Wheel-track composite chassis mode switching control method and system

Abstract

The invention relates to the technical field of artificial intelligence, and discloses a wheel-track combined type chassis mode switching control method and system, and the method comprises the steps: obtaining an environment three-dimensional point cloud sequence and an environment image of a terrain environment where a target wheel-track combined type machine is located, carrying out the moving point cloud removal of the environment three-dimensional point cloud sequence, and obtaining a target three-dimensional point cloud; performing point cloud multi-scale sampling on the target three-dimensional point cloud to obtain multi-scale sampling data; according to the multi-scale sampling data and the environment image, identifying a terrain edge and constructing a three-dimensional terrain model; generating an initial chassis control strategy of the target wheel-track combined machine according to the three-dimensional terrain model, and calculating a terrain risk coefficient; and optimizing the initial control strategy based on the terrain risk coefficient to obtain an optimal control strategy. According to the invention, the precision of the mode switching control of the wheel-track composite chassis can be improved.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method and system for switching modes of a wheel-track composite chassis. Background Technology

[0002] Numerous ground mobility platforms, including robots and special vehicles, are designed to operate in complex road environments. Among propulsion mechanisms, wheels offer high efficiency and economy, while tracks provide high mobility, low ground pressure, and good adaptability to unstructured surfaces, making them widely used on various ground mobility platforms. As platforms place increasingly higher demands on mobility performance, single wheeled or tracked propulsion mechanisms struggle to meet the diverse needs of complex roads. Therefore, wheel-tracked hybrid propulsion mechanisms have emerged and rapidly become a research hotspot.

[0003] The wheel-track hybrid chassis is a new type of mobile platform that combines the high-speed mobility of wheels with the strong terrain adaptability of tracks. It can intelligently control and flexibly switch between the two modes, balancing high-speed mobility and adaptability to complex terrain. However, existing technologies rely solely on speed or position sensors and do not fully integrate terrain environmental characteristics (such as slope and surface hardness). Furthermore, in complex terrain environments, sensor signals are easily affected by vibration and interference, leading to a disconnect between the control strategy and actual working conditions, and consequently, poor accuracy in switching modes of the wheel-track hybrid chassis. Summary of the Invention

[0004] This invention provides a mode switching control method and system for a wheel-track composite chassis, the main purpose of which is to solve the problem of low accuracy in mode switching of wheel-track composite chassis.

[0005] To achieve the above objectives, the present invention provides a mode switching control method for a wheel-track composite chassis, comprising: Acquire the environmental 3D point cloud sequence and environmental image of the terrain environment where the target wheel-tracked hybrid machine is located, and remove the moving point cloud from the environmental 3D point cloud sequence to obtain the target 3D point cloud; Multi-scale sampling of the target's three-dimensional point cloud is performed to obtain multi-scale sampling data of the terrain environment. Based on the multi-scale sampling data and the environmental image, the terrain edges of the terrain environment are identified, and a three-dimensional terrain model of the terrain environment is constructed based on the terrain edges. An initial chassis control strategy for the target wheel-track hybrid machine is generated based on the three-dimensional terrain model, and the terrain risk coefficient of the chassis control strategy is calculated. The initial control strategy is optimized based on the terrain risk coefficient to obtain the optimal control strategy for the target wheel-tracked hybrid machine in the terrain environment.

[0006] Optionally, the step of removing moving point clouds from the environmental 3D point cloud sequence to obtain the target 3D point cloud includes: Generate a point cloud depth map sequence of the terrain environment based on the three-dimensional point cloud sequence of the environment, and calculate the point cloud residual map between the point cloud depth map sequences; Perform upper and lower level self-correcting convolutions on the point cloud residual map and the point cloud depth map sequence to obtain environmental context convolutional features; The context convolutional features are encoded and decoded to obtain encoded and decoded features, and the dynamic point cloud of the environmental 3D point cloud sequence is calculated based on the encoded and decoded features. The dynamic point cloud is removed from the environmental 3D point cloud sequence to obtain the target 3D point cloud.

[0007] Optionally, the step of performing upper and lower level self-correcting convolutions on the point cloud residual map and the point cloud depth map sequence to obtain environmental context convolutional features includes: The point cloud residual map and the point cloud depth map of the current frame in the point cloud depth map sequence are superimposed to obtain the target feature map; The target feature map is subjected to multi-layer convolutional activation and convolution processing respectively to obtain the first feature map and the second feature map; The first feature map and the second feature map are concatenated to obtain the concatenated feature; The concatenated features are fused with the target feature map to obtain the environmental context convolutional features.

