A method, system, device, and medium for assisted driving
By acquiring the characteristic information and risk prediction model of puddles, the driver assistance system can accurately determine the actual impact of puddles on vehicle driving and adjust the driving mode to avoid splashing water onto pedestrians. This solves the problem of inaccurate judgment in existing technologies and improves the accuracy and safety of the driver assistance system.
Patent Information
- Application Number
- CN202611174689.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-04
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies cannot accurately determine whether water will splash onto pedestrians when a vehicle passes through a puddle in rainy or flooded conditions, reducing the accuracy of driver assistance systems in judging the surrounding impact in flooded scenarios.
By acquiring road condition information, spatial environment information, and vehicle driving information, the system identifies the characteristic information of puddles, including the radius of curvature, center depth, and opening radius. It then uses a risk prediction model to determine the risk level when a vehicle passes through a puddle and adjusts the driving mode according to the risk level to control the vehicle's safe passage through the puddles.
It enables accurate identification and risk assessment of puddles, improves the accuracy of the driver assistance system in judging the surrounding impact in waterlogged scenarios, and ensures that vehicles can safely pass through puddles without splashing pedestrians.
Smart Images

Figure CN123104657A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of artificial intelligence technology, and in particular to a method, system, device and medium for assisting driving. Background Technology
[0002] Currently, with the rapid development of intelligent driving technology, urban navigation assisted driving functions have been gradually applied to real-world road scenarios. Vehicles need to comprehensively perceive and intelligently make decisions about their surroundings while driving. In rainy or flooded conditions, water splashing when vehicles drive over puddles is a common uncivilized driving behavior in urban areas. This not only creates a negative experience for pedestrians but can also cause secondary safety hazards such as pedestrians slipping and falling while trying to avoid the splashing water.
[0003] However, existing technologies use image recognition to identify whether there are puddles ahead, but they cannot estimate the size of the puddles, and therefore cannot determine whether the water splashed when the vehicle passes through the puddles will wet pedestrians on the roadside, reducing the accuracy of the driver assistance system in judging the impact on the surroundings in waterlogged scenarios. Summary of the Invention
[0004] The summary section introduces a series of simplified concepts, which will be further explained in detail in the detailed description section. This summary section is not intended to limit the key and essential technical features of the claimed technical solution, nor is it intended to determine the scope of protection of the claimed technical solution.
[0005] In a first aspect, embodiments of this application provide an assisted driving method, the method comprising: Obtain road condition information, surrounding spatial environment information, and vehicle driving information for the area to be moved; Based on the road condition information, the characteristic information of the puddles in the area to be moved is determined. The characteristic information includes at least the radius of curvature of the puddles, the center depth of the puddles, and the opening radius of the puddles. The risk level of the vehicle traveling to the area to be moved is determined based on the feature information, the spatial environment information, the driving information, and the risk prediction model. The driving mode of the vehicle through the area to be moved is determined based on the risk level. Control the vehicle to pass through the location to be moved in the driving mode.
[0006] In one embodiment of the present invention, the road condition information is image information and point cloud information; determining the feature information of puddles in the area to be moved based on the road condition information includes: Based on the image information and the point cloud information, the original coordinate information of the water puddle in the reference coordinate system is determined. The original coordinate information includes multiple first coordinate points at the edge of the water puddle and multiple second coordinate points at the bottom of the water cooler. The reference coordinate system includes a first coordinate direction, a second coordinate direction, and a third coordinate direction. The third coordinate direction is the vertical direction. Based on multiple first coordinate point information and multiple second coordinate point information, the original coordinate information is fitted to the spherical equation using least squares to determine the coordinate point information of the center of the puddle and the radius of curvature. The center depth of the puddle is determined based on the reference coordinate point information, the sphere center coordinate point information, and the radius of curvature; wherein, the coordinate point in the first coordinate direction of the reference coordinate point information is the same as the coordinate point in the first coordinate direction of the sphere center coordinate point information, the coordinate point in the second coordinate direction of the reference coordinate point information is the same as the coordinate point in the second coordinate direction of the sphere center coordinate point information, and the coordinate point in the third coordinate direction of the reference coordinate point information is the same as the coordinate point in the third coordinate direction of the ground reference surface at the location of the puddle.
[0007] In one embodiment of the present invention, determining the risk level of the vehicle traveling to the area to be moved based on the feature information, the spatial environment information, the driving information, and a risk prediction model includes: The feature information and the driving information are input into the risk prediction model to determine the estimated splash range of the area to be moved. If it is determined based on the spatial environment information that there are no pedestrians in the estimated splash range, then the risk level is determined to be the first risk level. The first risk level indicates that when the vehicle passes through the area to be moved, there is no risk that water from the puddle will splash onto pedestrians. If it is determined based on the spatial environment information that there are pedestrians in the estimated splash range, then the risk level is determined to be the second risk level. The second risk level represents the risk that water in the puddle will splash onto pedestrians when the vehicle passes through the area to be moved.
[0008] In one embodiment of the present invention, the risk prediction model is constructed based on the following: Multiple sample data are acquired, including sample feature information of puddles, sample driving information of vehicles, and reference splash range formed when the vehicle passes through the puddles with the sample driving information. The risk prediction model is formed by training an initial neural network based on multiple sample data.
