, by defining the network input as the dynamic information of the vehicle, including: vehicle speed v(t), load F(t), ground inclination angle θ(t) and road surface type characteristics r(t), and the road surface type characteristics are collected and identified using ground pressure sensors;

[0113] The neural network model can be expressed as: h(t) = f(W·x(t) + b)

[0114] Where f is the forward propagation function of the neural network, W is the trained weight matrix, h(t) is the lift amount of the front axle, and x(t) = [v(t)F(t), θ(t)r(t)] is the input feature vector.

[0115] Specifically, the deep neural network model also includes reinforcement learning optimization, fuzzy control and adaptive control. Through reinforcement learning, the tractor adaptively learns the best front axle lifting strategy in a dynamic environment. In reinforcement learning, the tractor control task is modeled as a Markov decision process, where: state space S: includes environmental states, such as load, vehicle speed, front axle tilt angle, and road type;

[0116] Action space A: Control actions for lifting and lowering the front axle, such as raising, lowering or maintaining the current state;

[0117] Reward function R(s,a): rewards are given based on the current state and the selected action, with the goal of maximizing the long-term cumulative reward; if the front axle is successfully raised and lowered to maintain a horizontal state, a high reward is given; if an overreaction occurs or the front axle becomes unstable, a negative reward is given;

[0118] The value function Q(s,a) of reinforcement learning represents the expected return given state s and action a, which is updated by the Bellman equation:

[0119] Q(s,a)←Q(s,a)+α[R(s,a)+γamaxQ(s′,a)-Q(s,a)]

[0120] Where α is the learning rate, γ is the discount factor, and s′ is the next state after the current action is executed;

[0121] Fuzzy control and adaptive control:

[0122] The relationship between input and output is modeled through fuzzy rules. If the input of the system is the vehicle speed v(t) and the load F(t), and the output is the lifting amount of the front axle h(t), the control amount of the lifting of the front axle can be obtained through fuzzy reasoning and defuzzification process.

[0123] Specifically, a hybrid control method is adopted to combine deep learning, reinforcement learning and fuzzy control, and switch different control strategies under different working conditions. Fuzzy control is used for smooth response under low speed and low load conditions, while reinforcement learning is used for adaptive adjustment in complex dynamic environments.

[0124] Combination of adaptive fuzzy control and deep learning: When the environment is relatively stable, fuzzy control is used to avoid the computational overhead of deep neural networks; when the environment changes greatly, deep learning is used for strategy optimization. Fuzzy control provides smooth control outputs, while deep learning extracts deep dynamic patterns from large amounts of data.

[0125] Specifically, a decision layer is set in hybrid control to dynamically select the control algorithm;

[0126] If the current state of the system is S(t), by analyzing S(t), we can get the complexity index C(t) of the current environment and decide which control algorithm to use:

[0127] C(t)=w1·load(t)+w2·speed(t)+w3·terrain(t)

[0128] Among them, C(t) represents the complexity of the current environment, w1, w2, w3 are the weights of each environmental factor, and are dynamically adjusted according to environmental changes;

[0129] If C(t) is large, it means that the system faces a more complex operating environment, and deep reinforcement learning is preferred for adaptive control;

[0130] Control selection strategy:

[0131]

[0132] Where T1 is the complexity threshold.

[0133] Specifically, the sensor and feedback system adopts multimodal sensor fusion to integrate information from different types of sensors such as vision, lidar, and pressure sensors, and adopts extended Kalman filtering or particle filtering to fuse signals from multiple sensors to optimize the prediction accuracy of the front axle lifting and lowering;

[0134] If the state vector of the system is x(t), including the front axle position, speed, and tilt angle, and the sensor observation value is z(t), then the state update and observation update formulas of the system are as follows:

[0135] Status prediction:

[0136]

[0137] Where F is the state transfer matrix, B is the control matrix, and u(t) is the control input;

[0138] Observation Update:

[0139] K(t)=P(t|t-1)H T (HP(t|t-1)H T +R) -1

[0140]

[0141] Where K(t) is the Kalman gain, P(t|t-1) is the covariance of the state prediction, RRR is the observation noise covariance, and H is the observation matrix.