[0008] Optionally, the step of performing multi-scale sampling of the target 3D point cloud to obtain multi-scale sampling data of the terrain environment includes: The target 3D point cloud is divided into meshes to obtain point cloud mesh data; The point cloud grid data is subjected to maximum resampling, minimum resampling and average resampling respectively to obtain the maximum resampling point cloud data, minimum resampling point cloud data and average resampling point cloud data; The maximum resampled point cloud data, the minimum resampled point cloud data, and the average resampled point cloud data are combined to obtain multi-scale sampling data.

[0009] Optionally, identifying the terrain edges of the terrain environment based on the multi-scale sampling data and the environmental image includes: The environmental image is processed to obtain a grayscale image, and the basic image features of the grayscale image are extracted. Extract the point cloud geometric feature vectors from the multi-scale sampling data, and construct the image basic features and the fusion features of the point cloud geometric feature vectors; The fused features are subjected to multi-layer dilated convolution to obtain the target terrain environment features; Based on the terrain features, edge prediction is performed on the environment image to obtain the terrain edges of the environment.

[0010] Optionally, constructing a three-dimensional terrain model of the terrain environment based on the terrain edges includes: The surface of the target 3D point cloud is reconstructed to obtain a 3D surface; The terrain is adapted to the three-dimensional surface based on the terrain edge to obtain a three-dimensional terrain model of the terrain environment.

[0011] Optionally, generating the initial chassis control strategy for the target wheel-track hybrid machine based on the three-dimensional terrain model includes: The three-dimensional terrain model is divided into meshes to obtain a mesh node model; The terrain elevation difference of the target wheel-tracked hybrid machine is calculated based on the grid node model, and the ground type is determined. The optimal chassis mode of the target wheel-track hybrid machine is calculated based on the terrain elevation difference and the ground type. The initial chassis control strategy for the target wheel-track hybrid machine is generated based on the optimal chassis mode.

[0012] Optionally, calculating the terrain risk coefficient of the chassis control strategy includes: The grid node path of the target wheel-track hybrid machine is determined based on the chassis control strategy. Calculate the elevation risk coefficient and slope risk coefficient of the chassis control strategy based on the grid node path; A terrain risk coefficient is constructed based on the elevation risk coefficient and the slope risk coefficient.

[0013] Optionally, optimizing the initial control strategy based on the terrain risk coefficient to obtain the optimal control strategy for the target wheel-tracked hybrid machine in the given terrain environment includes: When the terrain risk coefficient is greater than the preset risk coefficient threshold, return to the above steps of generating the initial chassis control strategy of the target wheel-tracked hybrid machine based on the three-dimensional terrain model to obtain the updated chassis control strategy; Calculate the terrain risk coefficient of the updated chassis control strategy; When the terrain risk coefficient of the updated chassis control strategy is less than or equal to the risk coefficient threshold, the updated chassis control strategy is taken as the optimal control strategy.

[0014] To address the above problems, the present invention also provides a wheel-track hybrid chassis mode switching control system, the system comprising: The moving point cloud removal module is used to acquire the environmental three-dimensional point cloud sequence and environmental image of the terrain environment where the target wheel-tracked composite machine is located, and to remove the moving point cloud from the environmental three-dimensional point cloud sequence to obtain the target three-dimensional point cloud. A multi-scale sampling module is used to perform multi-scale sampling of the target three-dimensional point cloud to obtain multi-scale sampling data of the terrain environment. A three-dimensional terrain model construction module is used to identify the terrain edges of the terrain environment based on the multi-scale sampling data and the environmental image, and to construct a three-dimensional terrain model of the terrain environment based on the terrain edges. The terrain risk coefficient calculation module is used to generate an initial chassis control strategy for the target wheel-tracked hybrid machine based on the three-dimensional terrain model, and to calculate the terrain risk coefficient of the chassis control strategy. The optimal control strategy generation module is used to optimize the initial control strategy based on the terrain risk coefficient to obtain the optimal control strategy for the target wheel-tracked hybrid machine in the terrain environment.

