A multi-modal collaborative tractor intelligent chassis control method and system

Abstract

This application discloses a multimodal collaborative intelligent chassis control method and system for tractors. The method includes: acquiring visual image data, chassis pressure data, and position coordinate data, and fusing the above data to construct an environmental state matrix; determining the obstacle type based on the environmental state matrix, acquiring a set of obstacle characteristic parameters, calculating the collision risk between the tractor and the obstacle, and calculating a risk coefficient; if the risk coefficient is greater than a preset safety threshold, calculating the suspension stiffness and damping coefficient based on the obstacle type and terrain characteristics to obtain suspension control parameters; acquiring surface hardness data for the area to be operated, and adjusting the suspension control parameters based on the surface hardness data to generate a chassis response parameter set. This method solves the problem of tractor chassis being unable to adaptively adjust chassis parameters in complex terrain, achieving the effect of enabling tractors to automatically adjust chassis parameters based on real-time data, thereby improving operating efficiency and safety.

Description

Technical Field

[0001] The present application relates to the field of agricultural power machinery, and in particular to a multi-modal collaborative tractor intelligent chassis control method and system. Background Art

[0002] As a core driver of modern agricultural development, intelligent agricultural machinery is directly linked to improvements in agricultural production efficiency and operational safety. Tractors, the most important power machinery in agricultural production, are crucial for the advancement of their intelligent chassis control technology, which determines the overall progress of agricultural mechanization.

[0003] Current tractor chassis control methods primarily rely on single sensors or simple mechanical adjustments, lacking the ability to integrate multi-source information and intelligent decision-making. Typical control methods include local adjustments based on inclination or speed sensors. For example, these can measure the vehicle's tilt angle to adjust the suspension hydraulic cylinder's stroke or adjust the transmission gear according to a preset speed.

[0004] As agricultural operations become increasingly complex and precision requirements continue to rise, current tractor chassis control methods are no longer able to meet the demands of modern agricultural production. Relying on single sensors or fixed mechanical adjustments, they lack multi-source environmental perception and intelligent collaborative decision-making capabilities. This makes it impossible to accurately and adaptively adjust chassis parameters in real time during complex farmland operations, severely limiting operational efficiency and safety. Summary of the Invention

[0005] The embodiments of the present application provide a multi-modal collaborative tractor intelligent chassis control method and system, which solves the problem that the tractor chassis cannot adaptively adjust chassis parameters in complex terrain, and achieves the effect of enabling the tractor to automatically adjust chassis parameters according to real-time data, thereby improving operating efficiency and safety.

[0006] The embodiment of the present application provides a multi-modal collaborative tractor intelligent chassis control method and system. The multi-modal collaborative tractor intelligent chassis control method includes:

[0007] Acquire an original data set including visual image data, chassis pressure data, and position coordinate data, and perform data fusion on the original data set to construct an environmental state matrix;

[0008] Determining an obstacle type according to the environmental state matrix, and obtaining an obstacle characteristic parameter set of the obstacle type;

[0009] Calculating the collision risk between the tractor and the obstacle based on the obstacle characteristic parameter set, and calculating the risk coefficient;

[0010] If the risk factor is greater than a preset safety threshold, the suspension stiffness and damping coefficient are calculated according to the obstacle type and terrain characteristics to obtain suspension control parameters;

[0011] Surface hardness data of the area to be operated is obtained, and the suspension control parameters are adjusted according to the surface hardness data to generate a chassis response parameter set.

[0012] Optionally, the step of acquiring an original data set including visual image data, chassis pressure data, and position coordinate data, and performing data fusion on the original data set to generate an environmental state matrix includes:

[0013] determining terrain edge features according to the visual image data, and determining a terrain type based on the terrain edge features;

[0014] Acquiring terrain features of the terrain type, and associating the terrain features with the chassis pressure data and the position coordinate data to generate a fused data set;

[0015] Calculating a spatial position offset of the tractor based on the fused dataset, and if the spatial position offset is greater than a preset offset threshold, adjusting a weight of the terrain feature and updating the fused dataset;

[0016] The environmental state matrix is ​​generated according to the fused data set.

[0017] Optionally, the step of generating the environmental state matrix according to the fused data set includes:

[0018] Dynamically adjusting image edge detection parameters based on the multidimensional data in the fused data set, and performing edge detection on the visual image data;

[0019] If it is detected that there is an area in the image where the height change exceeds a preset change threshold, the area is marked as a terrain relief area to obtain a terrain relief mark set;

[0020] The terrain relief mark set is time-synchronized and associated with the corresponding chassis pressure data and the position coordinate data to generate the environmental state matrix.

[0021] Optionally, the step of determining the obstacle type according to the environmental state matrix and obtaining a set of obstacle characteristic parameters of the obstacle type includes:

[0022] Determine an obstacle and its type based on the visual feature vector, spatial position data, and height gradient information in the environmental state matrix;

[0023] The vertical height difference and horizontal distance between the obstacle and the current position of the tractor are obtained, and the obstacle feature parameter set including the obstacle type, the vertical height difference and the horizontal distance is generated.

[0024] Optionally, the step of calculating the collision risk between the tractor and the obstacle based on the obstacle characteristic parameter set and calculating the risk coefficient includes:

[0025] Comparing the vertical height difference and horizontal distance in the obstacle characteristic parameter set with the corresponding preset distance safety threshold to determine whether there is a collision risk;

[0026] If there is a collision risk, the real-time driving speed of the tractor is obtained, and the real-time driving speed is associated with the vertical height difference and the horizontal distance to generate a comprehensive risk parameter combination;

[0027] The parameters in the comprehensive risk parameter combination are adjusted according to preset weight rules to generate the risk coefficient.

[0028] Optionally, if the risk factor is greater than a preset safety threshold, the step of calculating the suspension stiffness and damping coefficient according to the obstacle type and terrain characteristics to obtain the suspension control parameters includes:

[0029] If the risk factor is greater than the preset safety threshold, assigning a basic suspension stiffness parameter and a basic damping coefficient parameter based on the obstacle type and the terrain characteristics;

[0030] Adjusting the basic suspension stiffness parameter based on the vertical height difference in the obstacle characteristic parameter set to obtain a suspension stiffness parameter;

[0031] Adjusting the basic damping coefficient parameter based on the risk coefficient to obtain a damping coefficient parameter;

[0032] The suspension control parameter is obtained according to the adjusted suspension stiffness parameter and the damping coefficient parameter.

[0033] Optionally, the step of obtaining surface hardness data of the area to be operated, and adjusting the suspension control parameters according to the surface hardness data to generate a chassis response parameter set includes:

[0034] Classifying the area to be operated into a first type of terrain and a second type of terrain according to the surface hardness data;

[0035] If the terrain is the first type, adjusting the suspension stiffness parameter of the suspension control parameter according to a first preset proportional coefficient to generate a target suspension stiffness parameter;

[0036] If the terrain is the second type, adjusting the damping coefficient parameter of the suspension control parameter according to the second preset proportional coefficient to generate a target damping coefficient parameter;

[0037] The chassis response parameter set is generated according to the target suspension stiffness parameter and the damping coefficient parameter.

[0038] In addition, to achieve the above objectives, an embodiment of the present invention further provides a multi-modal collaborative tractor intelligent chassis control system, the system comprising:

[0039] A multi-source data acquisition and fusion module is used to obtain original data sets including visual image data, chassis pressure data and position coordinate data, and perform data fusion on the original data sets to construct an environmental state matrix;

[0040] The obstacle identification and risk assessment module is used to determine the obstacle type based on the environmental state matrix, obtain obstacle characteristic parameters, and calculate the risk coefficient;

[0041] a suspension control decision module, configured to adjust suspension control parameters according to the risk coefficient and the ground surface hardness data to generate a chassis response parameter set;

[0042] The control execution module is used to control the operation of the chassis suspension system according to the chassis response parameter set.