[0009] In one embodiment of the present invention, determining the driving mode of the vehicle passing through the area to be moved based on the risk level includes: If the risk level is the first risk level, then the vehicle is controlled to pass through the area to be moved according to the current driving mode. The first risk level indicates that when the vehicle passes through the area to be moved, there is no risk of water from the puddle splashing onto pedestrians. If the risk level is the second risk level, the driving mode of the vehicle is adjusted. The second risk level indicates that when the vehicle passes through the area to be moved, there is a risk that water from the puddle will splash onto pedestrians.
[0010] In one embodiment of the present invention, adjusting the driving mode of the vehicle includes: Obtain road surface information around the area to be moved; If the road surface information meets the preset conditions that allow the vehicle to change lanes, then control the vehicle to move to the lane adjacent to the area to be moved. If the road surface information does not meet the preset conditions that allow the vehicle to change lanes, then the vehicle is controlled to decelerate and pass through the area to be moved.
[0011] Secondly, this application proposes an assisted driving system, the system comprising: a data acquisition module, a risk prediction module, and a vehicle control module; The data acquisition module is configured to acquire road condition information of the area to be moved, spatial environment information of the area to be moved, and vehicle driving information. The risk prediction module is configured to: determine the characteristic information of the presence of puddles in the area to be moved based on the road condition information, wherein the characteristic information includes at least the radius of curvature of the puddles, the center depth of the puddles, and the opening radius of the puddles; and determine the risk level of the vehicle traveling to the area to be moved based on the characteristic information, the spatial environment information, the driving information, and the risk prediction model. The vehicle control module is configured to: determine the driving mode of the vehicle passing through the area to be moved based on the risk level; and control the vehicle to pass through the location to be moved in the driving mode.
[0012] Thirdly, an electronic device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program stored in the memory to implement the steps of an assisted driving method as described in any of the first aspects above.
[0013] Fourthly, this application also proposes a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of any of the first aspects of the assisted driving method.
[0014] Fifthly, this application proposes a computer program product, including a computer program that, when executed by a processor, implements the steps of an assisted driving method according to any one of the first aspects.
[0015] In summary, the assisted driving method of this application embodiment acquires road condition information of the area to be moved, spatial environment information of the area to be moved, and vehicle driving information. Based on the road condition information, it determines the characteristic information of the presence of puddles in the area to be moved. The characteristic information includes at least the radius of curvature of the puddles, the center depth of the puddles, and the opening radius of the puddles, thus achieving accurate identification of puddles. Based on the characteristic information, the spatial environment information, the driving information, and a risk prediction model, it determines the risk level of the vehicle driving to the area to be moved. The risk prediction model infers based on the characteristic information of the puddles, the spatial environment information, and the driving information. Compared with the prior art, which can only output the existence of puddles, it can accurately determine the actual impact of puddles on vehicle driving, thereby effectively improving the accuracy of the assisted driving system in judging the surrounding impact in water accumulation scenarios.
[0016] The assisted driving method proposed in this application, other advantages, objectives and features of this application will be partly apparent from the following description, and partly understood by those skilled in the art through study and practice of this application. Attached Figure Description
[0017] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit this specification. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A flowchart illustrating an assisted driving method provided in an embodiment of this application; Figure 2 This is a schematic diagram of the structure of an assisted driving system provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of an electronic driver assistance device provided in an embodiment of this application. Detailed Implementation
[0018] To better understand the technical solutions provided in the embodiments of this specification, the technical solutions of the embodiments of this specification will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of this specification and the specific features in the embodiments are detailed descriptions of the technical solutions of the embodiments of this specification, rather than limitations on the technical solutions of this specification. In the absence of conflict, the embodiments of this specification and the technical features in the embodiments can be combined with each other.
[0019] In this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, without necessarily requiring or implying any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. The term "two or more" includes two or more cases.
[0020] Please see Figure 1 This is a schematic diagram of a driver assistance method provided in an embodiment of this application, specifically including: S110. Obtain road condition information of the area to be moved, spatial environment information of the area to be moved, and vehicle driving information; For example, the vehicle has an image acquisition unit and a lidar detection unit, which are located on the upper inner side of the vehicle's windshield. During the vehicle's movement, the vehicle can acquire road condition information of the area to be moved, spatial environment information of the area to be moved, and vehicle driving information. The road condition information reflects whether there are puddles and their basic morphological features in the area to be moved, and is obtained by the image acquisition unit and lidar detection unit sensing the road surface in front of the vehicle's driving direction. The spatial environment information reflects whether there are pedestrians around the area to be moved and their position and movement status, and is obtained by the image acquisition unit and millimeter-wave radar detection unit sensing the area around the vehicle. The vehicle driving information reflects the vehicle's current movement status and is acquired through the vehicle's CAN (Controller Area Network) bus. The driving information includes at least the current vehicle speed and longitudinal acceleration.
[0021] S120. Based on the road condition information, determine the characteristic information of the water pit in the area to be moved, wherein the characteristic information includes at least the radius of curvature of the water pit, the center depth of the water pit, and the opening radius of the water pit. For example, determining the characteristic information of puddles in the area to be moved based on the road condition information involves: treating the puddles in the area to be moved as a spherical cap structure, i.e., the concave part formed by cutting a sphere with a plane; and extracting three key parameters, i.e., characteristic information, that can completely characterize the physical features of the puddles by fitting the spatial geometry of the concave part. Specifically, the radius of curvature reflects the steepness of the puddle concavity; a smaller radius of curvature indicates a steeper puddle, and the more easily water is squeezed and splashed when a vehicle passes over it. The center depth reflects the puddle's water storage capacity; a greater depth means more water is stored in the puddle. The opening radius reflects the lateral extension range of the puddle on the road surface; a larger opening radius indicates a wider road surface covered by the puddle, and a larger contact area between the tires and the puddle when a vehicle passes over it.