[0015] This invention removes moving point clouds from the 3D point cloud sequence of the target wheel-tracked hybrid machine's terrain environment, thus avoiding interference from moving targets and improving the accuracy of subsequent optimal control strategy generation. Multi-scale sampling of the target's 3D point cloud yields multi-scale sampling data. Based on this data and environmental imagery, terrain edges are identified, and a 3D terrain model is constructed. This model visually reflects the terrain structure of the target wheel-tracked hybrid machine's environment, providing a foundation for subsequent control. An initial chassis control strategy is generated based on the 3D terrain model, and the terrain risk coefficient of the strategy is calculated, allowing for the assessment of risks in the terrain environment and the implementation of reasonable avoidance measures. The initial control strategy is optimized based on the terrain risk coefficient to obtain the optimal control strategy, which can be updated simultaneously for elevation changes at different locations, enabling precise switching of the target wheel-tracked hybrid machine's chassis mode. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating a mode switching control method for a wheel-track composite chassis provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of a process for removing moving point clouds from a three-dimensional point cloud sequence of an environment, provided in an embodiment of the present invention. Figure 3 This is a schematic diagram of a process for identifying terrain edges based on multi-scale sampling data and environmental images, according to an embodiment of the present invention. Figure 4This is a functional block diagram of a wheel-track composite chassis mode switching control system provided in an embodiment of the present invention; The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0017] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0018] This application provides a method for switching modes on a wheel-track hybrid chassis. The executing entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the wheel-track hybrid chassis mode switching control method can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0019] Reference Figure 1 The diagram shown is a flowchart illustrating a wheel-track composite chassis mode switching control method according to an embodiment of the present invention. In this embodiment, the wheel-track composite chassis mode switching control method includes: S1. Obtain the environmental three-dimensional point cloud sequence and environmental image of the terrain environment where the target wheel-tracked composite machine is located, and remove the moving point cloud from the environmental three-dimensional point cloud sequence to obtain the target three-dimensional point cloud.

[0020] In this embodiment of the invention, the target wheel-tracked composite machine is a special equipment that combines two walking devices, wheels and tracks, such as rescue robots, off-road machinery, polar exploration equipment, mine inspection robots, and other types of equipment.

[0021] In detail, the target wheel-track hybrid machine is located in a variety of complex terrain environments, including highways, urban roads, mountains, slopes, gravel roads, ruins, etc., which require the coordination of tracks or wheels to improve traction and obstacle crossing ability.

[0022] In this embodiment of the invention, two sets of three-dimensional point cloud data collected at different time frames within a preset time interval are used to obtain an environmental three-dimensional point cloud sequence. For example, the environmental three-dimensional point clouds collected at the current time frame and at time frames 30 seconds, one minute, or any set time interval are used to capture the full terrain (such as slope, obstacles, and gullies) of the terrain environment using vehicle-mounted LiDAR or lightweight LiDAR, so as to comprehensively and accurately capture terrain undulations. The environmental image is a two-dimensional image of the terrain environment, which can be obtained by setting a high-definition camera on the target wheel-tracked hybrid machine, or by obtaining environmental images through satellite remote sensing or UAV aerial photography.

[0023] Furthermore, moving point cloud removal removes the environmental 3D point cloud represented by the moving target from the environmental 3D point cloud sequence, thereby obtaining the target 3D point cloud representing the static target, in order to remove the interference of the moving target on the target wheel-track composite machine control.

[0024] Specifically, see Figure 2 As shown, the process of removing moving point clouds from the environmental 3D point cloud sequence to obtain the target 3D point cloud includes: S11. Generate a point cloud depth map sequence of the terrain environment based on the three-dimensional point cloud sequence of the environment, and calculate the point cloud residual map between the point cloud depth map sequences; S12. Perform upper and lower level self-correcting convolution on the point cloud residual map and the point cloud depth map sequence to obtain the environmental context convolution features. S13. Perform feature encoding and decoding on the context convolutional features to obtain encoded and decoded features, and calculate the dynamic point cloud of the environmental 3D point cloud sequence based on the encoded and decoded features; S14. Remove the dynamic point cloud from the environmental 3D point cloud sequence to obtain the target 3D point cloud.

[0025] In this embodiment of the invention, the depth map is the points of the three-dimensional point cloud in each three-dimensional spatial coordinate system in the two-dimensional image coordinate system, thereby distinguishing the geometric contours of objects through the depth of the depth map and improving the efficiency of moving point cloud removal.

[0026] Specifically, the point cloud depth map of the terrain environment can be generated using the following formula: ; in, Indicates the depth map of the point cloud. The coordinates of the location This represents the environmental 3D point cloud in the environmental 3D point cloud sequence. This indicates the preset upper limit of the radar field of view. This indicates the preset radar field of view size. Indicates the height of the depth map. This indicates the width of the depth map.

[0027] In detail, by converting each environmental 3D point cloud in the environmental 3D point cloud sequence into the coordinates of the corresponding point cloud depth map based on the radar field of view parameters acquired by the lidar, point cloud depth maps at different time frames can be generated, thus obtaining a depth map sequence.

[0028] Furthermore, the absolute value of the difference between the pixel values ​​at the same coordinate point in the depth map sequence is calculated. The ratio between the absolute value and the pixel value corresponding to the current time frame is the residual value corresponding to the coordinate point. The residual value of each coordinate point is calculated in turn, and the residual value is used as the pixel value to generate a point cloud residual map.