[0043] In addition, to achieve the above-mentioned purpose, an embodiment of the present invention also provides a terminal device, including a memory, a processor, and a multimodal collaborative tractor intelligent chassis control program stored in the memory and runnable on the processor. When the processor executes the multimodal collaborative tractor intelligent chassis control program, the method described above is implemented.

[0044] In addition, to achieve the above-mentioned purpose, an embodiment of the present invention further provides a computer-readable storage medium, on which a multimodal collaborative tractor intelligent chassis control program is stored. When the multimodal collaborative tractor intelligent chassis control program is executed by a processor, the method described above is implemented.

[0045] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0046] (1) The present invention adopts multi-sensor fusion technology to integrate multi-source raw data such as visual images, chassis pressure and position coordinates, effectively solving the technical problems of one-sided and inaccurate single sensor data and redundant conflicts between different sensor data, and realizes comprehensive, accurate and unified perception of the tractor's surrounding environment and its own status, thus building a reliable environmental model for subsequent intelligent control.

[0047] (2) The present invention uses a comprehensive analysis technology based on the environmental state matrix, combined with visual features, spatial position and other information to perform obstacle identification and feature extraction, effectively solving the technical problems of inaccurate obstacle identification and incomplete feature extraction in related technologies, and accurately determining the obstacle type and obtaining detailed feature parameters, providing a key basis for subsequent risk assessment and control.

[0048] (3) The present invention calculates the collision risk coefficient based on a quantitative calculation model of obstacle characteristic parameters and comprehensively considers multiple characteristic parameters, effectively solving the technical problems of difficulty in quantitatively evaluating collision risk and inaccurate evaluation results, and realizes intuitive and accurate quantification of collision risk, providing reliable quantitative indicators for control system decision-making.

[0049] (4) The intelligent control technology of the present invention dynamically adjusts the suspension parameters according to the obstacle type, terrain characteristics and collision risk, effectively solving the technical problem that the suspension system cannot actively adjust according to the real-time environment and has poor adaptability. It enables the tractor chassis to actively adapt to different driving environments, optimize driving performance and reduce damage.

[0050] (5) The present invention adopts a fine control technology that combines surface hardness data to perform secondary adjustment of suspension control parameters, which effectively solves the technical problem that the adjustment of suspension parameters based only on obstacles and terrain is not fine enough and cannot fully adapt to different surface conditions. It further optimizes the tractor chassis response and improves operation accuracy and efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a flow chart of the first embodiment of the multi-modal collaborative tractor intelligent chassis control method of the present application;

[0052] Figure 2 This is a flow chart of the second embodiment of the multi-modal collaborative tractor intelligent chassis control method of the present application;

[0053] Figure 3 This is a schematic diagram of the terminal structure of the hardware operating environment involved in an embodiment of the present application. DETAILED DESCRIPTION

[0054] To address the problem of tractor chassis' inability to adaptively adjust chassis parameters in complex terrain, this application uses multi-sensor fusion technology to integrate multi-source raw data to construct an environmental state matrix. A comprehensive analysis technique based on the environmental state matrix accurately determines obstacle type and extracts detailed characteristic parameters. A quantitative calculation model based on obstacle characteristic parameters intuitively and accurately quantifies collision risk. Intelligent control technology dynamically adjusts suspension parameters based on obstacle type, terrain characteristics, and collision risk, enabling the chassis to proactively adapt to the environment. Finally, a refined control technique that secondary adjusts suspension parameters based on surface hardness data optimizes chassis response. Ultimately, the tractor is able to automatically adjust chassis parameters based on real-time data, effectively adapting to complex terrain and thereby improving operational efficiency and safety.

[0055] To better understand the above technical solutions, exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein. Instead, these embodiments are provided to enable a more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.

[0056] In order to better understand the above technical solution, the above technical solution will be described in detail below with reference to the accompanying drawings and specific implementation methods.

[0057] Embodiment 1: In this embodiment, a multi-modal collaborative tractor intelligent chassis control method is provided.

[0058] Reference Figure 1 The multi-modal collaborative tractor intelligent chassis control method of this embodiment includes the following steps:

[0059] Step S100: acquiring an original data set including visual image data, chassis pressure data, and position coordinate data, and performing data fusion on the original data set to construct an environmental state matrix;

[0060] In this embodiment, multi-source sensors are used to acquire visual image data, chassis pressure data, and GPS position coordinate data of the tractor's surroundings. A multimodal data fusion algorithm aligns and correlates these data from different sources and formats, generating an environmental state matrix that comprehensively reflects the tractor's current state. Multi-sensor data fusion effectively addresses the issues of incomplete and inaccurate single-sensor data, as well as redundant and conflicting data between different sensors. It achieves comprehensive perception of the tractor's surroundings and its own state, providing a reliable data foundation for subsequent steps.

[0061] As an optional implementation, terrain edge features are determined based on the visual image data, and the terrain type is determined based on the terrain edge features. Terrain edge features refer to pixel-level features in the visual image data that can identify terrain boundaries or sudden changes in height, such as ditch edges, field ridge outlines, or the dividing line of undulating terrain.

[0062] Exemplarily, pixel-level features of visual image data are extracted through image processing algorithms. If the pixel value of a certain area is higher than a preset pixel threshold of 200, the area is marked as a terrain edge feature. After processing, multiple potential terrain edge feature areas are identified in the visual image data. These areas may correspond to ditches, ridges, or boundaries of undulating terrain in the fields. The identified terrain edge features are used as input and sent to a pre-trained convolutional neural network model for classification to determine the terrain type. By learning a large amount of labeled terrain image data, the convolutional neural network model can identify different terrain types based on terrain edge features, such as flat land, sloping land, ditch land, etc.

[0063] As another optional implementation, after determining the terrain type, corresponding terrain features are obtained, and the terrain features are associated with chassis pressure data and position coordinate data to generate a fused data set.

[0064] For example, if the terrain type is identified as "ditch terrain," chassis pressure data at that time is obtained. When the tractor enters the ditch terrain area, the chassis pressure data will change significantly, such as a sudden increase in pressure, indicating that the tractor is passing through a concave terrain. The identified ditch terrain type is then associated with the chassis pressure data to record the pressure state of the chassis under this terrain. The terrain type, chassis pressure data, and position coordinate data are combined into a fused dataset. This fused dataset not only contains information about the terrain's geometric features but also reflects the actual load state of the tractor under this terrain, providing more comprehensive and accurate data support for subsequent chassis control decisions.

[0065] As another optional implementation, the spatial offset of the tractor is calculated based on the fused dataset. If the spatial position offset is greater than a preset offset threshold, the weight of the terrain feature is adjusted and the fused dataset is updated.

[0066] For example, the continuous position points recorded by GPS Mapping to a unified coordinate system, the straight-line distance between two points is calculated using the Euclidean distance formula as the initial spatial offset. The heading angle is then inferred from the GPS velocity vector, and the offset direction is corrected to eliminate errors caused by steering. The calculated offset is then correlated with the terrain features and pressure data in the fused dataset for verification. If the terrain type changes suddenly and the pressure sensor data indicates a significant uneven load distribution, the current offset is considered valid. Otherwise, it is considered noise or error and filtered out. Furthermore, the offset threshold is dynamically adjusted based on terrain complexity. For example, a lower threshold, such as 0.3 meters, is set in flat terrain to filter out minor fluctuations, while a higher threshold, such as 0.5 meters, is set in complex terrain with ditches or slopes to accommodate greater positional variations. When the calculated spatial offset exceeds the threshold corresponding to the current terrain type, a terrain feature weighting adjustment mechanism is triggered. By increasing the weight of terrain features in the fused dataset (for example, from 0.4 to 0.7) and reducing the weight of pressure data (for example, from 0.6 to 0.3), the subsequent control strategy is more focused on adapting to the impact of terrain changes. Finally, the adjusted weights are re-fused with the terrain features, pressure data, and the corrected offset to generate an updated fused dataset.