[0022] S130. Based on the feature information, the spatial environment information, the driving information, and the risk prediction model, determine the risk level of the vehicle driving to the area to be moved; For example, the feature information of the puddle and the vehicle's driving information obtained through the vehicle's CAN bus are input into a pre-trained risk prediction model. The risk prediction model infers the relationship between a large number of sample feature information of puddle and sample driving information of vehicles and the corresponding reference splash range, and outputs the estimated splash range corresponding to the area to be moved. On this basis, combined with the pedestrian position contained in the obtained spatial environment information, it is determined whether the pedestrian is located within the estimated splash range, thereby determining the risk level of the area to be moved.
[0023] S140. Determine the driving mode of the vehicle when passing through the area to be moved based on the risk level. For example, after determining the risk level of the area to be moved, the vehicle needs to decide which driving mode to use to pass through the area based on that risk level. Specifically, the risk level characterizes whether there is a risk of water from puddles splashing onto pedestrians when the vehicle passes through the area; therefore, different risk levels correspond to different driving modes. This allows the vehicle to execute the appropriate driving mode when facing different risk levels.
[0024] S150, Control the vehicle to pass through the position to be moved in the driving mode.
[0025] For example, after determining the driving mode that the vehicle should adopt when passing through the area to be moved, the vehicle's control system performs longitudinal and / or lateral motion control on the vehicle according to the control strategy corresponding to the driving mode, so that the vehicle actually passes through the area to be moved in the driving mode, thereby completing the entire assisted driving closed loop from risk identification to risk response, ensuring that the vehicle passes through the area to be moved with puddles in a way that matches the risk level safely and compliantly.
[0026] In summary, the assisted driving method proposed in this application acquires road condition information of the area to be moved, spatial environment information of the area to be moved, and vehicle driving information. Based on the road condition information, it determines the characteristic information of the presence of puddles in the area to be moved. The characteristic information includes at least the radius of curvature of the puddles, the center depth of the puddles, and the opening radius of the puddles, thus achieving accurate identification of puddles. Based on the characteristic information, the spatial environment information, the driving information, and a risk prediction model, it determines the risk level of the vehicle driving to the area to be moved. The risk prediction model infers based on the characteristic information of the puddles, the spatial environment information, and the driving information. Compared with existing technologies that can only output the existence of puddles, it can accurately determine the actual impact of puddles on vehicle driving, thereby effectively improving the accuracy of the assisted driving system in judging the surrounding impact in waterlogged scenarios.
[0027] In some examples, the road condition information is image information and point cloud information; based on the road condition information, determining the feature information of puddles in the area to be moved includes: Based on the image information and the point cloud information, the original coordinate information of the water puddle in the reference coordinate system is determined. The original coordinate information includes multiple first coordinate points at the edge of the water puddle and multiple second coordinate points at the bottom of the water cooler. The reference coordinate system includes a first coordinate direction, a second coordinate direction, and a third coordinate direction. The third coordinate direction is the vertical direction. Based on multiple first coordinate point information and multiple second coordinate point information, the spherical equation is fitted by least squares to determine the coordinate point information of the center of the puddle and the radius of curvature. The center depth of the puddle is determined based on the reference coordinate point information, the sphere center coordinate point information, and the radius of curvature; wherein, the coordinate point in the first coordinate direction of the reference coordinate point information is the same as the coordinate point in the first coordinate direction of the sphere center coordinate point information, the coordinate point in the second coordinate direction of the reference coordinate point information is the same as the coordinate point in the second coordinate direction of the sphere center coordinate point information, and the coordinate point in the third coordinate direction of the reference coordinate point information is the same as the coordinate point in the third coordinate direction of the ground reference surface at the location of the puddle.
[0028] For example, the road condition information specifically includes image information and point cloud information. The image information is acquired through an image acquisition unit (such as a vehicle-mounted camera), and the point cloud information is acquired through a lidar detection unit. Specifically, after obtaining the image information and point cloud information, semantic segmentation processing is first performed on the image information. Each pixel in the image information is classified through a semantic segmentation network to identify the set of pixels belonging to the puddle region, generating a binary mask for the puddle. This binary mask is used to represent the two-dimensional contour of the puddle in the image coordinate system. Based on this, the binary mask of the puddle is mapped to the three-dimensional space where the lidar point cloud information is located. The original coordinate information corresponding to the puddle is extracted from the lidar point cloud information. The original coordinate information includes multiple first coordinate points located at the edge of the puddle and multiple second coordinate points located at the bottom of the puddle. The reference coordinate system is a three-dimensional spatial coordinate system, including a first coordinate direction, a second coordinate direction, and a third coordinate direction, where the third coordinate direction is the vertical direction and is used to represent the height information of the spatial points.
[0029] After obtaining the original coordinate information of the puddle in the reference coordinate system, a spherical cap structure is used as the geometric model of the puddle. That is, the spatial shape of the puddle is assumed to be the concave portion formed by a sphere being truncated by a plane. Based on multiple first and second coordinate points from the original coordinate information, a spherical equation is constructed, and a least-squares fit is performed on this equation to obtain the coordinates of the puddle's center point and the optimal sphere radius. This optimal radius is the radius of curvature of the puddle, which characterizes the steepness of the concavity. A smaller radius of curvature indicates a steeper puddle, and the more easily water is squeezed and splashed when a vehicle passes over it. The spherical equation is: Where (x, y, z) are the original coordinate information. It contains the coordinates of the sphere's center, and R is the radius of curvature.
