Unmanned agricultural machine path navigation method based on farmland environment perception judgment

Abstract

The invention discloses an unmanned agricultural machine path navigation method based on farmland environment perception judgment, and relates to the technical field of path navigation, and the method comprises the steps: continuously perceiving farmland topographic relief features, gradient changes, and agricultural machine current attitude and wheel speed information before and during the operation of an agricultural machine; identifying a terrain disturbance potential risk area based on the sensing data, and predicting the offset and rollover risk value of the terrain fluctuation to the agricultural machine trajectory; according to the method, a pre-correction model fusing terrain prediction data and agricultural machinery dynamic characteristics is called, a pre-correction path containing the terrain compensation amount is generated, and navigation control parameters are adjusted in advance, so that the pre-correction path containing the terrain compensation amount is actively generated by sensing terrain fluctuation and predicting risk and offset in advance; the problem of causal malposition of risk prior to control is avoided from the source.

Description

Technical Field

[0001] The present invention belongs to the field of path navigation technology, and specifically relates to a path navigation method for unmanned agricultural machinery based on farmland environment perception and judgment. Background Art

[0002] Traditional unmanned agricultural machinery navigation methods rely heavily on preset paths and fixed parameters, ignoring the dynamic complexity of farmland terrain. Farmland terrain is characterized by irregular undulations due to cultivation and natural subsidence, with random slope variations. This can easily lead to deviations in the machinery's trajectory, reduced operating accuracy, and even the risk of rollover due to loss of attitude control. Furthermore, traditional navigation parameter correction lags, making it unable to proactively respond to terrain disturbances and offering poor adaptability. Summary of the Invention

[0003] The purpose of the present invention is to provide a path navigation method for unmanned agricultural machinery based on farmland environment perception and judgment, which solves the technical problem in the existing technology that it is unable to perceive terrain undulations in advance, predict risks and offsets, and actively generate a pre-corrected path containing terrain compensation, thereby avoiding the causal dislocation of risk before control from the source.

[0004] A method for unmanned agricultural machinery path navigation based on farmland environment perception and judgment, comprising: Before and during agricultural machinery operations, it continuously senses the undulating features of farmland terrain, slope changes, and the current posture and wheel speed of the agricultural machinery; Identify potential risk areas of terrain disturbance based on perception data and predict the impact of terrain fluctuations on the trajectory of agricultural machinery and the risk of rollover. Calling a pre-correction model that integrates terrain prediction data and agricultural machinery dynamic characteristics to generate a pre-corrected path with terrain compensation and adjust navigation control parameters in advance; Control the agricultural machinery to travel along the pre-corrected path, compare the actual trajectory with the pre-corrected path deviation in real time, iterate and correct the parameters, cross-validate the consistency between the terrain prediction risk value and the actual posture change, and dynamically update the navigation strategy.

[0005] Furthermore, continuous perception of farmland terrain undulations and slope changes specifically includes: The system uses a fusion sensing method of vehicle-mounted LiDAR and millimeter-wave radar. The LiDAR collects three-dimensional point cloud data of the terrain with a preset grid accuracy, while the millimeter-wave radar captures the surface micro-topography in real time. Filter the collected point cloud data to remove interference points including but not limited to weeds and rocks, and retain valid data reflecting the actual terrain; The slope value and aspect angle of each grid are calculated to generate a terrain slope heat map, where areas with slopes exceeding the set threshold are marked as key perception areas to increase the perception frequency of these areas.

[0006] Furthermore, when identifying potential risk areas of terrain disturbance based on perception data, it specifically includes: Extract terrain undulation characteristic parameters, including the wavelength, amplitude, and slope change rate of continuous undulations. Areas with wavelengths smaller than a preset multiple of the agricultural machinery wheelbase and amplitudes greater than a set value are marked as high-disturbance potential areas. Combined with the current posture data of the agricultural machinery, the roll moment coefficient of the agricultural machinery when driving in the area is calculated. The area where the roll moment coefficient exceeds the set threshold is determined as a potential risk area for terrain disturbance; Risk areas are graded and divided into multiple levels according to the roll moment coefficient. Different levels of areas correspond to different prediction accuracy requirements.

[0007] Furthermore, the pre-correction model integrating terrain prediction data with the dynamic characteristics of agricultural machinery includes: Input parameters include the trajectory deviation of terrain prediction, rollover risk value, the wheelbase of the agricultural machine, the center of gravity height, the stiffness of the suspension system and the current wheel speed; The core algorithm of the model is the compensation calculation logic based on fuzzy PID control, in which the terrain compensation amount is positively correlated with the slope change rate and negatively correlated with the wheel speed of the agricultural machinery; The output includes the steering angle compensation value of the pre-corrected path, the wheel speed adjustment coefficient and the suspension system damping parameters, which work together to offset the impact of terrain disturbances.

[0008] Furthermore, identifying potential risk areas of terrain disturbance and predicting the offset and rollover risk values ​​includes: Based on the initially perceived terrain data, the slope gradient and curvature of the terrain are calculated, and the trajectory deviation and rollover risk of the agricultural machinery passing through the area are predicted; Call the terrain-trajectory correlation data of historical operations in the same area and calculate the deviation rate between the current predicted offset and the historical average offset; If the deviation rate exceeds the preset threshold, the sampling density of terrain perception is increased, the slope gradient and curvature are recalculated, the prediction model parameters are corrected, and the prediction is repeated until the deviation rate is lower than the threshold. The risk value and offset at this time are used as the valid prediction results.

[0009] Furthermore, when calling the pre-correction model to generate the pre-correction path and adjust the control parameters, a cyclic verification step is included: The pre-correction model combines the wheelbase, center of gravity height and terrain compensation of the agricultural machinery to generate the initial pre-correction path and control parameters; Simulate the agricultural machinery to drive along the initial path in a virtual terrain environment and calculate the rollover risk coefficient and energy consumption index of the simulated trajectory; If the rollover risk factor is higher than the safety threshold or the energy consumption index exceeds the upper limit, the weight ratio of the terrain compensation amount is adjusted, the path and parameters are regenerated, and the simulation verification is repeated until both indicators meet the standards and the final pre-corrected path is determined.