[0029] In this embodiment of the invention, the step of performing upper and lower level self-correcting convolutions on the point cloud residual map and the point cloud depth map sequence to obtain environmental context convolutional features includes: The point cloud residual map and the point cloud depth map of the current frame in the point cloud depth map sequence are superimposed to obtain the target feature map; The target feature map is subjected to multi-layer convolutional activation and convolution processing respectively to obtain the first feature map and the second feature map; The first feature map and the second feature map are concatenated to obtain the concatenated feature; The concatenated features are fused with the target feature map to obtain the environmental context convolutional features.

[0030] In this embodiment of the invention, multi-layer convolutional activation is applied to the target features. Figure 1 The process involves 1×1 convolution, 3×3 convolution, activation function, and 3×3 convolution to obtain the first feature map. The convolution process is applied to the target feature map. Figure 1 ×1 convolution and 3×3 convolution are used to obtain the second feature map.

[0031] Furthermore, feature fusion involves performing 1×1 convolution on the concatenated features and adding them to the target feature map to obtain the environmental context convolutional features with the same dimensions as the target feature map.

[0032] In this embodiment of the invention, residual images can be fused through contextual convolution features, while contextual features are used to improve the comprehensiveness of feature extraction, thereby improving the accuracy of subsequent target 3D point cloud generation.

[0033] Furthermore, feature extraction is performed using pre-built encoders and decoders. The encoder performs downsampling on the convolutional features of the environmental context and increases the number of channels to improve feature dimensionality and depth extraction. The decoder upsamples the feature map output by the encoder to gradually restore the image resolution and finally obtain the encoded and decoded features.

[0034] Preferably, before the encoder performs downsampling, key information of the high-resolution feature map is retained through skip connections; after entering the decoder stage, this information is fused with the upsampled features to enrich the detailed expression of the features.

[0035] The encoder can be composed of stacked convolutional layers and pooling layers (Max Pooling / AvgPooling), while the decoder can be a reverse convolution operation.

[0036] In this embodiment of the invention, the encoding and decoding features are mapped to the feature space of point cloud classification through the activation function in the fully connected layer. The environmental 3D point cloud with the classification result of dynamic point cloud is removed from the environmental 3D point cloud of the current frame, and replaced with the mean of the neighborhood point cloud coordinates within a preset range to obtain the target 3D point cloud.

[0037] In this embodiment of the invention, the point cloud data corresponding to the moving target can be removed by using the target's three-dimensional point cloud, thereby avoiding interference from the moving target and improving the accuracy of subsequent optimal control strategy generation.

[0038] S2. Perform multi-scale sampling of the target three-dimensional point cloud to obtain multi-scale sampling data of the terrain environment.

[0039] In this embodiment of the invention, multi-scale sampling of point clouds generates point cloud data at different resolutions to analyze the impact of resolution on terrain detection and provide a basis for subsequent identification of terrain edges.

[0040] In this embodiment of the invention, the step of performing multi-scale sampling of the target 3D point cloud to obtain multi-scale sampling data of the terrain environment includes: The target 3D point cloud is divided into meshes to obtain point cloud mesh data; The point cloud grid data is subjected to maximum resampling, minimum resampling and average resampling respectively to obtain the maximum resampling point cloud data, minimum resampling point cloud data and average resampling point cloud data; The maximum resampled point cloud data, the minimum resampled point cloud data, and the average resampled point cloud data are combined to obtain multi-scale sampling data.

[0041] In this embodiment of the invention, the mesh generation is based on a preset mesh side length and the coordinates of the target 3D point cloud to generate a uniform mesh. Resampling is then performed based on the z-axis coordinates of the point cloud within the mesh. For example, maximum resampling uses the maximum z-axis coordinate in each mesh and the x and y-axis coordinates at the mesh center as resampling points to obtain maximum resampled point cloud data; minimum resampling uses the minimum z-axis coordinate in each mesh and the x and y-axis coordinates at the mesh center as resampling points to obtain minimum resampled point cloud data; and average resampling uses the average z-axis coordinate in each mesh and the average x and y-axis coordinates as resampling points to obtain average resampled point cloud data.

[0042] S3. Identify the terrain edges of the terrain environment based on the multi-scale sampling data and the environmental image, and construct a three-dimensional terrain model of the terrain environment based on the terrain edges.