[0142] An agricultural tractor body stabilizer comprises a front axle body 6, wherein a central transmission assembly 4 is arranged in the middle of the front axle body 6, and the two ends of the front axle body 6 are respectively connected to a left front axle 2 and a right front axle 3, two groups of steering cylinders 5 which are transmission-connected to the left front axle 2 and the right front axle 3 are symmetrically distributed and fixed on the front side of the front axle body 6, a frame plate body 1 is arranged on the top of the front axle body 6, and connecting plates 7 are fixed on the front and rear sides of the frame plate body 1, and the frame plate body 1 is connected and fixed to the front axle body 6 via the connecting plates 7.

[0143] 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 it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, 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. Agricultural tractor body stabilization system, characterized in that: include: The front axle hydraulic lifting system uses hydraulic cylinders and control valves to raise and lower the front axle. The system combines hydraulic pressure compensation and flow compensation technologies to automatically adjust the lifting range according to changes in the front axle load. The pressure and flow compensation control system adjusts the hydraulic flow and pressure through the hydraulic control unit to respond to the load changes of the front axle in real time; Sensor and feedback system, real-time feedback data of front axle lifting and lowering. By installing tilt sensors, load sensors and ground pressure sensors on the front axle, the system monitors the posture, load and ground contact of the front axle in real time; The intelligent control unit integrates all sensor signals, determines the working status of the front axle through data fusion algorithm, and adjusts the hydraulic system. The control unit adopts fuzzy control, PID control or adaptive control algorithm to adjust the hydraulic system in real time according to different road conditions.

2. The agricultural tractor body stabilization system according to claim 1, characterized in that: The front axle hydraulic lifting system adopts a more efficient hydraulic pump and control valve combined with a control model to optimize the dynamic response of the hydraulic lifting system. The control model: Where L is the inertia of the hydraulic system, A is the flow control area of ​​the hydraulic system, K p It is the compression stiffness of the hydraulic pump. When optimizing the hydraulic system, the ratio of A to L is reduced to increase K. p value, reducing the response time t r .

3. The agricultural tractor body stabilization system according to claim 2, characterized in that: The intelligent control unit introduces a deep neural network model based on real-time terrain changes, load conditions, and vehicle speed data to dynamically adjust the lifting strategy; The lifting amount h(t) of the front axle based on the deep neural network model, where the input features include vehicle speed v(t), load F(t), and ground inclination angle θ(t); Based on the deep neural network model: h(t) = f(v(t), F(t), θ(t); W) Among them, f is the forward propagation function of the neural network, W is the weight matrix obtained by training, and h(t) is the lifting amount of the front axle; The training process can be completed by minimizing the loss function, and the loss function L is usually the mean squared error: Among them, h i (t) is the actual lifting amount, h pred (t) is the predicted rise or fall, and N is the total number of samples.

4. The agricultural tractor body stabilization system according to claim 3, characterized in that: The sensor and feedback system adopts multi-sensor fusion technology to improve the perception accuracy of the system and fuses the signals of multiple sensors through the Kalman filter; The state equation and observation equation of the Kalman filter are as follows: State equation: xk=Axk-1+Buk+wk Among them, xk is the current state, such as the position and speed of the front axle, A is the system state transfer matrix, B is the control matrix, uk is the control input, and wk is the process noise; Observation equation: zk=Hxk+vk Where zk is the observation value, the reading from the sensor, H is the observation matrix, and vk is the observation noise; During the Kalman filtering process, the filter optimizes the state estimation by minimizing the covariance of the observation noise vk and the process noise wk, and finally gives the optimal front axle lift estimate.

5. The agricultural tractor body stabilization system according to claim 4, characterized in that: The deep neural network model also includes a front axle lifting optimization control strategy based on road surface characteristics, by defining the network input as the dynamic information of the vehicle, including: vehicle speed v(t), load F(t), ground inclination angle θ(t) and road surface type characteristics r(t), and the road surface type characteristics are collected and identified by ground pressure sensors; The neural network model can be expressed as: h(t) = f(W·x(t) + b) Where f is the forward propagation function of the neural network, W is the trained weight matrix, h(t) is the lift amount of the front axle, and x(t) = [v(t)F(t), θ(t)r(t)] is the input feature vector.