[0010] Furthermore, when comparing deviations in real time and iteratively correcting parameters, multiple rounds of adjustments are included: When the deviation between the actual trajectory and the pre-corrected path exceeds a first threshold, adjusting the steering angle and wheel speed parameters proportionally; Continuously monitor the rate of change of the adjusted trajectory deviation. If the rate of change is positive, enter the second round of adjustment and double the steering angle adjustment; After each round of adjustment, the degree of consistency between the actual rollover risk value and the terrain predicted risk value is calculated. If the degree of consistency is lower than the preset standard, the system returns to re-identify the terrain risk area, updates the pre-corrected model parameters, and adjusts again until both the deviation and the degree of consistency meet the requirements.

[0011] Furthermore, when generating the pre-corrected path, a parameter adjustment range optimization step based on terrain similarity is also included: The farmland is divided into a plurality of continuous navigation adjustment zones, each of which corresponds to a set of preset navigation parameters; Calculate the similarity of terrain features between adjacent adjustment areas. If the similarity exceeds the set threshold, the adjacent adjustment areas will be merged into a parameter slow-changing area. In the parameter slowly changing area, the navigation parameters are smoothly transitioned according to the principle of monotonically increasing / decreasing the adjustment amplitude of the detection parameters to avoid sudden parameter changes.

[0012] Furthermore, when performing parameter adjustment on the potential risk area of ​​terrain disturbance, it also includes: Calculate the spatial distance between risk points in the area and identify dense clusters of risk points whose distance is less than a preset ratio of the wheelbase of agricultural machinery; Execution for risk point clusters: Generate parameter adjustment paths based on the principle of minimum gradient of navigation parameter changes between adjacent risk points; On the parameter adjustment path, a PID controller is used to fine-tune the parameters so that the deviation rate between the actual adjustment range and the theoretical predicted value does not exceed the preset range.

[0013] Furthermore, when comparing the actual trajectory with the pre-corrected path deviation in real time and iteratively correcting the parameters, the following is also included: When the parameter adjustment amplitudes of three or more adjacent adjustment areas are detected to exceed the threshold, the terrain similarity reassessment mechanism is triggered; Re-extract the terrain features of the area, perform similarity matching with the terrain features in historical operation data, and identify similar unmarked terrain parts; If there are unmarked similar parts, the area will be included in the parameter slow change area management, and the fuzzy adaptive control algorithm will be used to dynamically adjust the parameter change rate so that the parameter adjustment amplitude conforms to the monotonically increasing / decreasing law.

[0014] Compared with the prior art, the present invention has the following beneficial effects: (1) The present invention effectively reduces the trajectory deviation of agricultural machinery during driving, reduces the risk of rollover, improves the efficiency and quality of agricultural machinery operations, and ensures the safety of agricultural machinery and farmland through the steps of real-time perception of terrain information, risk prediction, advance path correction and parameter adjustment, and dynamic update of navigation strategies. This method can achieve accurate navigation of unmanned agricultural machinery in areas with local undulations and slope changes in plain farmland, as well as in farmland with complex terrain such as hilly and mountainous areas, which is conducive to ensuring that agricultural machinery can efficiently and safely complete field operations such as sowing, fertilizing, and harvesting according to the preset operation path.

[0015] (2) The existing technology only adjusts parameters after the offset occurs, which is essentially "responding to the deviation that has already occurred"; however, this method can proactively generate a pre-corrected path containing terrain compensation by sensing the terrain undulations in advance, predicting risks and offsets, and avoiding the causal dislocation of risk before control from the source. This change in thinking is a reconstruction of traditional navigation logic rather than a simple parameter optimization, which is conducive to avoiding risks during movement. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 Schematic diagram of the framework structure of the method of the present invention. DETAILED DESCRIPTION

[0017] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0018] See also Figure 1 This application provides a method for unmanned agricultural machinery path navigation based on farmland environment perception and judgment, including: Before and during agricultural machinery operations, it continuously senses the undulating features of farmland terrain, slope changes, and the current posture and wheel speed of the agricultural machinery; Identify potential risk areas of terrain disturbance based on perception data and predict the impact of terrain fluctuations on the trajectory of agricultural machinery and the risk of rollover. Calling a pre-correction model that integrates terrain prediction data and agricultural machinery dynamic characteristics to generate a pre-corrected path with terrain compensation and adjust navigation control parameters in advance; Control the agricultural machinery to travel along the pre-corrected path, compare the actual trajectory with the pre-corrected path deviation in real time, iterate and correct the parameters, cross-validate the consistency between the terrain prediction risk value and the actual posture change, and dynamically update the navigation strategy.

[0019] Among them, the undulating characteristics of farmland terrain refer to the uneven state of the farmland surface, including morphological features such as ups and downs of the terrain.

[0020] Among them, slope change refers to the change in the inclination of the farmland surface, that is, the change in height per unit distance.

[0021] Among them, the current posture of the agricultural machinery refers to the position and direction of the agricultural machinery in space, including the tilt angle, pitch angle, etc. of the agricultural machinery.

[0022] Among them, the potential risk area for terrain disturbance refers to the area that may cause interference and danger to the normal operation of agricultural machinery due to factors such as terrain undulation and slope changes.

[0023] Among them, the trajectory offset refers to the deviation between the actual driving trajectory of the agricultural machinery and the preset path.

[0024] Among them, the rollover risk value is a quantitative indicator used to assess the possibility of agricultural machinery rolling over under specific terrain conditions.

[0025] Among them, the pre-correction model is a mathematical model that integrates terrain prediction data and agricultural machinery dynamic characteristics and can correct the path in advance. Among them, the terrain compensation amount is the compensation value added to the pre-corrected path to offset the impact of terrain factors on the driving trajectory of agricultural machinery.

[0026] The pre-corrected path refers to a path that has been processed by a pre-corrected model and includes terrain compensation.

[0027] Among them, navigation control parameters are parameters used to control the driving direction and speed of agricultural machinery, such as steering angle, throttle size, etc.

[0028] Among them, cross-validation refers to the process of verifying the accuracy of the prediction by comparing the risk value of terrain prediction (such as predicted rollover risk) with the actual posture changes of agricultural machinery (such as actual tilt angle).