[0067] As another optional implementation, after updating the fused data set, the parameters of image edge detection are dynamically adjusted based on the multidimensional data in the fused data set, and edge detection is performed on the visual image data; if an area in the image with a height change exceeding a preset change threshold is detected, the area is marked as a terrain undulation area to obtain a terrain undulation mark set; the terrain undulation mark set is time-synchronized and associated with the corresponding chassis pressure data and the position coordinate data to generate the environmental state matrix.

[0068] Exemplarily, multidimensional data is extracted from the updated fusion data set, including dynamically adjusted terrain feature weights, chassis pressure distribution, and real-time position coordinates recorded by GPS. At the same time, the visual image data and depth information captured by the camera are combined, and the parameters of the image edge detection algorithm are dynamically optimized according to the current terrain feature weights. For example, the high and low thresholds of the Canny edge detection are lowered to enhance the sensitivity to the edges of complex terrain. The height change of each edge area in the image is calculated through the depth information, and the area with a height difference exceeding the preset height difference threshold of 0.2 meters is marked as a terrain undulation area, generating a terrain undulation mark set containing area coordinates and height change values.

[0069] The terrain undulation marker set, chassis pressure data, and GPS position coordinates are synchronized and aligned using timestamps to ensure that all data corresponds to the tractor status at the same moment. Next, terrain types such as flat, ditch, and slope are used as environmental classification labels and spatially associated with regional geometric features in the terrain undulation marker set, such as location, area, height gradient, chassis pressure distribution patterns such as the front and rear wheel pressure difference, and GPS position information, to construct a multidimensional data association model.

[0070] Environmental classification labels, regional geometric features, pressure distribution patterns, and location information are encapsulated into an environmental state matrix according to a unified data structure. Each row represents the complete environmental state at a point in time, including a timestamp, terrain type, and a set of terrain undulation areas. Each area records its center coordinates, boundary range and height change value, chassis pressure vector, and GPS coordinates, thus forming a comprehensive data set that can reflect the geometric characteristics, load status, and spatial position of the tractor's surrounding environment in real time, providing the chassis control module with accurate environmental perception input to support dynamic decision-making.

[0071] Optionally, after the environmental state matrix is ​​generated, the environmental state matrix can be checked for consistency. If there is data missing or abnormal, the data with the corresponding timestamp is extracted from the pre-established backup data to supplement it and determine the integrity of the final environmental state matrix.

[0072] Step S200: determining the obstacle type according to the environmental state matrix, and obtaining a set of obstacle characteristic parameters of the obstacle type;

[0073] In this embodiment, the visual feature vectors, spatial position data, and height gradient information in the environmental state matrix are used to classify and identify obstacles using algorithms such as convolutional neural networks. Feature parameters such as obstacle type, vertical height, and horizontal distance are extracted to form a set of obstacle feature parameters. This comprehensive analysis of the environmental state matrix effectively addresses the issues of inaccurate obstacle identification and incomplete feature extraction, enabling accurate determination of obstacle type and acquisition of detailed feature parameters.

[0074] As an optional implementation, a pre-trained convolutional neural network model is used to classify the visual features in the environment state matrix, identify obstacle types such as stones, tree stumps, and puddles, and extract the vertical height difference and horizontal distance between the obstacle and the tractor from the environment state matrix to generate a set of obstacle feature parameters.

[0075] For example, the model identifies the obstacle ahead as a stone. The laser rangefinder is used to obtain the horizontal distance between the tractor's current position and the obstacle. Assume that the distance to a stone is measured to be 5.2 meters. At the same time, the infrared height sensor is used to measure the vertical height difference to be 0.3 meters. The horizontal distance and vertical height difference are checked for consistency. Assuming that the preset horizontal distance threshold range is between 0.5-50 meters and the height difference threshold range is between -1.5 meters and 1.5 meters, the horizontal distance and vertical height difference of the stone are both within the corresponding preset threshold ranges, so no correction is required. The timestamp alignment tool is used to synchronize the data collected by different sensors to ensure the consistency of the data in the time dimension. For example, if the image recognition time is 10:00:05 and the distance measurement time is 10:00:06, the time is aligned to 10:00:05. After synchronizing the obstacle type, vertical height difference, and horizontal distance, a set of obstacle feature parameters is generated.

[0076] Optionally, the obstacle feature parameter set is stored in a pre-established database in chronological order, and the parameter set is grouped according to the obstacle type. For example, all stone-related data is grouped into one category, tree stumps into another category, and puddles into another category, forming a parameter set divided by type to facilitate targeted analysis of the impact of different obstacles. In addition, the parameter set can be further subdivided according to the regional characteristics of field operations. Assuming that in a certain operation area, stones are mostly distributed in the fields, and puddles mostly appear in low-lying areas, the data can be grouped again based on the location information to form a more detailed parameter set. This detailed classification will help to formulate differentiated operation strategies for different terrain areas in the future.

[0077] Alternatively, when measuring the horizontal distance and vertical height difference of obstacles, multiple sensors can be used to collaborate to improve accuracy. For example, if an infrared sensor and an ultrasonic sensor simultaneously measure the vertical height difference of a tree stump as 0.5 meter and 0.48 meter, the final result can be averaged or prioritized based on sensor confidence. This approach effectively reduces the error associated with a single sensor.

[0078] Step S300: Calculating the collision risk between the tractor and the obstacle based on the obstacle characteristic parameter set, and calculating the risk coefficient;

[0079] In this embodiment, a collision risk assessment algorithm is used to calculate the probability of collision between the tractor and the obstacle based on a set of obstacle characteristic parameters and the tractor's real-time speed, generating a quantified risk factor. This quantitative calculation of collision risk effectively addresses the difficulties associated with quantitative assessment and inaccurate assessment results. It intuitively and accurately quantifies collision risk, providing a reliable quantitative indicator for control system decision-making.

[0080] As an optional implementation, the preset distance safety threshold corresponding to the vertical height difference is the chassis clearance, and the preset distance safety threshold corresponding to the horizontal distance is the preset horizontal distance threshold. The vertical height difference and horizontal distance distribution are compared with the corresponding preset safety thresholds to determine whether there is a collision risk. If there is a collision risk, a weighted algorithm is used to calculate the risk coefficient in combination with the real-time driving speed of the tractor.

[0081] For example, if the vertical height difference of the obstacle is 0.4, the horizontal distance is 4.5 meters, the chassis clearance is 0.3 meters, and the preset horizontal distance threshold is 6 meters, it is obvious that the horizontal distance is less than the preset horizontal distance, and the vertical height difference is greater than the preset height difference threshold, and there is a collision risk. The current driving speed data is obtained. Assuming that the measured speed is 2.8 meters per second, the distance parameters of 4.5 meters, the vertical difference of 0.4 meters and the speed of 2.8 meters per second are associated and bound to form a comprehensive risk parameter combination. Based on the comprehensive risk parameter combination, each parameter is weighted. Since the horizontal distance directly affects the reaction time, the parameter weight of the horizontal distance is higher, while the weights of the speed and vertical difference are slightly lower. According to the preset weight distribution rules, these parameters are comprehensively evaluated to obtain a quantitative risk coefficient of 7.2. The higher the value, the greater the risk.

[0082] Alternatively, risk factors can be stored in a pre-established risk database. Assuming the risk database records the risk factor for each operation in chronological order, for example, in a field operation, risk values ​​of 7.2, 5.1, and 3.8 are recorded for multiple obstacles, forming a categorized risk assessment set. This storage method facilitates subsequent tracing and analysis of risk distribution over different time periods or regions. Furthermore, the risk database can also subdivide risk factors by operation area. For example, if the risk factor is generally higher at the edge of a field, the data from this area can be separately classified to facilitate subsequent targeted optimization of the operation path.