6. The agricultural tractor body stabilization system according to claim 5, characterized in that: The deep neural network model also includes reinforcement learning optimization, fuzzy control and adaptive control. Through reinforcement learning, the tractor can adaptively learn the best front axle lifting strategy in a dynamic environment. In reinforcement learning, the control task of the tractor is modeled as a Markov decision process, where: the state space S: includes environmental states, such as load, vehicle speed, front axle tilt angle, and road type; Action space A: Control actions for lifting and lowering the front axle, such as raising, lowering or maintaining the current state; Reward function R(s,a): rewards are given based on the current state and the selected action, with the goal of maximizing the long-term cumulative reward; if the front axle is successfully raised and lowered to maintain a horizontal state, a high reward is given; if an overreaction occurs or the front axle becomes unstable, a negative reward is given; The value function Q(s,a) of reinforcement learning represents the expected return given state s and action a, which is updated by the Bellman equation: Q(s,a)←Q(s,a)+α[R(s,a)+γamaxQ(s′,a)-Q(s,a)] Where α is the learning rate, γ is the discount factor, and s′ is the next state after the current action is executed; Fuzzy control and adaptive control: The relationship between input and output is modeled through fuzzy rules. If the input of the system is the vehicle speed v(t) and the load F(t), and the output is the lifting amount of the front axle h(t), the control amount of the lifting of the front axle can be obtained through fuzzy reasoning and defuzzification process.

7. The agricultural tractor body stabilization system according to claim 6, characterized in that: A hybrid control method is adopted, which combines deep learning, reinforcement learning and fuzzy control, and switches different control strategies under different working conditions. Fuzzy control is used for smooth response under low speed and low load conditions, while reinforcement learning is used for adaptive adjustment in complex dynamic environments. Combination of adaptive fuzzy control and deep learning: When the environment is relatively stable, fuzzy control is used to avoid the computational overhead of deep neural networks; when the environment changes greatly, deep learning is used for strategy optimization. Fuzzy control provides smooth control outputs, while deep learning extracts deep dynamic patterns from large amounts of data.

8. The agricultural tractor body stabilization system according to claim 7, characterized in that: In hybrid control, a decision layer is set to dynamically select the control algorithm; If the current state of the system is S(t), by analyzing S(t), we can get the complexity index C(t) of the current environment and decide which control algorithm to use: C(t)=w1·load(t)+w2·speed(t)+w3·terrain(t) Among them, C(t) represents the complexity of the current environment, w1, w2, w3 are the weights of each environmental factor, and are dynamically adjusted according to environmental changes; If C(t) is large, it means that the system faces a more complex operating environment, and deep reinforcement learning is preferred for adaptive control; Control selection strategy: Where T1 is the complexity threshold.

9. The agricultural tractor body stabilization system according to claim 8, characterized in that: The sensor and feedback system adopts multimodal sensor fusion to integrate information from different types of sensors such as vision, lidar, and pressure sensors, and adopts extended Kalman filtering or particle filtering to fuse signals from multiple sensors to optimize the prediction accuracy of front axle lifting; If the state vector of the system is x(t), including the front axle position, speed, and tilt angle, and the sensor observation value is z(t), then the state update and observation update formulas of the system are as follows: Status prediction: Where F is the state transfer matrix, B is the control matrix, and u(t) is the control input; Observation Update: K(t)=P(t|t-1)H T (HP(t|t-1)H T +R) -1 Where K(t) is the Kalman gain, P(t|t-1) is the covariance of the state prediction, RRR is the observation noise covariance, and H is the observation matrix.

10. An agricultural tractor body stabilizer, characterized in that: The invention comprises a front axle body (6), a central transmission assembly (4) is arranged in the middle of the front axle body (6), and the two ends of the front axle body (6) are respectively connected to the left front axle (2) and the right front axle (3), two groups of steering cylinders (5) which are transmission-connected to the left front axle (2) and the right front axle (3) are symmetrically distributed and fixed on the front side of the front axle body (6), a frame plate body (1) is arranged on the top of the front axle body (6), and connecting plates (7) are fixed on the front and rear sides of the frame plate body (1), and the frame plate body (1) is connected and fixed to the front axle body (6) via the connecting plates (7).

Images