[0029] The working principle of this application is as follows: before and during agricultural machinery operations, various sensors installed on the agricultural machinery are first used to continuously sense relevant information. For example, lidar and millimeter-wave radar can be used to scan farmland terrain to obtain terrain undulations and slope change data; inertial measurement units such as gyroscopes and accelerometers can be used to monitor the current posture of the agricultural machinery; wheel speed sensors can be used to collect wheel speed information, etc. These sensors work together to provide comprehensive and real-time data support for subsequent analysis and processing; Secondly, data processing algorithms can be used to analyze the collected sensory data. By identifying abnormal changes in the terrain, potential risk areas for terrain disturbances can be identified. Furthermore, combined with the agricultural machinery dynamics model, the amount of track deviation and rollover risk that may be caused by terrain fluctuations can be predicted based on the terrain's undulating characteristics and slope changes. For example, if a steep slope is detected in a certain area, calculations can predict the direction and distance of the track deviation, as well as the likelihood of rollover, that may occur in that area. Next, a pre-correction model is constructed using extensive experimental data and simulation analysis. Based on the predicted trajectory offset, rollover risk, and other information, the model calculates the corresponding terrain compensation and incorporates this information into the preset path, generating a pre-corrected path with terrain compensation. Furthermore, navigation control parameters are adjusted in advance based on the pre-corrected path and terrain characteristics. For example, before entering an area with a steep slope, parameters such as steering angle and throttle level are adjusted in advance to prepare the agricultural machinery for smooth passage through the area. The agricultural machinery is controlled to follow a pre-corrected path. During this process, a high-precision positioning system acquires the actual trajectory of the agricultural machinery in real time. The actual trajectory is compared with the pre-corrected path, and the deviation between the two is calculated. Based on the magnitude of the deviation, the navigation control parameters are iteratively adjusted to ensure that the agricultural machinery follows the pre-corrected path as closely as possible. Simultaneously, the consistency between the predicted terrain risk value and the actual posture changes is cross-validated. If any discrepancies are found, the terrain prediction model is promptly adjusted. This continuous iteration and dynamic updating of the navigation strategy enables the agricultural machinery to adapt to the complex and changing farmland terrain.

[0030] The innovation of this application is that it solves the problem that traditional unmanned agricultural machinery navigation methods are difficult to cope with complex farmland terrain changes. Through real-time perception of terrain information, risk prediction, advance path correction and parameter adjustment, dynamic update of navigation strategies and other steps, it is helpful to reduce the trajectory deviation of agricultural machinery during driving, reduce the risk of rollover, improve the efficiency and quality of agricultural machinery operations, and ensure the safety of agricultural machinery and farmland. Whether it is an area with local undulations and slope changes in plain farmland or a farmland with complex terrain such as hilly and mountainous areas, this method can achieve accurate navigation of unmanned agricultural machinery, which is conducive to ensuring that agricultural machinery can efficiently and safely complete field operations such as sowing, fertilizing, and harvesting according to the preset operation path; Furthermore, existing technologies only adjust parameters after the offset occurs, which is essentially "responding to deviations that have already occurred"; however, this method actively generates a pre-corrected path containing terrain compensation by sensing the terrain undulations in advance, predicting risks and offsets, and thus avoiding the causal dislocation problem of risk taking precedence over control from the source.

[0031] In some embodiments of this application, continuously sensing farmland terrain undulations and slope changes, relying solely on a single sensor can easily lead to blind spots or data errors. For example, lidar is easily obscured by dense weeds, and millimeter-wave radar fails to capture sufficient terrain detail. This can distort terrain data and affect the accuracy of subsequent risk assessments and path corrections.

[0032] In this regard, the present application further proposes the use of a fusion perception method of vehicle-mounted lidar and millimeter-wave radar. The lidar collects three-dimensional point cloud data of the terrain with a preset grid accuracy, and the millimeter-wave radar captures the surface micro-topography undulations in real time; the collected point cloud data is filtered to remove interference points including but not limited to weeds and stones, and retain valid data reflecting the real terrain; the slope value and aspect angle of each grid are calculated to generate a terrain slope heat map, in which areas where the slope exceeds the set threshold are marked as key perception areas to increase the perception frequency of this area.

[0033] This application uses LiDAR to collect three-dimensional point cloud data of terrain with a preset grid accuracy (e.g., 10 cm x 10 cm grid cells), thereby accurately capturing large-scale terrain undulations, such as the elevation changes of ridges and large depressions. Millimeter-wave radar, with its strong penetrating power, captures surface micro-topography in real time, such as subtle height differences in the soil surface caused by tillage and localized bulges caused by small rocks. The two data complement each other to fully reflect the terrain conditions. Secondly, a statistical filtering algorithm is used to calculate the distance deviation of each point cloud's neighboring points. Points with deviations outside a reasonable range are identified as weeds or rocks and removed. Radius filtering is then used to remove isolated discrete points, retaining continuous and valid data reflecting the actual terrain, providing a reliable basis for subsequent slope calculations. Finally, the slope value is calculated by the elevation difference and horizontal distance between adjacent points in the grid. The aspect angle indicates the direction of the slope (e.g., 0° is due north, 90° is due east). Slope values ​​are plotted on a grid map using a color gradient to create a heat map, visually displaying slope distribution. Areas with slopes exceeding a set threshold (e.g., 15°) are designated as key sensing areas, with the sensing frequency increased from 2 per second to 5 per second, ensuring real-time tracking of high-risk terrain changes.

[0034] The innovation of this application lies in that it solves the problems of incomplete perception of a single sensor, excessive data interference, and delayed perception in high-risk areas. Through multi-sensor fusion, data filtering optimization, and dynamic adjustment of perception frequency, it helps to accurately obtain the macro-undulations and micro-changes of farmland terrain, providing high-quality basic data for subsequent terrain risk identification and path pre-correction, and ensuring the accuracy and safety of unmanned agricultural machinery navigation in complex farmland environments.

[0035] In some embodiments of this application, traditional terrain risk identification methods typically rely on a single parameter (such as slope) for judgment, ignoring the coupling between the dynamic characteristics of terrain undulation and the posture of agricultural machinery. For example, continuous undulations with short wavelengths and large amplitudes (such as in areas with intersecting ridges and furrows) can cause periodic rolls in agricultural machinery, but slope alone cannot accurately identify such risks.

[0036] In this regard, the present application further proposes to extract characteristic parameters of terrain undulation, including the wavelength, amplitude and slope change rate of continuous undulations, where areas with wavelengths less than a preset multiple of the agricultural machinery wheelbase and amplitudes greater than a set value are marked as high-disturbance potential areas; combined with the current posture data of the agricultural machinery, the roll moment coefficient of the agricultural machinery when it is driving in the area is calculated, and areas where the roll moment coefficient exceeds the set threshold are determined as potential risk areas for terrain disturbance; risk areas are graded and divided into multiple levels according to the roll moment coefficient, and different levels of areas correspond to different prediction accuracy requirements.