[0030] After determining the optimal center coordinates and radius of curvature, the height of the ground reference plane at the center projection of the center coordinate information is obtained. The ground reference plane is a plane obtained by performing ground fitting processing on the point cloud information. It is obtained by fitting the road surface points in the point cloud information through a random sampling consensus algorithm and is used to characterize the normal road surface height of the area to be moved when there are no puddles. Specifically, the center projection is used as the reference coordinate point information, which specifically refers to the projection position of the center coordinate point information on the ground reference plane. The first coordinate direction coordinate value of the reference coordinate point information is the same as the first coordinate direction coordinate point of the center coordinate point information. The second coordinate direction coordinate point of the reference coordinate point information is the same as the second coordinate direction coordinate point of the center coordinate point information. The third coordinate direction coordinate point of the reference coordinate point information is the height of the ground reference plane at the projection position. Based on this, the difference between the height and the third coordinate direction coordinate value of the center coordinate point information is calculated to obtain the vertical distance from the center coordinate point information to the ground reference plane. Then, the difference between the radius of curvature and the vertical distance is calculated to obtain the center depth of the puddle, which is shown in Equation (1). Then, the difference between the radius of curvature and the difference data is taken as the center depth of the water pit. That is, the center depth is equal to the radius of curvature minus the vertical distance from the center of the sphere to the ground reference plane. This center depth is used to characterize the water storage capacity of the water pit. The greater the depth, the more water is stored in the water pit.
[0031] (1); Where d is the center depth, Let be the radius of curvature. The coordinates of the reference coordinate point are in the third coordinate direction.
[0032] Finally, the opening radius of the puddle area is determined based on the radius of curvature and the center depth. Specifically, it utilizes the geometric relationship of a spherical cap, where the opening radius is the square root of the difference between the radius of curvature and the center depth, minus the square of the radius of curvature. This opening radius characterizes the lateral extent of the puddle on the road surface. Alternatively, the opening radius can be converted from the visual 2D contour area S. Where r is the opening radius and S is the visual 2D contour area of the spherical cap. By fusing image information with point cloud information, the original coordinate information of the puddle in the reference coordinate system is obtained by utilizing their spatial mapping relationship. This overcomes the shortcomings of a single visual sensor in obtaining depth information and a single LiDAR in distinguishing between puddle and road surface, thus improving the accuracy of puddle perception. Secondly, by using the spherical cap structure as the geometric assumption model of the puddle and employing the least squares fitting method to solve for the optimal center coordinates and radius of curvature, the irregular real-world puddle is transformed into a standard geometric body with a clear mathematical expression. Thirdly, when solving for the center depth, the difference between the height of the ground reference plane projected at the center of the sphere and the height of the center of the sphere is combined to ensure that the center depth can truly reflect the actual water storage depth of the puddle, avoiding depth estimation errors caused by road surface slope or fitting deviation.
[0033] In some examples, the risk level of the vehicle traveling to the area to be moved is determined based on the feature information, the spatial environment information, the driving information, and the risk prediction model, including: The feature information and the driving information are input into the risk prediction model to determine the estimated splash range of the area to be moved. If it is determined based on the spatial environment information that there are no pedestrians in the estimated splash range, then the risk level is determined to be the first risk level. The first risk level indicates that when the vehicle passes through the area to be moved, there is no risk that water from the puddle will splash onto pedestrians. If it is determined based on the spatial environment information that there are pedestrians in the estimated splash range, then the risk level is determined to be the second risk level. The second risk level represents the risk that water in the puddle will splash onto pedestrians when the vehicle passes through the area to be moved.
[0034] For example, after obtaining the feature information of the puddle and the vehicle's driving information, the feature information of the puddle and the driving information of the vehicle are jointly input into a pre-trained risk prediction model to determine the estimated splash range when the vehicle travels to the area to be moved. The risk prediction model is a neural network model pre-trained based on the mapping relationship between a large number of sample feature information of puddles and sample driving information of vehicles and the corresponding reference splash distance. The vehicle's driving information includes the vehicle tire's water entry angle (relative to the puddle normal), water depth coefficient (estimated by point cloud reflectivity), and tire tread drainage coefficient (preset constant or according to vehicle model configuration). Its inputs are the radius of curvature of the puddle, the center depth of the puddle, the opening radius of the puddle, the vehicle's current speed, longitudinal acceleration, the vehicle tire's water entry angle, water depth coefficient, and tire tread drainage coefficient. The risk prediction model includes 4 hidden layers with the number of neurons [128, 256, 256, 128]. An identity mapping is introduced between the second and third hidden layers to prevent gradient vanishing, accelerate convergence, and improve generalization ability for extreme conditions (such as high-speed deep water). Activation function: GELU (Gaussian Error Linear Unit) is used, which has better smoothness and non-monotonicity than ReLU (Rectified Linear Unit), making it more suitable for fitting continuously changing curves in fluid dynamics. Regularization: Each layer is followed by Batch Normalization (BN) and Dropout (p=0.1) to prevent overfitting. Dropout (p=0.1) randomly deactivates neurons with a 10% probability, preventing overfitting and improving generalization ability. The risk prediction model uses a Mixture Density Network (MDN) head, outputting the mean and variance of a Gaussian distribution. When the input parameters are in a sparse training data region (such as an extremely deep puddle), the risk prediction model can output high variance, prompting the system to reduce confidence and adopt a more conservative defensive driving strategy. The output is the maximum splash distance and splash angle generated when the vehicle passes through the puddle in its current driving state. The maximum splash distance and splash angle together constitute the estimated splash range. Specifically, the maximum splash distance refers to the farthest splash distance of water from the edge of the puddle in the vehicle's driving direction; the splash angle refers to the deflection angle of the water's flight direction relative to the vehicle's driving direction.