[0083] Optionally, when calculating the risk factor using a weighted algorithm, the weighting can be adjusted based on the specific field operation environment. For example, when working on slopes, the weighting of vertical differences may be increased appropriately, as height changes have a greater impact on collision risk. This dynamic adjustment allows the risk factor to better reflect the actual operation scenario, enhancing the targeted nature of the risk assessment.

[0084] Step S400: If the risk factor is greater than a preset safety threshold, the suspension stiffness and damping coefficient are calculated according to the obstacle type and terrain characteristics to obtain suspension control parameters;

[0085] In this embodiment, when the risk factor exceeds a preset safety threshold, a parameter optimization formula is used to calculate the optimal suspension stiffness and damping coefficient based on obstacle type and terrain characteristics, such as terrain undulation and soil hardness. These parameters are then generated to adjust the chassis suspension system in real time to adapt to different operating environments. By dynamically adjusting suspension parameters based on obstacle type and terrain characteristics, this effectively addresses the suspension system's inability to proactively adjust to real-time conditions and its poor adaptability. This allows the tractor chassis to proactively adapt to different driving environments, optimizing driving performance.

[0086] As an optional implementation, when the risk factor is greater than a preset safety threshold, the basic suspension stiffness parameters and the basic damping coefficient parameters are assigned based on the obstacle type and terrain characteristics.

[0087] For example, a tractor operating on a slope detects a 0.4-meter-diameter rock 3 meters ahead. The calculated collision risk coefficient is 8.5, exceeding the safety threshold of 6.0, triggering the suspension parameter adjustment mechanism. The terrain slope is 15 degrees, indicating an uphill slope. Basic parameter coefficients matching the "rock" obstacle type and the "15-degree uphill" terrain characteristics are extracted from a pre-set parameter map. Because rocks are high-hardness obstacles, the stiffness coefficient is set at 1.2 times the standard value. To quickly attenuate impact vibrations, the damping coefficient is set at 1.5 times the standard value. A secondary correction is applied to the parameters based on the 15-degree uphill terrain. The slope correction factor, obtained from the parameter map, is 1.15, necessitating increased front suspension stiffness to prevent pitching when driving uphill. These coefficients are then combined with the standard parameter values. The standard stiffness value of 8,000 N / m is corrected to 9,600 N / m (8,000 × 1.2). After terrain correction, the resulting basic suspension stiffness parameter is 11,040 N / m (9,600 × 1.15). The standard damping value of 1200 N·s / m becomes 1800 N·s / m (1200 × 1.5) after obstacle correction, and the base damping coefficient parameter after terrain correction is 2070 N·s / m (1800 × 1.15). This value assignment takes into account both the transient high-energy characteristics of rock impact and the increased front wheel load during uphill driving, providing a reasonable starting point for subsequent parameter optimization that complies with physical constraints.

[0088] As another optional implementation, a preset parameter optimization formula is used to adjust the basic damping coefficient and basic suspension stiffness parameters to obtain the suspension control parameters. The parameter optimization formula for suspension stiffness is: , where K represents the suspension stiffness, represents the basic suspension stiffness parameter, represents the adjustment coefficient, Represents the vertical height difference; the parameter optimization formula of the damping coefficient is , where C represents the damping coefficient, represents the basic damping coefficient, represents the adjustment coefficient, and R represents the risk coefficient.

[0089] For example, when a rock obstacle is detected ahead and the terrain slope is 15 degrees, the calculated risk factor is 8.5, which exceeds the preset safety threshold of 6.0. At this time, the initial suspension parameters need to be optimized and adjusted. Assuming the initial basic suspension stiffness is =11040N / m, basic damping coefficient =2070N·s / m, and the adjustment coefficient is set to α=0.08 N / m based on experience. 2 , β = 0.15 N·s / m, and the vertical height difference between the tractor and the stone is measured by the lidar =0.3m. Substituting the parameters into the above formula, the suspension stiffness is calculated to be 10775N / m. When the vertical height difference is greater, the obstacle protrusion is more obvious, and the stiffness needs to be appropriately reduced to reduce the bumps of the vehicle body. The optimized suspension stiffness is 97.6% of the basic suspension stiffness parameter, which meets the requirements. The damping coefficient is 4709N·s / m. When the risk factor is higher, the collision impact energy is greater, and the damping coefficient needs to be significantly increased to quickly attenuate the vibration and ensure continuous contact between the tire and the ground. The optimized damping coefficient is 227.5% of the basic damping coefficient, which meets the requirements. The optimized parameters are input into the vehicle dynamics model to simulate the vehicle body vibration acceleration when passing through stones from the initial 3.2m / s 2 Down to 1.8m / s 2 The tire dynamic load coefficient dropped from 0.45 to 0.28, and the risk factor simultaneously dropped to 4.9, meeting safety threshold requirements. When the tire dynamic load coefficient was 0.45, it was close to the danger threshold of 0.5, which could easily cause the tire to lift off the ground or excessive deformation. After optimization, it was 0.28, indicating more stable tire-ground contact. The dynamic load coefficient of agricultural machinery tires is typically required to be less than 0.35 on hard roads and less than 0.4 on soft roads. The optimized value of 0.28 meets safety requirements. This optimization process achieves flexible buffering and smooth passage of high-risk obstacles by dynamically balancing stiffness and damping.

[0090] It should be noted that the parameter optimization formula of the suspension stiffness and the parameter optimization formula of the damping coefficient are suitable for relatively flat road environments. If in extreme scenarios, it is necessary to introduce nonlinear adjustments to the parameter optimization formula of the suspension stiffness, such as α The damping limit can be set in extreme scenarios, such as , or adopt a segmented adjustment method, such as when R < 10, β = 0.15; when R ≥ 10, β = 0.1, to balance the attenuation efficiency and response speed.

[0091] As another optional implementation, an online learning mechanism can be introduced to dynamically adjust the adjustment coefficients α and β in the parameter optimization formula according to the actual operation results to improve the accuracy and adaptability of parameter calculation.

[0092] For example, the feedback information of the electro-hydraulic control system can be combined to realize intelligent fine-tuning of the suspension parameters to ensure the smoothness and comfort of the chassis response. For example, in the initial stage, the empirical coefficients α=0.08 and β=0.15 are used. When dealing with stones with a height difference of 0.3m, the suspension stiffness is calculated to be 10775N / m and the damping coefficient is 4709N·s / m. However, the actual operation shows that the vehicle body vibrates continuously for 0.8 seconds after passing through and the tire contact time accounts for only 82%, indicating that the damping coefficient is too high, resulting in excessive energy attenuation. The data acquisition module is started to record key features such as the suspension cylinder pressure fluctuation range of 12-18MPa and the vehicle body vertical acceleration spectrum of 10-15Hz energy accounting for 45%. The real-time deviation data of the pressure sensor and the displacement sensor are obtained through the electro-hydraulic control unit. Based on these feedbacks, an incremental loss function is constructed. , α is optimized to 0.074 through gradient descent iteration, and the frequency band attenuation term is introduced into the damping model , adjust β to 0.12-0.03f. The optimized parameters make the vehicle acceleration from 2.1m / s 2 Down to 1.4m / s 2 The tire contact time is increased to 87%. The coefficient is automatically fine-tuned every 15 minutes according to the displacement deviation. After 8 hours of operation, the coefficient converges to 、 , parameter calculation error was reduced from 23% to 8%, significantly improving operational stability. This mechanism achieves intelligent adaptive optimization of suspension parameters by integrating multi-dimensional feedback such as pressure, vibration, and displacement in real time.

[0093] Step S500: Acquire surface hardness data of the area to be operated, and adjust the suspension control parameters according to the surface hardness data to generate a chassis response parameter set.