[0037] The working principle of this application is that the system first extracts terrain features from the filtered point cloud data: Wavelength of continuous undulations: Calculate the horizontal distance between adjacent wave crests (or troughs). Areas with a wavelength less than 1.5 times (preset multiple) the wheelbase of the agricultural machinery indicate that the frequency of terrain changes is faster than the responsiveness of the agricultural machinery's wheelbase, which can easily lead to one-side tire suspension or uneven force.

[0038] Amplitude: The vertical height difference between the peak and the trough. Areas with an amplitude greater than 15cm (set value) may cause the agricultural machinery to shake violently or tilt.

[0039] Slope change rate: the amount of change in slope within a unit distance. Areas with a large rate of change (such as the junction of steep slopes and gentle slopes) will cause the center of gravity of agricultural machinery to shift rapidly.

[0040] When a region satisfies both the wavelength being smaller than the threshold and the amplitude being larger than the threshold, it is marked as a high disturbance potential region.

[0041] Furthermore, combined with the agricultural machinery posture data (tilt angle, pitch angle) collected in real time by the IMU, the system establishes a dynamic model to calculate the roll moment coefficient. The specific calculation formula is: , where h is the height of the center of gravity of the agricultural machinery, g is the acceleration of gravity, is the current roll angle, is the wheelbase, is the current pitch angle. When K is greater than the preset threshold, it indicates that the terrain in this area may cause a significant increase in the risk of agricultural machinery rollover, and it is determined to be a potential risk area.

[0042] Furthermore, the risk regions are divided according to the roll moment coefficient into: low-risk region (K ≤ 0.3): the prediction accuracy requirement is ±10%; medium-risk region (0.3 < K ≤ 0.6): the prediction accuracy requirement is ±5%; high-risk region (K > 0.6): the prediction accuracy requirement is ±2%. Different levels of regions adopt different data sampling frequencies and algorithm complexities. For example, the high-resolution mode of the millimeter-wave radar will be enabled in the high-risk region, and the number of Kalman filter iterations will be increased.

[0043] The innovation of this application lies in that through multi-parameter fusion analysis, it solves the problem of risk misjudgment caused by traditional single-parameter judgment. It can not only capture the high-frequency undulating terrain risks ignored by traditional methods, realize short-wavelength terrain recognition, but also combine real-time attitude data for dynamic risk assessment, more accurately reflect the actual operating risks of agricultural machinery, and at the same time dynamically adjust resource allocation according to the risk level, thus significantly enhancing the adaptability of driverless agricultural machinery to complex terrains.

[0044] In some embodiments of this application, traditional agricultural machinery navigation correction models mostly use fixed parameters or single terrain factors for compensation, ignoring the coupling relationship between terrain dynamic changes and the dynamic characteristics of the agricultural machinery itself. For example, only adjusting the steering according to the slope without considering the height of the center of gravity of the agricultural machinery will result in insufficient compensation; using fixed wheel speed parameters to deal with undulating terrain is prone to rollover risks.

[0045] In response to this, this application further proposes that the input parameters of the pre-correction model include the trajectory offset of terrain prediction, rollover risk value, wheelbase of the agricultural machinery, center of gravity height, suspension system stiffness, and current wheel speed. Specifically, the input of the pre-correction model covers two types of key data: one is terrain prediction data, including the trajectory offset of terrain prediction (the expected deviation distance of the agricultural machinery caused by the terrain) and rollover risk value (the risk probability calculated based on slope and undulating characteristics); the other is the dynamic parameters of the agricultural machinery, including the wheelbase of the agricultural machinery (the distance between the left and right wheels, affecting rollover stability), center of gravity height (determining the torque size during tilting), suspension system stiffness (reflecting shock absorption ability), and current wheel speed (affecting the balance between inertial force and terrain force). These parameters jointly construct the basic data set of the interaction between terrain and agricultural machinery. The core algorithm of the model is the compensation calculation logic based on fuzzy PID control, where the terrain compensation amount is positively correlated with the slope change rate and negatively correlated with the wheel speed of the agricultural machinery. Specifically, fuzzy PID combines the flexibility of fuzzy control with the precision of PID control, dynamically adjusting control parameters through preset fuzzy rules (such as "large slope change rate → large compensation increase" and "high wheel speed → limited compensation increase"). The calculation of terrain compensation follows a dual correlation rule: a positive correlation with the slope change rate (the more dramatic the slope change, the greater the compensation, such as increasing the steering compensation angle on steep slopes); and a negative correlation with the wheel speed. Higher wheel speeds result in smaller compensation increases, helping to prevent loss of control caused by over-adjustment at high speeds.

[0046] The output includes the steering angle compensation value of the pre-corrected path, the wheel speed adjustment coefficient and the suspension system damping parameters, which work together to offset the impact of terrain disturbances.

[0047] Specifically, the model outputs three key compensation parameters: one is the steering angle compensation value: used to adjust the steering mechanism of the agricultural machinery to offset the trajectory deviation caused by the terrain, such as increasing the steering compensation angle to the right on a left-slope terrain; the second is the wheel speed adjustment coefficient: by reducing or increasing the drive wheel speed (such as the coefficient in the steep slope area is less than 1, slowing down the driving speed), balancing the inertia force and terrain resistance; the third is the suspension system damping parameter: increasing or decreasing the shock absorption force of the suspension system (such as increasing the damping on undulating terrain to reduce vehicle body bumps). The three work together to synchronously compensate for steering, power, and posture, forming an all-round offset to terrain disturbances.

[0048] The innovations of this application are: first, it overcomes the "inadequate terrain adaptation" problem caused by fixed parameter compensation and dynamically generates compensation that matches the terrain; second, it avoids the one-sidedness of single-factor compensation and improves compensation accuracy through multi-parameter fusion; and third, it implements "early compensation" rather than "delayed correction," reducing trajectory deviation and rollover risk. For example, in complex farmland scenarios such as hills and terraces, the deviation between the actual trajectory of the agricultural machinery and the pre-corrected path can be reduced to within a preset value, reducing the risk of rollover and helping to improve the terrain adaptability and operational safety of unmanned agricultural machinery.