[0035] Based on the spatial environment information, it is determined whether there are pedestrians within the estimated splash range, and the risk level of the vehicle traveling to the area to be moved is determined according to the determination result. The spatial environment information includes the location information of pedestrians sensed by the image acquisition unit and the millimeter-wave radar detection unit, which reflects the actual spatial coordinates of the pedestrians around the area to be moved. The specific determination method is as follows: The pedestrian's location information is spatially compared with the estimated splash range. If the pedestrian's location is outside the spatial range jointly defined by the maximum splash distance and the splash angle, or the risk probability of water splashing from the puddle to the pedestrian is less than the risk probability threshold (0.7), meaning the pedestrian is in a safe area, then it is determined that the pedestrian does not exist within the estimated splash range. In this case, the risk level is determined to be the first risk level. The first risk level indicates that when the vehicle passes through the area to be moved, there is no risk of water splashing from the puddle to the pedestrian, and the vehicle does not need to perform any additional avoidance maneuvers. If the pedestrian's location is within the spatial range jointly defined by the maximum splash distance and the splash angle, meaning the pedestrian is in a potential splash area, then it is determined that the pedestrian exists within the estimated splash range. In this case, the risk level is determined to be the second risk level. The second risk level indicates that when the vehicle passes through the area to be moved, there is a risk of water splashing from the puddle to the pedestrian, and the vehicle needs to further determine an avoidance strategy to eliminate or reduce this risk. By using the above method, the splash range output by the risk prediction model is spatially determined with the perceived pedestrian position, thus achieving an accurate prediction of whether a vehicle will splash pedestrians when passing through a puddle.
[0036] In some examples, the risk prediction model is constructed based on the following: Multiple sample data are acquired, including sample feature information of puddles, sample driving information of vehicles, and reference splash range formed when the vehicle passes through the puddles with the sample driving information. The risk prediction model is formed by training an initial neural network based on multiple sample data.
[0037] For example, the risk prediction model is constructed using a technique that separates offline simulation from online inference. The computationally intensive fluid dynamics simulation is completed offline, enabling the trained risk prediction model to achieve millisecond-level real-time inference during vehicle operation. Specifically, the SPH (Smooth Particle Hydrodynamics) method is used in the simulation environment setup because it excels at handling free surface flow and large deformation splashing problems. First, sample feature information of puddles affecting the splash distance and sample vehicle driving information are determined. Each puddle's sample feature information and vehicle driving information corresponds to a sampling point in a multi-dimensional parameter space, representing a specific combination of puddle geometry and vehicle motion state. Based on this, the Latin hypercube sampling method is used to perform hierarchical equal-probability sampling of this space, generating approximately 50,000-100,000 independent simulation conditions. The Latin hypercube sampling method divides each parameter dimension into multiple equal-probability intervals and performs sampling without replacement within each interval, ensuring that sampling points are evenly distributed throughout the parameter space and avoiding over-concentration of sampling points in a local area, which could lead to blind spots in the training data. Among them, the sample feature information includes the sample center depth d of the puddle. [5,60]mm (step size 5mm), the sample radius of curvature R of the puddle. [0.5,15]m (logarithmic sampling) and shape factor (eccentricity) [0,0.5]. The vehicle's sample driving information includes speed v. [10,90]km / h (step length 5km / h), angle of entry into water θ [ [20º, 20º], tire type (summer tire or rain tire). Environmental variables are road surface friction coefficient and water depth coverage ratio.
[0038] After obtaining sample feature information of multiple sets of puddles and sample driving information of vehicles, the sample feature information of each set of puddles and sample driving information of vehicles are input into a cloud server cluster. Utilizing the parallel computing capabilities of the cloud server cluster, a fluid dynamics simulation based on the smooth particle fluid dynamics method is independently performed on the sample feature information of each set of puddles and the sample driving information of vehicles. This simulates the entire process of water splash particle generation, movement, and dissipation after the vehicle tires come into contact with the puddles, generating water splash particle trajectory data corresponding to the sample feature information of each set of puddles and the sample driving information of vehicles. The water splash particle trajectory data records the sequence of position changes of each water splash particle in space during the simulation time. After obtaining the water splash particle trajectory data, automated post-processing is performed on each set of simulation results to extract the reference splash distance of the water splash particles in the vehicle's driving direction from the water splash particle trajectory data. The sample feature information of each group of puddles and the sample driving information of vehicles, along with their corresponding reference splash ranges, are correlated to form multiple sample datasets for supervised learning. Each set of sample datasets includes the sample feature information of puddles and the sample driving information of vehicles as input features, and the reference splash range as output. A script is written to analyze the particle cloud, extracting the 95th quantile distance, the angle with the highest particle density, and the particle flux at a specific distance. Sample datasets that do not converge or exhibit numerical explosions (approximately 1-2%) are removed. A mirror-symmetric transformation is performed on the existing sample dataset (for the left and right puddles), and Gaussian noise is added to simulate sensor measurement errors, expanding the sample dataset to 200,000 samples. Based on the multiple sets of sample datasets, an initial neural network is iteratively trained. Using the sample feature information of puddles and the sample driving information of vehicles in the sample dataset as input, and the corresponding reference splash range as the supervision signal, the network weights of the initial neural network are adjusted through a backpropagation algorithm until the error between the actual reference splash range output by the network and the corresponding reference splash range label converges to a preset range. The trained initial neural network is then determined as the risk prediction model.