[0094] In this embodiment, multi-point inspections are performed in the area to be operated to obtain detailed surface hardness data. Based on a preset surface hardness classification standard, the area to be operated is classified as either soft terrain or hard obstacle terrain. If the terrain is soft, an instruction to reduce the suspension stiffness is generated; if the terrain is hard, an instruction to increase the damping coefficient is generated, resulting in a targeted adjustment instruction set. Based on this targeted adjustment instruction set, the electro-hydraulic control module adjusts the stiffness and damping characteristics of the chassis suspension in real time, reducing the suspension stiffness parameter for soft terrain and increasing the damping coefficient parameter for hard obstacles. This results in an adjusted suspension parameter combination, which is then used to generate the chassis response parameter set.

[0095] As an optional embodiment, the area to be operated is divided into a first type of terrain and a second type of terrain based on the surface hardness data. If the terrain is the first type, the suspension stiffness parameter of the suspension control parameter is adjusted according to a first preset proportional coefficient to generate a target suspension stiffness parameter. If the terrain is the second type, the damping coefficient parameter of the suspension control parameter is adjusted according to a second preset proportional coefficient to generate a target damping coefficient parameter. The chassis response parameter set is generated based on the target suspension stiffness parameter and the damping coefficient parameter. The first type of terrain and the second type of terrain represent soft soil and hard obstacles, respectively. In actual applications, other types of terrain may also be included based on the actual operation.

[0096] For example, the terrain scanning tool collects surface hardness data in real time, and the dual-threshold classification method is used to divide the area to be operated into two types of typical terrain. The terrain scanning tool simultaneously collects three sets of characteristic parameters: ultrasonic echo attenuation rate, vibration wave conduction velocity, and soil resistivity. When it is detected that the ultrasonic attenuation rate continuously exceeds 0.8μs / cm and the vibration wave velocity is lower than 200m / s, the system determines that the current area is the first type of terrain, soft soil terrain. At this time, the terrain classification signal T=1 is automatically generated and the flexible adaptation mode is triggered. In this mode, the electro-hydraulic control system calls the parameter adjustment formula , the suspension stiffness parameter =8000N / m according to the first preset proportional coefficient =0.20 to generate a target suspension stiffness parameter of 6400N / m. At the same time, the electro-hydraulic valve opening pressure threshold is reduced by 15% to enhance suspension compliance. When the lidar detects a local hardness mutation exceeding 50N / cm² / m and the duration exceeds 0.5m, the system determines it as a second-class terrain, a hard obstacle, generates a terrain classification signal T=2 and activates the impact buffer mode. At this time, the damping adjustment formula is called , the damping coefficient =1200N·s / m according to the second preset proportional coefficient =0.15 to generate a target damping coefficient of 1380N·s / m, and simultaneously enable a dual-stage buffer structure to improve energy absorption efficiency.

[0097] Optionally, the chassis response parameter set can be adjusted according to the multi-terrain hybrid mode. When the terrain scanning unit detects the transition area between soft soil and rock, the system uses a weighted fusion algorithm Calculate the comprehensive adjustment coefficient, where the weight , Dynamic allocation based on the confidence level of terrain classification. For example, in the area where soft soil and rock meet, the stiffness is automatically adjusted to 7200N / m, between 6400-8000N / m, and the damping coefficient is increased to 1300N·s / m, achieving a smooth parameter transition. Through hardware-level parameter adjustment and control strategy innovation, this technology improves the suspension system's adaptability to typical terrain by more than 40%, shortening the parameter adjustment response time to less than 200ms, effectively solving the problem of stable operation of agricultural machinery in complex terrain. It deeply couples the surface hardness characteristic parameters with the suspension system dynamics model, and maximizes operating efficiency by solving the optimal parameter combination in real time.

[0098] As another optional implementation, during the adjustment process, chassis attitude change data and load distribution data are acquired simultaneously via an attitude sensor. These data are then fused to generate a dynamic response characteristic dataset. This dynamic response characteristic dataset is then compared with a preset response threshold. If the threshold is exceeded, a secondary fine-tuning instruction is generated, resulting in an optimized chassis response parameter set.

[0099] For example, the attitude sensor can collect real-time chassis tilt angle data. For example, when a tractor travels at 3 km / h on a 10° slope, the attitude sensor detects the chassis roll angle increasing from an initial 0° to 10°, while also recording the dynamic change in front-to-back height difference from 0.1m to 0.3m. The load monitoring unit simultaneously captures the increase in the left axle load from the standard value of 12kN to 15.2kN, while the right axle load decreases from 12kN to 8.8kN, a load offset of 30.3%. This multi-dimensional data is fused and processed, and after using a Kalman filter algorithm to eliminate sensor noise, a dynamic response dataset is generated, containing key features such as "roll angle exceeding the limit (10° > 8° threshold)" and "load offset exceeding the limit (30.3% > 25% threshold)."

[0100] Based on this data set, the system initiates a two-tiered comparison and analysis mechanism: First, at the base response layer, it verifies whether individual parameters exceed their limits. When a roll angle of 10° exceeds the preset safety threshold of 8°, a primary warning is immediately triggered. Then, at the integrated response layer, a fuzzy logic algorithm is used to assess the risk of multi-parameter coupling, calculating a combined risk factor of R=0.82. Given a preset safety threshold of 0.7, it confirms the need for secondary fine-tuning. The control unit then generates a fine-tuning command containing temporal and spatial coordinate information, implementing a composite adjustment strategy for the left suspension: The front suspension stiffness is dynamically reduced from its initial value of 8000 N / m to 6800 N / m, corresponding to an adjustment factor of 0.85. Simultaneously, the rear suspension damping coefficient is increased from 1200 N·s / m to 1440 N·s / m, corresponding to an adjustment factor of 1.2, creating an asymmetric parameter adjustment mode.

[0101] Optionally, during the adjustment execution phase, the electro-hydraulic control system adopts a feedforward-feedback composite control architecture, completing parameter adjustments within 200ms. The results after adjustment show: the chassis roll angle is reduced from 10° to 7.2°, the load offset is reduced from 30.3% to 18.7%, and the tire contact time difference is shortened from 120ms to 45ms. The system continuously monitors the dynamic response characteristics after adjustment. When the roll angle change rate is found to be less than 0.5° / s for 3 seconds, the parameter retention mode is automatically activated. If new out-of-limit characteristics are detected during subsequent operations (such as a sudden increase in vertical acceleration to 3.5m / s² due to a pothole), the secondary fine-tuning mechanism is reactivated, forming a complete adaptive control closed loop.

[0102] In this embodiment, dynamic adaptive control of the tractor chassis is achieved through multi-source data fusion and intelligent decision-making. Visual images, chassis pressure, and position coordinate data are simultaneously collected to construct an environmental state matrix to identify obstacle types and characteristic parameters and calculate collision risk factors. When the risk exceeds a limit, the suspension stiffness and damping coefficients are dynamically adjusted based on obstacle type and terrain characteristics. Parameters are also optimized secondary using surface hardness data, forming a "perception-decision-execution" closed-loop control system. This improves the tractor's active obstacle avoidance response speed in complex operating environments, reducing collision risks. Furthermore, adaptive suspension parameter adjustment improves chassis stability, significantly enhancing operational safety and maneuverability in unstructured terrain.

[0103] Example 2: Based on Example 1, another example of this application is proposed, referring to Figure 2 After obtaining the surface hardness data of the area to be operated and adjusting the suspension control parameters according to the surface hardness data to generate a chassis response parameter set, the following steps are included:

[0104] Step S600: continuously collecting chassis status data, load change data, and environmental change data of the tractor's operating range to form a real-time monitoring data stream;

[0105] In this embodiment, a multi-source sensor array achieves omnidirectional perception of the tractor's operating status. Chassis status data refers to the 3D attitude angle, vibration acceleration, and displacement collected by an inertial measurement unit. Load change data refers to the vertical load distribution on all four axles, monitored in real time by an axle pressure sensor array. Environmental change data consists of a 3D point cloud of terrain and obstacle outlines acquired through the fusion of a LiDAR and stereo vision system.