[0049] In some embodiments of the present application, the traditional terrain risk prediction method relies only on a single initial perception data for calculation, without combining it with historical operation data for verification, and the sampling density is fixed, which can easily lead to large deviations between the prediction results and the actual results - for example, due to seasonal farming changes in the same area (such as changes in ridge furrow depth after spring plowing), the initial predicted trajectory offset may be significantly different from the historical data, but it cannot be dynamically corrected, which directly affects the navigation accuracy.

[0050] In this regard, the present application further proposes to calculate the slope gradient and curvature of the terrain based on the initial perception of terrain data, and predict the trajectory offset and rollover risk value of agricultural machinery when passing through the area; call the terrain-trajectory correlation data of historical operations in the same area, and calculate the deviation rate between the offset of this prediction and the historical average offset; if the deviation rate exceeds the preset threshold, increase the sampling density of terrain perception, recalculate the slope gradient and curvature, correct the prediction model parameters and predict again until the deviation rate is lower than the threshold, and use the risk value and offset at this time as the valid prediction result.

[0051] The application works by first using initial terrain data collected by LiDAR and millimeter-wave radar to calculate the terrain's slope gradient (the change in slope per unit distance, reflecting the rate of change in terrain steepness) and curvature (the degree of curvature of the terrain surface; greater curvature indicates more severe terrain fluctuations). A dynamic model combines this slope gradient and curvature with agricultural machinery parameters (such as wheelbase and center of gravity height) to preliminarily predict the trajectory deviation (such as the lateral offset distance caused by a raised tire on one side) and rollover risk (the risk probability calculated based on the roll moment coefficient) of the agricultural machinery as it passes through the area. Secondly, the terrain-trajectory correlation database of historical operations in the area (including the predicted offset, actual offset, and rollover risk records under the same historical terrain conditions) is called to calculate the deviation rate between the initial predicted offset and the historical average offset: For example, if historical data shows that the average offset in a certain area is 0.3 meters and the initial prediction is 0.5 meters, the deviation rate is 66.7%; Finally, if the deviation rate exceeds a preset threshold (e.g., 30%), this indicates that the initial perception data may have missed subtle changes in the terrain (e.g., a new ridge). In this case, the terrain perception sampling density is increased (e.g., increasing the LiDAR grid resolution from 10 cm x 10 cm to 5 cm x 5 cm), data is recollected, and the slope gradient and curvature are calculated. Simultaneously, the prediction model parameters are modified based on the new data (e.g., adjusting the weight of the terrain curvature on the offset), and the prediction is repeated. This process is repeated until the deviation rate falls below the threshold, and the risk value and offset at this point are finally output as the valid result. The innovation of this application lies in its closed-loop mechanism of initial prediction, historical data verification, and dynamic adjustment, which addresses the problems of traditional prediction, such as large deviations caused by reliance on single-shot data, a lack of historical verification mechanisms, and a fixed sampling density that cannot adapt to terrain changes. This ensures that the predicted trajectory offset and rollover risk values ​​dynamically adapt to the actual terrain conditions. In complex farmland environments such as hills and terraces, this significantly improves the accuracy and safety of unmanned agricultural machinery navigation, providing a reliable basis for risk prediction for stable agricultural machinery operations.

[0052] In some embodiments of the present application, in the process of calling the pre-correction model to generate a path and adjust parameters, traditional methods often directly use the initial model output results without considering the parameter adaptation deviation under complex terrain, which may cause the generated path to have potential rollover risks or excessive energy consumption. For example, when the initial terrain compensation amount is set unreasonably, the simulated rollover risk of agricultural machinery driving on a steep slope may exceed the safety range, or the energy consumption may surge due to frequent adjustment of power parameters.

[0053] In this regard, the present application further proposes a pre-correction model that combines the wheelbase, center of gravity height and terrain compensation of the agricultural machinery to generate an initial pre-correction path and control parameters; simulates the agricultural machinery traveling along the initial path in a virtual terrain environment, and calculates the rollover risk coefficient and energy consumption index of the simulated trajectory; if the rollover risk coefficient is higher than the safety threshold or the energy consumption index exceeds the upper limit, the weight ratio of the terrain compensation amount is adjusted, the path and parameters are regenerated, and the simulation verification is repeated until both indicators meet the standards to determine the final pre-correction path.

[0054] The working principle of this application is that the pre-correction model first combines the wheelbase of the agricultural machinery (affecting lateral stability), the center of gravity height (determines the rollover threshold) and the terrain compensation amount (correction value to offset the terrain undulations) to generate the initial pre-correction path (such as the steering angle compensation path for slopes) and control parameters (such as throttle opening and steering sensitivity). Then, the driving of agricultural machinery is simulated in a virtual terrain environment: a virtual scene is constructed based on the three-dimensional terrain data scanned by the lidar, the initial path and control parameters are input, and the rollover risk coefficient (comprehensive tilt angle, quantitative value of tire grip) and energy consumption index (total power output per unit distance) of the agricultural machinery on each road section are calculated through dynamic simulation. If the rollover risk coefficient in the simulation results exceeds a safety threshold (e.g., 0.7) or the energy consumption index exceeds a preset upper limit (e.g., 1.2 kWh per kilometer), the weighting of terrain compensation is adjusted (e.g., increasing the weight of slope compensation and decreasing the weight of undulation compensation), the path and parameters are regenerated, and the virtual simulation is repeated. This process is repeated until the rollover risk coefficient falls below the safety threshold and the energy consumption index is within a reasonable range. The final path is the pre-corrected path that meets both safety and efficiency requirements.

[0055] The innovation of this application lies in that the cyclic verification step effectively solves the potential safety hazards and excessive energy consumption problems of the initial pre-corrected path. Through virtual simulation, parameter adaptation defects are exposed in advance and iterative optimization is carried out, which is conducive to ensuring the driving safety of agricultural machinery in complex terrain, while avoiding ineffective energy waste, and improving the path adaptation accuracy and operation economy of unmanned agricultural machinery in complex farmland environments such as hilly and sloping land.

[0056] In some embodiments of the present application, in the process of real-time correction of agricultural machinery navigation parameters, the traditional single adjustment method is prone to problems of insufficient or excessive parameter correction. For example, only a proportional adjustment is performed based on the trajectory deviation, which may cause the deviation to continue to expand due to the complexity of the terrain, and ignoring the consistency between actual risk and predicted risk will make the correction direction disconnected from actual needs, affecting navigation accuracy and safety.