[0039] The loss function design employs a weighted multi-task loss function, including mean squared error, binary cross-entropy, and a regularization term. Mean squared error is used for regression tasks. Binary cross-entropy is used for probabilistic classification tasks (label defined as 1 if the pedestrian is within the splash zone, 0 otherwise). The regularization term prevents excessively large weights. Early training focuses on geometric parameter fitting, while later training emphasizes risk probability classification accuracy. The optimizer uses AdamW (Adam with Weight Decay, an adaptive moment estimation optimization algorithm with weight decay). The scheduling strategy is Cosine Annealing Warm Restarts (a learning rate scheduling strategy combining cosine annealing and warm restart), restarting every 10 epochs (training rounds) to help escape local optima. The batch size is 256 or 512 (depending on GPU memory). An early stopping mechanism monitors the validation set loss; if there is no decrease after 20 consecutive epochs, training stops. Deployment optimizations include: Quantization-Aware Training (QAT): Introducing INT8 (8-bit integer) quantization awareness in the final stage of training to simulate a low-precision inference environment, ensuring that the accuracy loss is less than 1% when the model is deployed to an onboard DSP (Digital Signal Processor) or NPU (Neural Processing Unit). Removing low-contribution neuron channels further compresses the model size to less than 500KB.
[0040] By employing the Latin hypercube sampling method to perform hierarchical and equal-probability sampling of the multidimensional parameter space, uniform coverage of the sample feature information of puddles and the sample driving information of vehicles is ensured within the multidimensional parameter space. This avoids the problem of prediction failure of the risk prediction model under specific working conditions due to uneven data distribution, and improves the generalization ability of the risk prediction model for puddles of different shapes and different vehicle speeds. By executing a large number of fluid dynamics simulation operations in parallel through a cloud server cluster, the tens of thousands of simulation tasks that would have taken years to complete in a single-machine environment are shortened to a few hours or days, improving efficiency. The risk prediction model trained on this basis involves only simple matrix operations of lightweight neural networks in its inference process, and can complete forward inference on an onboard embedded chip with millisecond latency, thereby improving prediction accuracy and computational efficiency.
[0041] In some examples, determining the vehicle's driving pattern through the area to be moved based on the risk level includes: If the risk level is the first risk level, then the vehicle is controlled to pass through the area to be moved according to the current driving mode. The first risk level indicates that when the vehicle passes through the area to be moved, there is no risk of water from the puddle splashing onto pedestrians. If the risk level is the second risk level, the driving mode of the vehicle is adjusted. The second risk level indicates that when the vehicle passes through the area to be moved, there is a risk that water from the puddle will splash onto pedestrians.
[0042] For example, a vehicle needs to decide which driving mode to use when passing through a puddles based on the risk level. Specifically, the risk level indicates whether there is a risk of water splashing onto pedestrians when the vehicle passes through the puddles. Therefore, different risk levels correspond to different driving modes. When the risk level is the first risk level, it indicates that there is no risk of water splashing onto pedestrians when the vehicle passes through the puddles. That is, based on spatial environmental information, it is determined that there are no pedestrians within the estimated splash range. In this case, there is no need to adjust the vehicle's current driving mode; the vehicle can simply pass through the puddles according to the current driving mode. When the risk level is the second risk level, it indicates that there is a risk of water splashing onto pedestrians when the vehicle passes through the puddles. That is, based on spatial environmental information, it is determined that there are pedestrians within the estimated splash range. In this case, the vehicle's driving mode needs to be adjusted to avoid splashing water onto pedestrians. In this way, the vehicle can execute a matching driving mode when facing different risk levels, ensuring traffic efficiency when there is no risk and adjusting the driving mode when there is risk to avoid water splashing onto pedestrians.
[0043] In some examples, adjusting the vehicle's driving mode includes: Obtain road surface information around the area to be moved; If the road surface information meets the preset conditions that allow the vehicle to change lanes, then control the vehicle to move to the lane adjacent to the area to be moved. If the road surface information does not meet the preset conditions that allow the vehicle to change lanes, then the vehicle is controlled to decelerate and pass through the area to be moved.
[0044] For example, under the second risk level, the vehicle needs to adjust its driving mode to avoid splashing water from puddles onto pedestrians. First, the vehicle acquires road surface information around the area to be moved. This information reflects the traffic conditions and road geometry of the lanes surrounding the area, obtained through image acquisition and lidar detection units sensing the sides and rear of the vehicle. It includes at least the lane line attributes of adjacent lanes, the presence of other obstacles in adjacent lanes, and the movement state of those obstacles. After acquiring this road surface information, the vehicle determines whether preset conditions for lane changing are met. These preset conditions include that the lane lines of the adjacent lanes are dashed, there are no other vehicles in the adjacent lanes, or the distance between any existing vehicles and the vehicle is greater than a preset safe distance threshold, and the width of the adjacent lanes is sufficient for the vehicle to pass safely. If the road surface information simultaneously meets all the aforementioned preset conditions, it is determined that the road surface information allows the vehicle to change lanes. At this time, the vehicle controls itself to move from its current lane to a lane adjacent to the area to be moved, i.e., it avoids puddles in the area to be moved through lateral displacement, thereby preventing direct contact between the vehicle's tires and the puddles and fundamentally eliminating the source of water splashes. During the lane change process, a fifth-order polynomial trajectory is generated, causing the vehicle's lateral offset to... Sufficient to allow the tires to completely avoid puddles (considering vehicle width and safety margin), while also restraining lateral acceleration. accelerometer The vehicle plans a smooth lateral movement trajectory based on its current speed and lateral offset, enabling it to complete lane changes with lateral acceleration that does not exceed comfort constraints, thus ensuring the safety and smoothness of lane-changing operations.