[0106] As an optional implementation, a distributed sensor network enables simultaneous multi-source data collection: an inertial measurement unit continuously acquires chassis 3D attitude angle and vibration acceleration data, a set of axle pressure sensors monitors load distribution in real time, and a lidar and stereo vision system collaborate to generate a 3D point cloud of the operating environment. The data fusion processing unit utilizes ROS middleware for cross-sensor time synchronization and uses an extended Kalman filter algorithm to spatially and temporally align the inertial measurement unit data with the axle pressure data, constructing a real-time monitoring data stream containing 12-dimensional feature parameters.

[0107] For example, when a tractor traverses a muddy paddy field, the inertial measurement unit detects an increase in roll angle from 0° to 4.2°, with a peak vibration acceleration of 3.8 m / s². The pressure sensor indicates that the right axle load has decreased from 12.5 kN to 9.1 kN, while the left axle load has increased to 13.7 kN. The lidar identifies a 0.25 m deep trench 3 m ahead. The system fuses this data to generate a real-time monitoring stream containing features such as "roll angle exceeded," "load offset 21.6%," and "trench terrain warning," achieving a data integrity rate of 99.3%.

[0108] Step S700: Classifying the environmental change data, and if the terrain is determined to be a sudden change according to the classification result, generating a suspension parameter adjustment instruction;

[0109] In this embodiment, the environmental change data is classified to determine the terrain, which includes flat ground, sloped land, and sudden terrain. When it is determined to be sudden terrain, a suspension parameter adjustment instruction is generated.

[0110] As an optional implementation, collaborative sensing using LiDAR and visual sensors enables terrain classification and mutation detection. This involves introducing a dynamic terrain database based on historical operation data. This database categorizes terrain into three categories: flat ground (slope <3° and roughness <0.3), sloping terrain (3° ≤ slope <15° and roughness 0.3-0.6), and sudden terrain (slope ≥15° or roughness ≥0.6 or sudden elevation difference >0.3m). It also stores 200 sets of terrain characteristics and parameter adjustment data from historical operation scenarios. During real-time operation, the LiDAR scans a 10m area ahead at a 10Hz frequency to generate a terrain height profile, while the visual sensor simultaneously captures grayscale images and calculates texture edge density. Terrain feature values ​​are calculated every two seconds: geometric features are derived from the current slope and elevation mutation rate calculated using a laser point cloud. Texture features are derived by calculating the ground roughness index using the standard deviation of the visual image's grayscale gradient. Euclidean distance matching is performed between the current feature values ​​and historical records of similar terrain in the database. The 10 data sets with the smallest distances are selected as references to determine the historical matching degree. When any geometric or texture feature detected exceeds a preset threshold (e.g., slope change rate > 5° / s or Rt mutation > 40%), and the historical matching score shows a similarity of more than 75% between the current feature and historical data related to sudden terrain, the system identifies the situation as sudden terrain and extracts the optimal parameter combination (stiffness / damping adjustment coefficient, ground clearance, etc.) from the historical record with the highest matching score to generate adjustment instructions. If no matching record is found, the default sudden terrain adjustment strategy is used.

[0111] For example, a tractor approached a field ridge at 3 km / h. A lidar detected a sudden drop in ground height from 0.2m to -0.25m in 0.4 seconds, creating an 18° downhill slope, 5m ahead. The calculated slope change rate was 22.5° / s (18° / 0.4s), and the height mutation rate was 0.11m / s, both exceeding preset thresholds. The visual sensor captured the ground texture changing from a soft mud layer to a hard earth ridge, with the standard deviation of the grayscale gradient increasing from 18 to 42, a change of 133%. The current feature values ​​(slope 18°, Rt = 0.72, height mutation of 0.45m) were matched against 87 historical records of sudden terrain changes in the database. The similarity with a previously recorded "hard field ridge downhill" scenario in the same plot was 89%, with a Euclidean distance of 0.12. The historical record indicated the optimal parameters were a stiffness adjustment coefficient of 0.6, a damping adjustment coefficient of 1.5, and a ground clearance of +80mm. Suspension parameter adjustment instructions were generated based on these parameters.

[0112] Step S800: adjusting the suspension stiffness parameter and the damping coefficient according to the suspension parameter adjustment instruction to generate an adjusted suspension parameter combination;

[0113] In this embodiment, upon receiving a suspension parameter adjustment command, the electro-hydraulic control module first retrieves the base adjustment coefficients from the terrain-parameter mapping table: For ridge terrain, the stiffness adjustment coefficient is 0.78, corresponding to a baseline value of 8000 N / m → 6240 N / m; the damping adjustment coefficient is 1.28, corresponding to a baseline value of 1200 N·s / m → 1536 N·s / m. An adaptive fuzzy PID controller is then activated for parameter fine-tuning. The controller inputs are the vibration acceleration error, a target value of 0.5 m / s², a measured value of 1.2 m / s², and its rate of change of -0.8 m / s³; the output is the parameter adjustment increment. The system employs a two-tiered adjustment strategy: the primary adjustment stage completes coarse parameter adjustments within 180 ms, while the secondary adjustment stage uses feedforward compensation to eliminate the nonlinear hysteresis characteristics of the hydraulic system.

[0114] For example, while climbing a ridge, after primary adjustment reduced stiffness to 6240 N / m and increased damping to 1536 N·s / m, the measured vibration acceleration still reached 0.9 m / s². Based on the error signal, the fuzzy PID controller further fine-tuned the damping coefficient to 1612 N·s / m, increasing the adjustment by +76 N·s / m, ultimately reducing the vibration acceleration to 0.48 m / s² and increasing the tire contact time from 68% to 84%.

[0115] Step S900: fusing the chassis state data and the load change data to obtain dynamic state data;

[0116] As an optional implementation, after adjusting the tractor chassis according to the suspension parameter adjustment instructions, chassis status data and load change data are acquired and fused to generate dynamic status data. The chassis status data includes vibration acceleration and suspension travel utilization, and the load change data includes axle load, load balance index, and center of gravity offset.

[0117] For example, when crossing a ridge, uneven axle load is detected, which is manifested as the left front wheel load is 120kN, the right rear wheel load is 110kN, the load balance index is 0.18, and the vibration acceleration is 0.8m / s 2 After adjusting the suspension parameters to 8400 N / m stiffness and 1560 N·s / m damping, the sensor provided real-time feedback: the load on the left front wheel increased to 135 kN, the load on the right rear wheel decreased to 95 kN, and the load balance index was optimized to 0.15. Each indicator was normalized, assuming that the reasonable range for the load balance index was [0, 0.3] and the reasonable range for the vibration acceleration was [0, 2.5] m / s. 2 , the reasonable range of center of gravity offset is [0,0.2]m, the reasonable range of suspension travel utilization is [0,1], and the range normalization formula is used The calculated optimized load balance index normalized value is 0.5, the vibration acceleration normalized value is 0.48, the center of gravity offset normalized value is 0.4, and the suspension travel utilization normalized value is 0.65. Weighted fusion of these data, using a load balance index weight of 0.3, a vibration acceleration weight of 0.4, a center of gravity offset weight of 0.2, and a travel utilization weight of 0.1, yields the dynamic state data S = 0.3 × 0.5 + 0.4 × 0.48 + 0.2 × 0.4 + 0.1 × 0.65 = 0.487. Because S is less than the dynamic state threshold of 0.5, but close to it, it is considered "critically stable." The current parameters are maintained, but load monitoring is initiated every 50 milliseconds. At the same time, the rear-wheel drive torque is reduced by 5% to reduce the risk of front wheel overload, ensuring smooth traversal of the ridge.

[0118] Step S1000: If the dynamic state data is within a preset state threshold range, the suspension parameter combination is maintained.