[0057] In this regard, the present application further proposes that when the deviation between the actual trajectory and the pre-corrected path exceeds a first threshold, the steering angle and wheel speed parameters are adjusted proportionally; the rate of change of the adjusted trajectory deviation is continuously monitored, and if the rate of change is positive (the deviation expands), a second round of adjustment is entered, and the steering angle adjustment amplitude is doubled; after each round of adjustment, the degree of consistency between the actual rollover risk value and the terrain predicted risk value is calculated. If the degree of consistency is lower than the preset standard, the vehicle returns to re-identify the terrain risk area, updates the pre-corrected model parameters, and adjusts again until both the deviation and the degree of consistency meet the requirements.

[0058] The working principle of this application is that when the deviation between the actual trajectory and the pre-corrected path exceeds a first threshold (such as 0.2 meters), the system proportionally adjusts the steering angle and wheel speed parameters - for example, for every 0.1 meter increase in deviation, the steering angle is increased by 1°, and at the same time, the outer wheel speed is appropriately reduced to suppress the deviation, thus reducing the trajectory deviation through preliminary adjustments; Secondly, the system continuously monitors the rate of change of the trajectory deviation (the increase or decrease in the deviation per unit time) after adjustment. If the rate of change is positive (i.e., the deviation is still increasing), it indicates that the initial adjustment was insufficient. The system then initiates a second round of adjustments, doubling the steering angle adjustment (e.g., increasing the original 1° increase by 2°) and simultaneously adjusting the wheel speed differential. This is done by strengthening corrective measures to curb the trend of increasing deviation. Furthermore, after each round of adjustments, the system calculates the degree of agreement between the actual rollover risk value (calculated based on the real-time posture data of the agricultural machinery) and the terrain prediction risk value (e.g., the difference between the two as a percentage of the predicted value). If the agreement falls below a preset standard (e.g., below 80%), it indicates a deviation in the terrain prediction or model parameters. The system must return to the terrain risk identification stage, recollect terrain data, update the pre-correction model parameters (e.g., adjust the terrain compensation weight), and then make adjustments based on the new parameters until the trajectory deviation is within the threshold and the risk agreement is met. The innovation of this application lies in that this mechanism effectively solves the problems of insufficient adjustment force, directional deviation and risk disconnection in traditional parameter correction. Through multiple rounds of progressive adjustment and risk cross-validation, it is helpful to ensure that the actual trajectory of agricultural machinery in complex terrain is highly consistent with the pre-corrected path, while ensuring the accuracy of risk assessment, which helps to significantly improve the stability and safety of unmanned agricultural machinery navigation, and is especially suitable for farmland operation scenarios with varied terrain such as hills and terraces.

[0059] In some embodiments of the present application, when generating a pre-corrected path, if the navigation parameters suddenly change in adjacent areas (such as a sudden increase or decrease in the steering angle or wheel speed), it will cause the driving posture of the agricultural machinery to fluctuate violently, which will not only affect the operation accuracy (such as uneven sowing row spacing), but may also cause the risk of rollover due to inertia force. Especially in areas where the terrain changes continuously, the traditional parameter division method is difficult to adapt to the gradual characteristics of the terrain.

[0060] In this regard, the present application further proposes to divide the farmland into multiple continuous navigation adjustment zones, each adjustment zone corresponding to a set of preset navigation parameters; calculate the similarity of terrain features between adjacent adjustment zones, and if the similarity exceeds a set threshold, merge the adjacent adjustment zones into a parameter slow-changing zone; within the parameter slow-changing zone, the navigation parameters are smoothly transitioned according to the principle of monotonically increasing / decreasing the detection parameter adjustment amplitude to avoid parameter mutations.

[0061] The application works by dividing farmland into multiple continuous navigation adjustment zones, each corresponding to a set of preset navigation parameters. For example, zones are divided into 5m x 5m units. Based on the average slope and undulation frequency of the terrain in that area, initial steering angle compensation values ​​(e.g., a 3° slope corresponds to a 2° steering angle) and wheel speed reference values ​​(e.g., a wheel speed of 1.2m / s on flat terrain) are preset to provide a basic unit for parameter adjustment.

[0062] Calculate the similarity of terrain features between adjacent adjustment zones by comparing the degree of agreement (e.g., using a cosine similarity algorithm) of core parameters such as slope gradient, curvature change rate, and undulation wavelength. If the similarity exceeds a set threshold (e.g., 80%), indicating similar terrain features between the two zones, they are merged into a parameter-slowly changing zone. For example, if two adjacent zones both have gentle slopes and the slope difference is less than 2°, they are merged into the same slowly changing zone. Within the slowly changing parameter range, navigation parameters transition smoothly according to the principle of monotonically increasing / decreasing the adjustment range of the detected parameters. For example, within the slowly changing range from a flat area to a 5° gentle slope, the steering angle compensation value gradually increases from 0° to 3°, and the wheel speed gradually decreases from 1.2m / s to 0.9m / s. This gradual parameter transition is achieved through linear interpolation or exponential decay algorithms, which helps avoid machine posture oscillation caused by sudden parameter changes. The innovation of this application is that the above steps solve the problem of sudden adjustment in traditional parameter division: by associating and merging areas through terrain similarity, the navigation parameters are naturally transitioned with the terrain characteristics, reducing the fluctuation of agricultural machinery posture, ensuring the operation accuracy under complex terrain (such as the sowing deviation of hilly terraces can be controlled within ±3cm), and reducing the mechanical loss and safety risks caused by sudden changes in parameters. It is especially suitable for farmland with continuously changing terrain (such as gently sloping farmland and stepped terraces), which is beneficial to significantly improve the smoothness and reliability of the pre-corrected path.

[0063] In some embodiments of the present application, when performing parameter adjustments in potential risk areas of terrain disturbance, traditional methods often adopt a point-by-point independent adjustment strategy for densely distributed risk points, which can easily lead to frequent sudden changes in navigation parameters (such as large fluctuations in steering angle and wheel speed within a short distance), causing oscillation of agricultural machinery posture, and ignoring the spatial correlation between risk points, which may lead to adjustment direction conflicts (such as opposite steering directions required by adjacent points), reducing navigation stability and operation accuracy.