[0045] If the road surface information does not meet the preset conditions for allowing the vehicle to change lanes, such as the adjacent lane line being a solid line, other vehicles existing in the adjacent lane and their distance from the vehicle being less than or equal to a preset safe distance threshold, or the width of the adjacent lane being insufficient to allow the vehicle to pass safely, then the road surface information determines that the vehicle does not allow to change lanes. In this case, the vehicle cannot bypass the puddle by lateral displacement and instead executes a deceleration strategy. Specifically, the vehicle determines a safe passage speed based on the puddle's characteristic information and the vehicle's driving information. This safe passage speed is configured such that the estimated splash range generated when the vehicle passes through the puddle at this speed is less than or equal to the actual distance between the pedestrian and the puddle, ensuring that the furthest splash distance of water will not reach the pedestrian's location, thereby avoiding water splashing onto the pedestrian. After determining the safe passage speed, the vehicle plans a longitudinal deceleration trajectory based on the difference between the current vehicle speed and the safe passage speed. The deceleration of the longitudinal deceleration trajectory is constrained within a preset comfort deceleration range to ensure the smoothness of the deceleration process. The vehicle performs longitudinal motion control according to the stated longitudinal deceleration trajectory, reducing its speed to less than or equal to the stated safe passage speed before actually reaching the puddle location. It then passes through the area to be moved at this safe speed, thereby reducing the risk of splashing water on pedestrians when lane changes are not possible. The decision threshold is dynamically adjusted using the uncertainty variance of the model output. When the model is uncertain, the risk probability threshold for triggering avoidance is automatically reduced (e.g., from 0.7 to 0.5).
[0046] One to two seconds before the action is executed, a "Water detected, pedestrian avoidance in progress" icon is displayed on the HUD (Head-Up Display) accompanied by a soft warning sound. The driver then takes over (by turning the steering wheel or pressing the accelerator pedal), immediately disengaging from automatic control and logging the process. The planned trajectory (lateral) and speed curve (longitudinal) are converted into steering angle and throttle / brake pressure commands, which are sent to the drive-by-wire chassis. The vehicle's actual execution status is monitored; if the deviation from the planned trajectory exceeds a threshold, the planning process is retried. Once the vehicle has completely passed the puddle area and there is no further risk, it automatically resumes the original CNOA (CityNavigation On-Road Assist) cruise control mode. By prioritizing the feasibility of lane changes and providing two different levels of avoidance strategies—lane change and deceleration—the system controls vehicles to move to adjacent lanes to bypass puddles when lane changes are feasible, fundamentally eliminating the source of water splashes without affecting traffic efficiency. When lane changes are not feasible, the system controls vehicles to decelerate over puddles to reduce the intensity of water splashes, ensuring that the splash range is less than or equal to the distance between pedestrians and puddles, thus preventing water splashes on pedestrians even when detours are not possible.
[0047] like Figure 2As shown, this application proposes an assisted driving system, which includes: a data acquisition module 21, a risk prediction module 22, and a vehicle control module 23; The data acquisition module 21 is configured to acquire road condition information of the area to be moved, spatial environment information around the area to be moved, and vehicle driving information. The risk prediction module 22 is configured to: determine the characteristic information of the presence of puddles in the area to be moved based on the road condition information, wherein the characteristic information includes at least the radius of curvature of the puddles, the center depth of the puddles, and the opening radius of the puddles; and determine the risk level of the vehicle driving to the area to be moved based on the characteristic information, the spatial environment information, the driving information, and the risk prediction model. The vehicle control module 23 is configured to: determine the driving mode of the vehicle passing through the area to be moved based on the risk level; and control the vehicle to pass through the location to be moved in the driving mode.
[0048] The effects of applying the aforementioned method in the above system can be found in the description of the aforementioned method embodiments, and will not be repeated here.
[0049] like Figure 3 As shown, this application embodiment also provides an electronic device 300, including a memory 310, a processor 320, and a computer program 311 stored in the memory 310 and executable on the processor. When the processor 320 executes the computer program 311, it implements the steps of any of the above-described assisted driving methods.
[0050] Since the electronic device described in this embodiment is a device used to implement an assisted driving device in the embodiments of this application, those skilled in the art can understand the specific implementation method and various variations of the electronic device in this embodiment based on the method described in the embodiments of this application. Therefore, how the electronic device implements the method in the embodiments of this application will not be described in detail here. Any device used by those skilled in the art to implement the method in the embodiments of this application falls within the scope of protection of this application.
[0051] In practical implementation, when the computer program 311 is executed by the processor, it can achieve the following: Figure 1 Any of the corresponding implementation methods in the embodiments.
[0052] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.
[0053] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-readable program code.
[0054] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0055] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0056] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0057] This application also provides a computer program product, which includes computer software instructions that, when executed on a processing device, cause the processing device to execute the LDPC decoding method of a solid-state drive controller.
[0058] A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk (SSD)).
[0059] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.
[0060] In the several embodiments provided in this application, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between devices or units, and may be electrical, mechanical, or other forms.