[0043] In this embodiment of the invention, the terrain edge is the transition boundary between different terrain features in the terrain environment. It is a spatial area where terrain attributes (such as elevation, slope, and landform) rapidly change from one state to another. By observing the terrain edge, the terrain can be predicted in advance, triggering the chassis control mode switch in advance and avoiding sudden power interruption or loss of attitude control.

[0044] Specifically, see Figure 3 As shown, the step of identifying the terrain edges of the terrain environment based on the multi-scale sampling data and the environmental image includes: S31. Perform grayscale processing on the environmental image to obtain a grayscale image, and extract the basic image features of the grayscale image; S32. Extract the point cloud geometric feature vectors from the multi-scale sampling data, and construct the image basic features and the fusion features of the point cloud geometric feature vectors; S33. Perform multi-layer dilated convolution on the fused features to obtain the target terrain environment features; S34. Based on the terrain and environmental features, perform edge prediction on the environmental image to obtain the terrain edges of the terrain environment.

[0045] In this embodiment of the invention, grayscale processing involves converting the environmental image into a single-channel grayscale image, calculating the grayscale value of each pixel based on the channel values ​​of the R, G, and B channels, and then constructing a grayscale image. A pre-built convolutional network is then used to extract texture features (edges, texture roughness), spectral features, etc., from the grayscale image.

[0046] Furthermore, geometric features of the point cloud from the multi-scale sampling data are extracted, such as slope, curvature, normal vector, and elevation. Among them, dilated convolution inserts holes (zero-filling) into the ordinary convolution kernel, and controls the size of the receptive field by controlling the dilation rate, thereby expanding the receptive field of the convolution and better preserving spatial features.

[0047] In this embodiment of the invention, point cloud geometric feature vectors corresponding to sampling data at different scales of the fused features are superimposed with the basic image features to obtain the fused features. Multi-layer dilated convolution involves connecting dilated convolutional layers with different dilation rates in parallel to stitch feature maps from different receptive fields along the channel dimension, thereby extracting the terrain and environmental features of the fused features.

[0048] In detail, each pixel in the environmental image is mapped to the feature space of a preset terrain edge category based on the terrain environmental features, so as to obtain the terrain edge corresponding to each pixel.

[0049] In this embodiment of the invention, constructing a three-dimensional terrain model of the terrain environment based on the terrain edge includes: The surface of the target 3D point cloud is reconstructed to obtain a 3D surface; The terrain is adapted to the three-dimensional surface based on the terrain edge to obtain a three-dimensional terrain model of the terrain environment.

[0050] Specifically, surface reconstruction is performed by methods such as moving least squares and Poisson reconstruction, and then the real surface texture of the three-dimensional surface (such as vegetation cover, different landform distribution, etc.) is restored by the terrain edge to generate a three-dimensional terrain environment.

[0051] In this embodiment of the invention, the terrain structure of the target wheel-tracked hybrid machine can be intuitively reflected through the three-dimensional terrain environment, providing a basis for subsequent control of the target wheel-tracked hybrid machine.

[0052] S4. Generate the initial chassis control strategy of the target wheel-tracked hybrid machine based on the three-dimensional terrain model, and calculate the terrain risk coefficient of the chassis control strategy.

[0053] In this embodiment of the invention, the initial chassis control strategy is the logic for switching chassis modes of the target wheel-track hybrid machine. For example, when the ground adhesion is less than a preset threshold or the terrain slope is greater than a preset threshold, the tracks need to be switched.

[0054] Specifically, the step of generating the initial chassis control strategy for the target wheel-tracked hybrid machine based on the three-dimensional terrain model includes: The three-dimensional terrain model is divided into meshes to obtain a mesh node model; The terrain elevation difference of the target wheel-tracked hybrid machine is calculated based on the grid node model, and the ground type is determined. The optimal chassis mode of the target wheel-track hybrid machine is calculated based on the terrain elevation difference and the ground type. The initial chassis control strategy for the target wheel-track hybrid machine is generated based on the optimal chassis mode.

[0055] In this embodiment of the invention, the grid division is achieved by dividing the three-dimensional terrain model into equal parts using a preset grid size, resulting in a grid node model where each grid serves as a node.

[0056] Furthermore, a target grid node is selected in the grid node model as the target point of the target wheel-tracked composite machine, and the terrain elevation difference between the current position and the grid node where the target point is located, as well as the ground category corresponding to the three-dimensional terrain model between the current position and the target point, are calculated.

[0057] In detail, the optimal chassis mode is the optimal mode of the chassis when the target wheel-track hybrid machine is driving. For example, when the ground type is hard ground and the vehicle speed is ≥5km / h, the optimal chassis mode is wheeled. When the terrain slope is ≥15°, the optimal chassis mode is wheeled. When the terrain elevation difference is greater than half the wheel diameter, the optimal chassis mode is a hybrid switching mode.