[0064] In this regard, the present application further proposes to calculate the spatial spacing between the risk points in the area, and identify dense risk point clusters whose spacing is less than a preset ratio of the agricultural machinery wheelbase; execute on the risk point clusters: generate a parameter adjustment path based on the principle of minimum gradient of navigation parameter changes between adjacent risk points; on the parameter adjustment path, use a PID controller to fine-tune the parameters so that the deviation rate between the actual adjustment amplitude and the theoretical prediction value does not exceed the preset range.

[0065] This application works by first calculating the horizontal distance between any two risk points using the Euclidean distance formula. Risk points with a distance less than a preset ratio of the machine's wheelbase (e.g., 1 / 2 the wheelbase) are then grouped into a cluster. For example, if the machine's wheelbase is 2 meters, risk points with a distance less than 1 meter are considered densely distributed and require coordinated processing. This helps avoid parameter confusion caused by point-by-point adjustments.

[0066] Secondly, parameter adjustment paths are generated for risk point clusters. Based on the principle of minimizing the gradient of navigation parameter changes between adjacent risk points (i.e., minimizing the rate of change of parameter adjustments), a continuous adjustment path is planned. For example, if two adjacent risk points within a cluster require steering angle adjustments of +3° and +5°, respectively, the adjustment path will smoothly transition from +3° to +4° to +5°, rather than abruptly jumping. This ensures uniform changes in the machine's posture and reduces inertial shock. The PID controller then dynamically adjusts parameters such as steering angle and wheel speed through the coordinated action of the proportional phase (real-time deviation correction), the integral phase (eliminating accumulated errors), and the differential phase (suppressing overshoot). For example, if the actual adjustment deviation from the theoretical prediction reaches 4% (the preset threshold is 5%), the controller automatically adjusts the output to keep the deviation below 5%, ensuring the accuracy and stability of parameter adjustments.

[0067] The innovation of this application lies in its ability to solve the problems of parameter mutations and adjustment conflicts in traditional dense risk point processing: collaborative adjustment is achieved by identifying risk point clusters, which is conducive to avoiding frequent parameter fluctuations; paths are generated based on the gradient minimum principle to ensure smooth transition of agricultural machinery posture; combined with PID fine-tuning control accuracy, it ultimately improves the navigation stability and operation accuracy of unmanned agricultural machinery in complex and dense risk areas, which is conducive to reducing the risk of trajectory deviation and rollover caused by improper parameter adjustment.

[0068] In some embodiments of the present application, in the process of comparing the actual trajectory with the pre-corrected path deviation in real time and iteratively correcting the parameters, if the parameter adjustment amplitudes of multiple consecutive adjacent adjustment areas exceed the threshold, it means that the current terrain division may not fully reflect the terrain continuity in the area, which may easily lead to frequent fluctuations in parameter adjustments and affect the driving stability and operation accuracy of agricultural machinery.

[0069] In this regard, the present application further proposes that when the parameter adjustment amplitudes of three or more adjacent adjustment areas are continuously detected to exceed the threshold, a terrain similarity re-evaluation mechanism is triggered; the terrain features of the area are re-extracted, and the similarity is matched with the terrain features in the historical operation data to identify unmarked terrain similar parts; if there are unmarked similar parts, the area is included in the parameter slow-changing area management, and the fuzzy adaptive control algorithm is used to dynamically adjust the parameter change rate so that the parameter adjustment amplitude conforms to the monotonically increasing / decreasing law.

[0070] The working principle of this application is that when the parameter adjustment amplitudes of three or more consecutive adjustment areas are detected to exceed the threshold, the system determines that the current area may have insufficiently identified terrain correlations and immediately triggers the terrain similarity reassessment mechanism. This secondary verification of the rationality of the existing terrain division through the evaluation mechanism is helpful to avoid parameter adjustment anomalies caused by insufficient initial evaluation; Secondly, the terrain features of the area (such as slope gradient, undulation wavelength, curvature, etc.) are re-extracted and matched with the terrain features stored in the historical operation data in a multi-dimensional similarity manner (for example, using the cosine similarity algorithm to calculate the matching degree of the feature vectors). This matching can identify areas that were not marked in the initial division but actually have significant terrain similarities with the surrounding areas, such as an area in a continuous gentle slope that was mistakenly divided into multiple independent adjustment areas; Specifically, in areas with continuous gentle slopes, fuzzy logic reasoning is used to optimize the adjustment range of the steering angle and wheel speed in real time to ensure that the parameter adjustment strictly follows the monotonic increasing / decreasing law (for example, as the slope increases, the steering compensation angle continues to increase steadily). This helps avoid jumps or reverse adjustments, allowing agricultural machinery to maintain a smooth driving state in complex and continuous terrain.

[0071] The innovation of this application lies in addressing the problem of excessive parameter adjustments and frequent fluctuations caused by insufficient initial terrain similarity assessment. By dynamically identifying unlabeled similar terrain and incorporating it into slowly changing zones, combined with fuzzy adaptive control to achieve smooth parameter transitions, this significantly improves the navigation stability and operating accuracy of unmanned agricultural machinery in complex and continuous terrain. This is particularly suitable for agricultural operations in hilly and terraced fields with diverse terrain and large areas of continuous similar features.

[0072] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A method for unmanned agricultural machinery path navigation based on farmland environment perception and judgment, characterized in that: include: Before and during agricultural machinery operations, it continuously senses the undulating features of farmland terrain, slope changes, and the current posture and wheel speed of the agricultural machinery; Identify potential risk areas of terrain disturbance based on perception data and predict the impact of terrain fluctuations on the trajectory of agricultural machinery and the risk of rollover. Calling a pre-correction model that integrates terrain prediction data and agricultural machinery dynamic characteristics to generate a pre-corrected path with terrain compensation and adjust navigation control parameters in advance; Control the agricultural machinery to travel along the pre-corrected path, compare the actual trajectory with the pre-corrected path deviation in real time, iterate and correct the parameters, cross-validate the consistency between the terrain prediction risk value and the actual posture change, and dynamically update the navigation strategy.