[0119] As an optional implementation, when the dynamic state data S satisfies 0.7 ≤ S ≤ 1.0 for five consecutive sampling periods, and the absolute value of the chassis attitude angle change rate is less than 0.5° / s, the suspension parameter combination is deemed valid and a parameter hold command is generated. This command triggers the locking circuit in the electro-hydraulic control module, shutting off the parameter adjustment signal path via a solid-state relay and simultaneously activating a parameter hold timer. During the hold period, the fluctuation range of the evaluation value is continuously monitored. If the maximum fluctuation exceeds 0.05, the adjustment process is reactivated. If no adjustment is triggered after the timer expires, the system automatically enters a low-power monitoring mode, keeping only the inertial measurement unit and pressure sensor operational, and reducing the sampling rate to 10Hz. The chassis attitude angle change rate is a physical quantity that describes the speed of the tractor chassis's attitude changes in three-dimensional space. It typically refers to the time-dependent rate of change of the pitch, roll, and heading angles. The chassis attitude angle change rate is a dynamic metric for determining the validity of a suspension parameter combination. It directly quantifies the chassis's response speed and stability to sudden terrain changes, compensating for the hysteresis of static indicators and ensuring that the adjusted parameters can both suppress excessive tilt and adapt to complex operating scenarios.

[0120] For example, when operating on a flat field, the S value remained stable in the 0.82-0.85 range for 10 consecutive seconds, with a roll angle change rate of 0.3° / s, triggering a parameter hold command. After locking, the electro-hydraulic valve group maintained its current opening: stiffness 6240 N / m, damping 1612 N·s / m, and power consumption dropped from 125 W to 78 W. During the subsequent 30 minutes of operation, only one parameter adjustment was triggered due to an occasional rock encounter; stable control was maintained throughout the remainder of the operation.

[0121] In this embodiment, after generating a chassis response parameter set, chassis status data, load change data, and environmental change data within the tractor's operating range are continuously collected to form a real-time monitoring data stream. After determining that the terrain is abrupt, based on the environmental change data, suspension parameter adjustment instructions are generated. Suspension stiffness parameters and damping coefficients are adjusted based on these suspension parameter adjustment instructions to generate a suspension parameter combination. The chassis status data and load change data from the real-time monitoring data stream are then obtained and fused to generate dynamic status data. When the dynamic status data falls within a preset threshold, the tractor is determined to be stable and the suspension parameter combination is maintained. This multi-parameter fusion evaluation effectively shortens parameter adjustment response time, improves chassis pitch angle change rate control accuracy, and optimizes the load balance index, reducing the risk of tractor overturning when operating in complex terrain.

[0122] Example 3: In the embodiment of the present application, a multi-modal collaborative tractor intelligent chassis control device is proposed.

[0123] Reference Figure 3 , Figure 3 This is a schematic diagram of the terminal structure of the hardware operating environment involved in an embodiment of the present application.

[0124] like Figure 3 As shown, the control terminal may include: a processor 1001, such as a CPU, a network interface 1003, a memory 1004, and a communication bus 1002. The communication bus 1002 is used to enable communication between these components. The network interface 1003 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1004 may be a high-speed RAM memory or a non-volatile memory, such as a disk drive. The memory 1004 may also be a storage device independent of the processor 1001.

[0125] Those skilled in the art will understand that Figure 3 The terminal structure shown in the figure does not constitute a limitation to the terminal, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0126] like Figure 3 As shown, the memory 1004 as a computer storage medium may include an operating system, a network communication module, and a multi-modal collaborative tractor intelligent chassis control program.

[0127] exist Figure 3 In the hardware structure of the multi-modal collaborative tractor intelligent chassis control device shown, the processor 1001 can call the multi-modal collaborative tractor intelligent chassis control program stored in the memory 1004 and perform the following operations:

[0128] Acquire an original data set including visual image data, chassis pressure data, and position coordinate data, and perform data fusion on the original data set to construct an environmental state matrix;

[0129] Determining an obstacle type according to the environmental state matrix, and obtaining an obstacle characteristic parameter set of the obstacle type;

[0130] Calculating the collision risk between the tractor and the obstacle based on the obstacle characteristic parameter set, and calculating the risk coefficient;

[0131] If the risk factor is greater than a preset safety threshold, the suspension stiffness and damping coefficient are calculated according to the obstacle type and terrain characteristics to obtain suspension control parameters;

[0132] Surface hardness data of the area to be operated is obtained, and the suspension control parameters are adjusted according to the surface hardness data to generate a chassis response parameter set.

[0133] Optionally, the processor 1001 may call the multimodal collaborative tractor intelligent chassis control program stored in the memory 1004 and further perform the following operations:

[0134] determining terrain edge features according to the visual image data, and determining a terrain type based on the terrain edge features;

[0135] Acquiring terrain features of the terrain type, and associating the terrain features with the chassis pressure data and the position coordinate data to generate a fused data set;

[0136] Calculating a spatial position offset of the tractor based on the fused dataset, and if the spatial position offset is greater than a preset offset threshold, adjusting a weight of the terrain feature and updating the fused dataset;

[0137] The environmental state matrix is ​​generated according to the fused data set.

[0138] Optionally, the processor 1001 may call the multimodal collaborative tractor intelligent chassis control program stored in the memory 1004 and further perform the following operations:

[0139] Dynamically adjusting image edge detection parameters based on the multidimensional data in the fused data set, and performing edge detection on the visual image data;

[0140] If it is detected that there is an area in the image where the height change exceeds a preset change threshold, the area is marked as a terrain relief area to obtain a terrain relief mark set;

[0141] The terrain relief mark set is time-synchronized and associated with the corresponding chassis pressure data and the position coordinate data to generate the environmental state matrix.

[0142] Optionally, the processor 1001 may call the multimodal collaborative tractor intelligent chassis control program stored in the memory 1004 and further perform the following operations:

[0143] Determine an obstacle and its type based on the visual feature vector, spatial position data, and height gradient information in the environmental state matrix;

[0144] The vertical height difference and horizontal distance between the obstacle and the current position of the tractor are obtained, and the obstacle feature parameter set including the obstacle type, the vertical height difference and the horizontal distance is generated.

[0145] Optionally, the processor 1001 may call the multimodal collaborative tractor intelligent chassis control program stored in the memory 1004 and further perform the following operations:

[0146] Comparing the vertical height difference and horizontal distance in the obstacle characteristic parameter set with the corresponding preset distance safety threshold to determine whether there is a collision risk;

[0147] If there is a collision risk, the real-time driving speed of the tractor is obtained, and the real-time driving speed is associated with the vertical height difference and the horizontal distance to generate a comprehensive risk parameter combination;

[0148] The parameters in the comprehensive risk parameter combination are adjusted according to preset weight rules to generate the risk coefficient.

[0149] Optionally, the processor 1001 may call the multimodal collaborative tractor intelligent chassis control program stored in the memory 1004 and further perform the following operations:

[0150] If the risk factor is greater than the preset safety threshold, assigning a basic suspension stiffness parameter and a basic damping coefficient parameter based on the obstacle type and the terrain characteristics;

[0151] Adjusting the basic suspension stiffness parameter based on the vertical height difference in the obstacle characteristic parameter set to obtain a suspension stiffness parameter;

[0152] Adjusting the basic damping coefficient parameter based on the risk coefficient to obtain a damping coefficient parameter;

[0153] The suspension control parameter is obtained according to the adjusted suspension stiffness parameter and the damping coefficient parameter.

[0154] Optionally, the processor 1001 may call the multimodal collaborative tractor intelligent chassis control program stored in the memory 1004 and further perform the following operations:

[0155] Classifying the area to be operated into a first type of terrain and a second type of terrain according to the surface hardness data;

[0156] If the terrain is the first type, adjusting the suspension stiffness parameter of the suspension control parameter according to a first preset proportional coefficient to generate a target suspension stiffness parameter;

[0157] If the terrain is the second type, adjusting the damping coefficient parameter of the suspension control parameter according to the second preset proportional coefficient to generate a target damping coefficient parameter;

[0158] The chassis response parameter set is generated according to the target suspension stiffness parameter and the damping coefficient parameter.