[0058] In this embodiment of the invention, chassis mode rules can be pre-written, and then the optimal chassis mode can be determined according to the chassis mode rules. The optimal chassis mode is dynamically matched with the terrain environment to balance mobility and safety.

[0059] Furthermore, the terrain risk coefficient represents the risk of the target wheel-tracked hybrid machine during operation. For example, the higher the elevation of the area, the higher the risk of rollover or skidding; and this risk tends to increase sharply within a small range of elevation values.

[0060] Specifically, calculating the terrain risk coefficient of the chassis control strategy includes: The grid node path of the target wheel-track hybrid machine is determined based on the chassis control strategy. Calculate the elevation risk coefficient and slope risk coefficient of the chassis control strategy based on the grid node path; A terrain risk coefficient is constructed based on the elevation risk coefficient and the slope risk coefficient.

[0061] In this embodiment of the invention, the grid nodes of the target wheel-track hybrid machine traveling on the three-dimensional terrain model in the chassis control strategy are extracted to obtain the grid node path, and the mean node coordinates of the point cloud coordinates in each grid node are obtained.

[0062] Furthermore, the elevation risk coefficient is calculated using the following formula: ; in, Indicates the elevation risk coefficient. This indicates the preset scaling factor. This indicates the preset risk growth rate. This represents the elevation difference between consecutive nodes in a grid node path. This indicates the preset range of elevation differences.

[0063] The slope risk coefficient is calculated using the following formula: ; in, Indicates the slope risk coefficient. This represents the preset first impact factor. Indicates the first node in the mesh node path The coordinates of each node, This represents the total number of nodes in the grid node path. This represents the pre-defined second influence factor.

[0064] In this embodiment of the invention, the slope risk coefficient and the elevation risk coefficient are weighted and added together to obtain the terrain risk coefficient. The terrain risk coefficient can be used to determine the risks in the terrain environment and make reasonable avoidances. At the same time, it can be updated according to the elevation changes at different locations to achieve precise switching of the target wheel-tracked composite machine chassis mode.

[0065] S5. Optimize the initial control strategy based on the terrain risk coefficient to obtain the optimal control strategy for the target wheel-tracked hybrid machine in the terrain environment.

[0066] In this embodiment of the invention, the initial control strategy is optimized by adjusting and optimizing the switching of the chassis model in the initial control strategy according to the terrain risk coefficient, so that the switching of chassis mode conforms to the terrain environment, while avoiding risks in the terrain environment.

[0067] Specifically, optimizing the initial control strategy based on the terrain risk coefficient to obtain the optimal control strategy for the target wheel-tracked hybrid machine in its terrain environment includes: When the terrain risk coefficient is greater than the preset risk coefficient threshold, return to step S4 of the above process of generating the initial chassis control strategy of the target wheel-tracked hybrid machine based on the three-dimensional terrain model to obtain the updated chassis control strategy. Calculate the terrain risk coefficient of the updated chassis control strategy; When the terrain risk coefficient of the updated chassis control strategy is less than or equal to the risk coefficient threshold, the optimal control strategy is obtained.

[0068] In this embodiment of the invention, when the terrain risk coefficient is greater than a preset risk coefficient threshold, it indicates that the current chassis mode switching has a high risk, which may cause the target wheel-track hybrid machine to overturn in severe cases. Therefore, it is necessary to return to the steps of generating the initial chassis control strategy of the target wheel-track hybrid machine to obtain an updated chassis control strategy until the terrain risk coefficient of the updated chassis control strategy is less than or equal to the risk coefficient threshold, thereby improving the accuracy of wheel-track hybrid chassis mode switching.

[0069] This invention removes moving point clouds from the 3D point cloud sequence of the target wheel-tracked hybrid machine's terrain environment, thus avoiding interference from moving targets and improving the accuracy of subsequent optimal control strategy generation. Multi-scale sampling of the target's 3D point cloud yields multi-scale sampling data. Based on this data and environmental imagery, terrain edges are identified, and a 3D terrain model is constructed. This model visually reflects the terrain structure of the target wheel-tracked hybrid machine's environment, providing a foundation for subsequent control. An initial chassis control strategy is generated based on the 3D terrain model, and the terrain risk coefficient of the strategy is calculated, allowing for the assessment of risks in the terrain environment and the implementation of reasonable avoidance measures. The initial control strategy is optimized based on the terrain risk coefficient to obtain the optimal control strategy, which can be updated simultaneously for elevation changes at different locations, enabling precise switching of the target wheel-tracked hybrid machine's chassis mode.