2. The method for unmanned agricultural machinery path navigation based on farmland environment perception and judgment according to claim 1 is characterized in that: Continuously sensing farmland terrain undulations and slope changes includes: The system uses a fusion sensing method of vehicle-mounted LiDAR and millimeter-wave radar. The LiDAR collects three-dimensional point cloud data of the terrain with a preset grid accuracy, while the millimeter-wave radar captures the surface micro-topography in real time. Filter the collected point cloud data to remove interference points including but not limited to weeds and rocks, and retain valid data reflecting the actual terrain; The slope value and aspect angle of each grid are calculated to generate a terrain slope heat map, where areas with slopes exceeding the set threshold are marked as key perception areas to increase the perception frequency of these areas.

3. The method for unmanned agricultural machinery path navigation based on farmland environment perception and judgment according to claim 1 is characterized in that: Identifying potential risk areas of terrain disturbance based on perception data includes: Extract terrain undulation characteristic parameters, including the wavelength, amplitude, and slope change rate of continuous undulations. Areas with wavelengths smaller than a preset multiple of the agricultural machinery wheelbase and amplitudes greater than a set value are marked as high-disturbance potential areas. Combined with the current posture data of the agricultural machinery, the roll moment coefficient of the agricultural machinery when driving in the area is calculated. The area where the roll moment coefficient exceeds the set threshold is determined as a potential risk area for terrain disturbance; Risk areas are graded and divided into multiple levels according to the roll moment coefficient. Different levels of areas correspond to different prediction accuracy requirements.

4. The method for unmanned agricultural machinery path navigation based on farmland environment perception and judgment according to claim 1 is characterized in that: A pre-correction model that integrates terrain prediction data with the dynamic characteristics of agricultural machinery, specifically including: Input parameters include the trajectory deviation of terrain prediction, rollover risk value, the wheelbase of the agricultural machine, the center of gravity height, the stiffness of the suspension system and the current wheel speed; The core algorithm of the model is the compensation calculation logic based on fuzzy PID control, in which the terrain compensation amount is positively correlated with the slope change rate and negatively correlated with the wheel speed of the agricultural machinery; The output includes the steering angle compensation value of the pre-corrected path, the wheel speed adjustment coefficient and the suspension system damping parameters, which work together to offset the impact of terrain disturbances.

5. The method for unmanned agricultural machinery path navigation based on farmland environment perception and judgment according to claim 1 is characterized in that: Identifying potential risk areas of terrain disturbance and predicting the offset and rollover risk values ​​includes: Based on the initially perceived terrain data, the slope gradient and curvature of the terrain are calculated, and the trajectory deviation and rollover risk of the agricultural machinery passing through the area are predicted; Call the terrain-trajectory correlation data of historical operations in the same area and calculate the deviation rate between the current predicted offset and the historical average offset; If the deviation rate exceeds the preset threshold, the sampling density of terrain perception is increased, the slope gradient and curvature are recalculated, the prediction model parameters are corrected, and the prediction is repeated until the deviation rate is lower than the threshold. The risk value and offset at this time are used as the valid prediction results.

6. The method for unmanned agricultural machinery path navigation based on farmland environment perception and judgment according to claim 4 is characterized in that: When calling the pre-correction model to generate the pre-correction path and adjust the control parameters, a cyclic verification step is included: The pre-correction model combines the wheelbase, center of gravity height and terrain compensation of the agricultural machinery to generate the initial pre-correction path and control parameters; Simulate the agricultural machinery to drive along the initial path in a virtual terrain environment and calculate the rollover risk coefficient and energy consumption index of the simulated trajectory; If the rollover risk factor is higher than the safety threshold or the energy consumption index exceeds the upper limit, the weight ratio of the terrain compensation amount is adjusted, the path and parameters are regenerated, and the simulation verification is repeated until both indicators meet the standards and the final pre-corrected path is determined.

7. The method for unmanned agricultural machinery path navigation based on farmland environment perception and judgment according to claim 3 is characterized in that: When comparing deviations in real time and iteratively correcting parameters, multiple rounds of adjustments are included: When the deviation between the actual trajectory and the pre-corrected path exceeds a first threshold, adjusting the steering angle and wheel speed parameters proportionally; Continuously monitor the rate of change of the adjusted trajectory deviation. If the rate of change is positive, enter the second round of adjustment and double the steering angle adjustment; After each round of adjustment, the degree of consistency between the actual rollover risk value and the terrain predicted risk value is calculated. If the degree of consistency is lower than the preset standard, the system returns to re-identify the terrain risk area, updates the pre-corrected model parameters, and adjusts again until both the deviation and the degree of consistency meet the requirements.

8. The method for unmanned agricultural machinery path navigation based on farmland environment perception and judgment according to claim 1 is characterized in that: When generating the pre-corrected path, a parameter adjustment amplitude optimization step based on terrain similarity is also included: The farmland is divided into a plurality of continuous navigation adjustment zones, each of which corresponds to a set of preset navigation parameters; Calculate the similarity of terrain features between adjacent adjustment areas. If the similarity exceeds the set threshold, the adjacent adjustment areas will be merged into a parameter slow-changing area. In the parameter slowly changing area, the navigation parameters are smoothly transitioned according to the principle of monotonically increasing / decreasing the adjustment amplitude of the detection parameters to avoid sudden parameter changes.

9. The method for unmanned agricultural machinery path navigation based on farmland environment perception and judgment according to claim 8 is characterized in that: When performing parameter adjustments for areas with potential risk of terrain disturbance, the following are also included: Calculate the spatial distance between risk points in the area and identify dense clusters of risk points whose distance is less than a preset ratio of the wheelbase of agricultural machinery; Execution for risk point clusters: Generate parameter adjustment paths based on the principle of minimum gradient of navigation parameter changes between adjacent risk points; On the parameter adjustment path, a PID controller is used to fine-tune the parameters so that the deviation rate between the actual adjustment range and the theoretical predicted value does not exceed the preset range.

10. The method for unmanned agricultural machinery path navigation based on farmland environment perception and judgment according to claim 1, characterized in that: When comparing the actual trajectory with the pre-corrected path deviation in real time and iteratively correcting the parameters, it also includes: When the parameter adjustment amplitudes of three or more adjacent adjustment areas are detected to exceed the threshold, the terrain similarity reassessment mechanism is triggered; Re-extract the terrain features of the area, perform similarity matching with the terrain features in historical operation data, and identify similar unmarked terrain parts; If there are unmarked similar parts, the area will be included in the parameter slow change area management, and the fuzzy adaptive control algorithm will be used to dynamically adjust the parameter change rate so that the parameter adjustment amplitude conforms to the monotonically increasing / decreasing law.

Images