[0159] In addition, to achieve the above objectives, an embodiment of the present invention further provides a multi-modal collaborative tractor intelligent chassis control system, the system comprising:

[0160] A multi-source data acquisition and fusion module is used to obtain original data sets including visual image data, chassis pressure data and position coordinate data, and perform data fusion on the original data sets to construct an environmental state matrix;

[0161] The obstacle identification and risk assessment module is used to determine the obstacle type based on the environmental state matrix, obtain obstacle characteristic parameters, and calculate the risk coefficient;

[0162] a suspension control decision module, configured to adjust suspension control parameters according to the risk coefficient and the ground surface hardness data to generate a chassis response parameter set;

[0163] The control execution module is used to control the operation of the chassis suspension system according to the chassis response parameter set.

[0164] In addition, to achieve the above-mentioned purpose, an embodiment of the present invention also provides a terminal device, including a memory, a processor, and a multimodal collaborative tractor intelligent chassis control program stored in the memory and runnable on the processor. When the processor executes the multimodal collaborative tractor intelligent chassis control program, the multimodal collaborative tractor intelligent chassis control method as described above is implemented.

[0165] In addition, to achieve the above-mentioned purpose, an embodiment of the present invention further provides a computer-readable storage medium, on which a multimodal collaborative tractor intelligent chassis control program is stored. When the multimodal collaborative tractor intelligent chassis control program is executed by a processor, the multimodal collaborative tractor intelligent chassis control method as described above is implemented.

[0166] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0167] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0168] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0169] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0170] It should be noted that in the claims, any reference signs placed between parentheses shall not be construed as limiting the claims. The word "comprising" does not exclude the presence of components or steps not listed in the claim. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The present application may be implemented by means of hardware comprising several distinct components and by means of a suitably programmed computer. In a unit claim enumerating several means, several of these means may be embodied by one and the same item of hardware. The use of the words first, second and third etc. does not indicate any order. These words may be interpreted as names.

[0171] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present application.

[0172] Obviously, those skilled in the art may make various changes and modifications to the present application without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present application fall within the scope of the claims and their equivalents, the present application is intended to include such modifications and variations.

Claims

1. A multi-modal collaborative tractor intelligent chassis control method, characterized in that: The method comprises: Acquire an original data set including visual image data, chassis pressure data, and position coordinate data, and perform data fusion on the original data set to construct an environmental state matrix; Determining an obstacle type according to the environmental state matrix, and obtaining an obstacle characteristic parameter set of the obstacle type; Calculating the collision risk between the tractor and the obstacle based on the obstacle characteristic parameter set, and calculating the risk coefficient; If the risk factor is greater than a preset safety threshold, the suspension stiffness and damping coefficient are calculated according to the obstacle type and terrain characteristics to obtain suspension control parameters; Acquiring surface hardness data of the area to be operated, and adjusting the suspension control parameters according to the surface hardness data to generate a chassis response parameter set; The steps of acquiring an original data set including visual image data, chassis pressure data, and position coordinate data, and performing data fusion on the original data set to generate an environmental state matrix include: determining terrain edge features according to the visual image data, and determining a terrain type based on the terrain edge features; Acquiring terrain features of the terrain type, and associating the terrain features with the chassis pressure data and the position coordinate data to generate a fused data set; Calculating a spatial position offset of the tractor based on the fused dataset, and if the spatial position offset is greater than a preset offset threshold, adjusting a weight of the terrain feature and updating the fused dataset; generating the environmental state matrix according to the fused data set; Wherein, if the risk coefficient is greater than a preset safety threshold, the step of calculating the suspension stiffness and damping coefficient according to the obstacle type and terrain characteristics to obtain the suspension control parameters includes: If the risk factor is greater than the preset safety threshold, assigning a basic suspension stiffness parameter and a basic damping coefficient parameter based on the obstacle type and the terrain characteristics; Adjusting the basic suspension stiffness parameter based on the vertical height difference in the obstacle characteristic parameter set to obtain a suspension stiffness parameter; Adjusting the basic damping coefficient parameter based on the risk coefficient to obtain a damping coefficient parameter; The suspension control parameter is obtained according to the adjusted suspension stiffness parameter and the damping coefficient parameter.

2. The multi-modal coordinated tractor intelligent chassis control method according to claim 1, characterized in that: The step of generating the environmental state matrix according to the fused data set includes: Dynamically adjusting image edge detection parameters based on the multidimensional data in the fused data set, and performing edge detection on the visual image data; If it is detected that there is an area in the image where the height change exceeds a preset change threshold, the area is marked as a terrain relief area to obtain a terrain relief mark set; The terrain relief mark set is time-synchronized and associated with the corresponding chassis pressure data and the position coordinate data to generate the environmental state matrix.

3. The multi-modal coordinated tractor intelligent chassis control method according to claim 1, characterized in that: The step of determining the obstacle type according to the environmental state matrix and obtaining the obstacle characteristic parameter set of the obstacle type includes: Determine an obstacle and its type based on the visual feature vector, spatial position data, and height gradient information in the environmental state matrix; The vertical height difference and horizontal distance between the obstacle and the current position of the tractor are obtained, and the obstacle feature parameter set including the obstacle type, the vertical height difference and the horizontal distance is generated.

4. The multi-modal coordinated tractor intelligent chassis control method according to claim 1, characterized in that: The step of calculating the collision risk between the tractor and the obstacle based on the obstacle characteristic parameter set and calculating the risk coefficient includes: Comparing the vertical height difference and horizontal distance in the obstacle characteristic parameter set with the corresponding preset distance safety threshold to determine whether there is a collision risk; If there is a collision risk, the real-time driving speed of the tractor is obtained, and the real-time driving speed is associated with the vertical height difference and the horizontal distance to generate a comprehensive risk parameter combination; The parameters in the comprehensive risk parameter combination are adjusted according to preset weight rules to generate the risk coefficient.

5. The multi-modal coordinated tractor intelligent chassis control method according to claim 1, characterized in that: The step of obtaining the surface hardness data of the area to be operated, and adjusting the suspension control parameters according to the surface hardness data to generate a chassis response parameter set includes: Classifying the area to be operated into a first type of terrain and a second type of terrain according to the surface hardness data; If the terrain is the first type, adjusting the suspension stiffness parameter of the suspension control parameter according to a first preset proportional coefficient to generate a target suspension stiffness parameter; If the terrain is the second type, adjusting the damping coefficient parameter of the suspension control parameter according to the second preset proportional coefficient to generate a target damping coefficient parameter; The chassis response parameter set is generated according to the target suspension stiffness parameter and the damping coefficient parameter.

6. A multi-modal collaborative tractor intelligent chassis control system, characterized in that: A method for implementing a multi-modal collaborative tractor intelligent chassis control method according to any one of claims 1 to 5, the system comprising: A multi-source data acquisition and fusion module is used to obtain original data sets including visual image data, chassis pressure data and position coordinate data, and perform data fusion on the original data sets to construct an environmental state matrix; The obstacle identification and risk assessment module is used to determine the obstacle type based on the environmental state matrix, obtain obstacle characteristic parameters, and calculate the risk coefficient; a suspension control decision module, configured to adjust suspension control parameters according to the risk coefficient and the ground surface hardness data to generate a chassis response parameter set; The control execution module is used to control the operation of the chassis suspension system according to the chassis response parameter set.

7. A terminal device, characterized in that: The invention comprises a memory, a processor and a multimodal collaborative tractor intelligent chassis control program stored in the memory and executable on the processor, wherein when the processor executes the multimodal collaborative tractor intelligent chassis control program, the method described in any one of claims 1 to 5 is implemented.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a multi-modal collaborative tractor intelligent chassis control program, and when the multi-modal collaborative tractor intelligent chassis control program is executed by a processor, the method described in any one of claims 1 to 5 is implemented.

Images