[0070] like Figure 4 The diagram shown is a functional block diagram of a wheel-track composite chassis mode switching control system provided in an embodiment of the present invention.

[0071] The wheel-track hybrid chassis mode switching control system 100 of this invention can be installed in an electronic device. Depending on the functions implemented, the wheel-track hybrid chassis mode switching control system 100 may include a moving point cloud removal module 101, a multi-scale sampling module 102, a three-dimensional terrain model construction module 103, a terrain risk coefficient calculation module 104, and an optimal control strategy generation module 105. The module described in this invention can also be called a unit, referring to a series of computer program segments that can be executed by the processor of an electronic device and perform a fixed function, stored in the memory of the electronic device.

[0072] In this embodiment, the functions of each module / unit are as follows: The moving point cloud removal module 101 is used to acquire the environmental three-dimensional point cloud sequence and environmental image of the terrain environment where the target wheel-tracked composite machine is located, and to remove the moving point cloud from the environmental three-dimensional point cloud sequence to obtain the target three-dimensional point cloud. The multi-scale sampling module 102 is used to perform multi-scale sampling of the target three-dimensional point cloud to obtain multi-scale sampling data of the terrain environment. The three-dimensional terrain model construction module 103 is used to identify the terrain edges of the terrain environment based on the multi-scale sampling data and the environmental image, and to construct a three-dimensional terrain model of the terrain environment based on the terrain edges. The terrain risk coefficient calculation module 104 is used to generate an initial chassis control strategy for the target wheel-tracked hybrid machine based on the three-dimensional terrain model, and to calculate the terrain risk coefficient of the chassis control strategy. The optimal control strategy generation module 105 is used to optimize the initial control strategy based on the terrain risk coefficient to obtain the optimal control strategy of the target wheel-tracked hybrid machine in the terrain environment.

[0073] In detail, the modules described in the wheel-track composite chassis mode switching control system 100 in this embodiment of the invention employ the same methods as described above during use. Figure 1 The method used is the same as the wheel-track composite chassis mode switching control method described above, and can produce the same technical effect, so it will not be repeated here.

[0074] In the several embodiments provided by this invention, it should be understood that the disclosed devices, systems, and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0075] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0076] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0077] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0078] Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be embraced within the invention. No appended diagram markings in the claims should be construed as limiting the scope of the claims.

[0079] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0080] Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or systems described in the system embodiments may also be implemented by a single unit or system through software or hardware. Terms such as "first," "second," etc., are used to indicate names and do not indicate any specific order.

[0081] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A mode switching control method for a wheel-track composite chassis, characterized in that, The method includes: Acquire the environmental 3D point cloud sequence and environmental image of the terrain environment where the target wheel-tracked hybrid machine is located, and remove the moving point cloud from the environmental 3D point cloud sequence to obtain the target 3D point cloud; Multi-scale sampling of the target's three-dimensional point cloud is performed to obtain multi-scale sampling data of the terrain environment. Based on the multi-scale sampling data and the environmental image, the terrain edges of the terrain environment are identified, and a three-dimensional terrain model of the terrain environment is constructed based on the terrain edges. An initial chassis control strategy for the target wheel-track hybrid machine is generated based on the three-dimensional terrain model, and the terrain risk coefficient of the chassis control strategy is calculated. The initial control strategy is optimized based on the terrain risk coefficient to obtain the optimal control strategy for the target wheel-tracked hybrid machine in the terrain environment.

2. The wheel-track composite chassis mode switching control method as described in claim 1, characterized in that, The process of removing moving point clouds from the environmental 3D point cloud sequence to obtain the target 3D point cloud includes: Generate a point cloud depth map sequence of the terrain environment based on the three-dimensional point cloud sequence of the environment, and calculate the point cloud residual map between the point cloud depth map sequences; Perform upper and lower level self-correcting convolutions on the point cloud residual map and the point cloud depth map sequence to obtain environmental context convolutional features; The context convolutional features are encoded and decoded to obtain encoded and decoded features, and the dynamic point cloud of the environmental 3D point cloud sequence is calculated based on the encoded and decoded features. The dynamic point cloud is removed from the environmental 3D point cloud sequence to obtain the target 3D point cloud.

3. The wheel-track composite chassis mode switching control method as described in claim 2, characterized in that, The step of performing upper and lower level self-correcting convolutions on the point cloud residual map and the point cloud depth map sequence to obtain environmental context convolutional features includes: The point cloud residual map and the point cloud depth map of the current frame in the point cloud depth map sequence are superimposed to obtain the target feature map; The target feature map is subjected to multi-layer convolutional activation and convolution processing respectively to obtain the first feature map and the second feature map; The first feature map and the second feature map are concatenated to obtain the concatenated feature; The concatenated features are fused with the target feature map to obtain the environmental context convolutional features.