[0061] The units described as separate components may or may not be physically separate. The components shown as units 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 units can be selected to achieve the purpose of this embodiment according to actual needs.
[0062] Furthermore, the functional units in the various embodiments of this application 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 as a software functional unit.
[0063] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0064] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.
[0065] Although preferred embodiments have been described in this specification, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this specification.
[0066] Obviously, those skilled in the art can make various modifications and variations to this specification without departing from its spirit and scope. Therefore, if such modifications and variations fall within the scope of the claims and their equivalents, this specification is also intended to include such modifications and variations.
Claims
1. A driving assistance method, characterized in that, The method includes: Obtain road condition information, surrounding spatial environment information, and vehicle driving information for the area to be moved; Based on the road condition information, the characteristic information of the puddles in the area to be moved is determined. The characteristic information includes at least the radius of curvature of the puddles, the center depth of the puddles, and the opening radius of the puddles. The risk level of the vehicle traveling to the area to be moved is determined based on the feature information, the spatial environment information, the driving information, and the risk prediction model. The driving mode of the vehicle through the area to be moved is determined based on the risk level. Control the vehicle to pass through the location to be moved in the driving mode.
2. The assisted driving method according to claim 1, characterized in that, The road condition information includes image information and point cloud information; Based on the road condition information, the characteristic information of the presence of puddles in the area to be moved includes: Based on the image information and the point cloud information, the original coordinate information of the water puddle in the reference coordinate system is determined. The original coordinate information includes multiple first coordinate points at the edge of the water puddle and multiple second coordinate points at the bottom of the water cooler. The reference coordinate system includes a first coordinate direction, a second coordinate direction, and a third coordinate direction. The third coordinate direction is the vertical direction. Based on multiple first coordinate point information and multiple second coordinate point information, the spherical equation is fitted by least squares to determine the coordinate point information of the center of the puddle and the radius of curvature. The center depth of the puddle is determined based on the reference coordinate point information, the sphere center coordinate point information, and the radius of curvature; wherein, the coordinate point in the first coordinate direction of the reference coordinate point information is the same as the coordinate point in the first coordinate direction of the sphere center coordinate point information, the coordinate point in the second coordinate direction of the reference coordinate point information is the same as the coordinate point in the second coordinate direction of the sphere center coordinate point information, and the coordinate point in the third coordinate direction of the reference coordinate point information is the same as the coordinate point in the third coordinate direction of the ground reference surface at the location of the puddle.
3. The assisted driving method according to claim 1, characterized in that, Based on the feature information, the spatial environment information, the driving information, and the risk prediction model, the risk level of the vehicle driving to the area to be moved is determined, including: The feature information and the driving information are input into the risk prediction model to determine the estimated splash range of the area to be moved. If it is determined based on the spatial environment information that there are no pedestrians in the estimated splash range, then the risk level is determined to be the first risk level. The first risk level indicates that when the vehicle passes through the area to be moved, there is no risk that water from the puddle will splash onto pedestrians. If it is determined based on the spatial environment information that there are pedestrians in the estimated splash range, then the risk level is determined to be the second risk level. The second risk level represents the risk that water in the puddle will splash onto pedestrians when the vehicle passes through the area to be moved.
4. The assisted driving method according to claim 3, characterized in that, The risk prediction model is constructed based on the following method: Multiple sample data are acquired, including sample feature information of puddles, sample driving information of vehicles, and reference splash range formed when the vehicle passes through the puddles with the sample driving information. The risk prediction model is formed by training an initial neural network based on multiple sample data.
5. The assisted driving method according to claim 1, characterized in that, Determining the vehicle's driving mode through the area to be moved based on the risk level includes: If the risk level is the first risk level, then the vehicle is controlled to pass through the area to be moved according to the current driving mode. The first risk level indicates that when the vehicle passes through the area to be moved, there is no risk of water from the puddle splashing onto pedestrians. If the risk level is the second risk level, the driving mode of the vehicle is adjusted. The second risk level indicates that when the vehicle passes through the area to be moved, there is a risk that water from the puddle will splash onto pedestrians.
6. The assisted driving method according to claim 5, characterized in that, Adjusting the driving mode of the vehicle includes: Obtain road surface information around the area to be moved; If the road surface information meets the preset conditions that allow the vehicle to change lanes, then control the vehicle to move to the lane adjacent to the area to be moved. If the road surface information does not meet the preset conditions that allow the vehicle to change lanes, then the vehicle is controlled to decelerate and pass through the area to be moved.
7. A driver assistance system, characterized in that, The system includes: a data acquisition module, a risk prediction module, and a vehicle control module; The data acquisition module is configured to acquire road condition information of the area to be moved, spatial environment information of the area to be moved, and vehicle driving information. The risk prediction module is configured to: determine the characteristic information of the presence of puddles in the area to be moved based on the road condition information, wherein the characteristic information includes at least the radius of curvature of the puddles, the center depth of the puddles, and the opening radius of the puddles; and determine the risk level of the vehicle traveling to the area to be moved based on the characteristic information, the spatial environment information, the driving information, and the risk prediction model. The vehicle control module is configured to: determine the driving mode of the vehicle passing through the area to be moved based on the risk level; and control the vehicle to pass through the location to be moved in the driving mode.
8. An electronic device, comprising: The memory and processor are characterized in that the processor is used to implement the steps of a driving assistance method as described in any one of claims 1-6 when executing a computer program stored in the memory.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the assisted driving method as described in any one of claims 1-6.
10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the assisted driving method as described in any one of claims 1 to 6.