4. The wheel-track composite chassis mode switching control method as described in claim 1, characterized in that, The step of performing multi-scale sampling of the target 3D point cloud to obtain multi-scale sampling data of the terrain environment includes: The target 3D point cloud is divided into meshes to obtain point cloud mesh data; The point cloud grid data is subjected to maximum resampling, minimum resampling and average resampling respectively to obtain the maximum resampling point cloud data, minimum resampling point cloud data and average resampling point cloud data; The maximum resampled point cloud data, the minimum resampled point cloud data, and the average resampled point cloud data are combined to obtain multi-scale sampling data.

5. The wheel-track composite chassis mode switching control method as described in claim 1, characterized in that, The step of identifying the terrain edges of the terrain environment based on the multi-scale sampling data and the environmental image includes: The environmental image is processed to obtain a grayscale image, and the basic image features of the grayscale image are extracted. Extract the point cloud geometric feature vectors from the multi-scale sampling data, and construct the image basic features and the fusion features of the point cloud geometric feature vectors; The fused features are subjected to multi-layer dilated convolution to obtain the target terrain and environmental features; Based on the terrain features, edge prediction is performed on the environmental image to obtain the terrain edges of the terrain environment.

6. The wheel-track composite chassis mode switching control method as described in claim 1, characterized in that, The construction of a three-dimensional terrain model of the terrain environment based on the terrain edges includes: The surface of the target 3D point cloud is reconstructed to obtain a 3D surface; Based on the terrain edges, the three-dimensional surface is adapted to obtain a three-dimensional terrain model of the terrain environment.

7. The wheel-track composite chassis mode switching control method as described in claim 1, characterized in that, The initial chassis control strategy for generating the target wheel-tracked hybrid machine based on the three-dimensional terrain model includes: The three-dimensional terrain model is divided into meshes to obtain a mesh node model; The terrain elevation difference of the target wheel-tracked hybrid machine is calculated based on the grid node model, and the ground type is determined. The optimal chassis mode of the target wheel-track hybrid machine is calculated based on the terrain elevation difference and the ground type. The initial chassis control strategy for the target wheel-track hybrid machine is generated based on the optimal chassis mode.

8. The wheel-track composite chassis mode switching control method as described in claim 1, characterized in that, The calculation of the terrain risk coefficient of the chassis control strategy includes: The grid node path of the target wheel-track hybrid machine is determined based on the chassis control strategy. Calculate the elevation risk coefficient and slope risk coefficient of the chassis control strategy based on the grid node path; A terrain risk coefficient is constructed based on the elevation risk coefficient and the slope risk coefficient.

9. The wheel-track composite chassis mode switching control method as described in claim 1, characterized in that, The optimization of the initial control strategy based on the terrain risk coefficient to obtain the optimal control strategy for the target wheel-tracked hybrid machine in the given terrain environment includes: When the terrain risk coefficient is greater than the preset risk coefficient threshold, return to the above steps of generating the initial chassis control strategy of the target wheel-tracked hybrid machine based on the three-dimensional terrain model to obtain the updated chassis control strategy; Calculate the terrain risk coefficient of the updated chassis control strategy; When the terrain risk coefficient of the updated chassis control strategy is less than or equal to the risk coefficient threshold, the updated chassis control strategy is taken as the optimal control strategy.

10. A mode switching control system for a wheel-track hybrid chassis, characterized in that, The system implementing the wheel-track composite chassis mode switching control method according to any one of claims 1-9 comprises: The moving point cloud removal module is used to acquire the environmental three-dimensional point cloud sequence and environmental image of the terrain environment where the target wheel-tracked composite machine is located, and to remove the moving point cloud from the environmental three-dimensional point cloud sequence to obtain the target three-dimensional point cloud. A multi-scale sampling module is used to perform multi-scale sampling of the target three-dimensional point cloud to obtain multi-scale sampling data of the terrain environment. A three-dimensional terrain model construction module is used to identify the terrain edges of the terrain environment based on the multi-scale sampling data and the environmental image, and to construct a three-dimensional terrain model of the terrain environment based on the terrain edges. The terrain risk coefficient calculation module is used to generate an initial chassis control strategy for the target wheel-tracked hybrid machine based on the three-dimensional terrain model, and to calculate the terrain risk coefficient of the chassis control strategy. The optimal control strategy generation module is used to optimize the initial control strategy based on the terrain risk coefficient to obtain the optimal control strategy for the target wheel-tracked hybrid machine in the terrain environment.

Images