Wading risk identification method, wading risk identification device and vehicle

Abstract

The invention provides a wading risk identification method, a wading risk identification device and a vehicle, and belongs to the technical field of wading detection of vehicles. The method is applied to a vehicle, and according to the embodiment of the invention, the wading related information including wading probability distribution determined through a first model, a wading risk score determined through a second model and a wading clustering result is determined through acquired multi-dimensional target state information; according to the method and the device, the target wading risk level corresponding to the vehicle is determined based on the wading related information, various wading related information can be determined through multi-dimensional target state information, the wading risk level of the vehicle can be accurately detected through multi-source information fusion and multi-level risk assessment, the comprehensiveness of vehicle risk perception is remarkably improved, and meanwhile, the wading risk level of the vehicle can be accurately detected. The method can effectively deal with a complex scene that a single sensor fails or data is incomplete, enhances the robustness and adaptability of the vehicle in extreme weather, and provides accurate safety early warning for wading driving of the vehicle.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of water detection of vehicles, and more particularly, to a water risk identification method, a water risk identification device and a vehicle in the technical field of water detection of vehicles. BACKGROUND

[0002] With the development of vehicle technology, more and more vehicles are equipped with intelligent systems to cope with various complex road environments.

[0003] In the related art, the intelligent system of the vehicle mainly relies on sensors such as cameras, radars and lidars to realize lane line recognition, obstacle detection, adaptive cruise and automatic emergency braking functions. However, in the scene of the water section, due to the unpredictability of the road, high risk level and dynamic change, etc., the current water risk level cannot be accurately detected, and there are problems of low accuracy and high driving risk.

[0004] Therefore, how to accurately detect the water risk level of the vehicle is a research hotspot. SUMMARY

[0005] The present application provides a water risk identification method, a water risk identification device and a vehicle, which can accurately detect the water risk level of the vehicle, and the technical solution is as follows: In a first aspect, a water risk identification method is provided, which is applied to a vehicle, and the method comprises: Obtaining target state information of the vehicle, the target state information comprising at least one of the following: water level information, image information, temperature information, humidity information, air pressure information, speed information, acceleration information; Determining water-related information corresponding to the vehicle based on the target state information, the water-related information comprising: a water probability distribution, a water risk score, and a water clustering result, the water probability distribution being determined by a first model, the first model being used to quantify input data into a probability density function, the water risk score being determined by a second model, the second model being used for fuzzy processing of input data; Determining a target water risk level corresponding to the vehicle based on the water-related information.

[0006] In this implementation, the wading-related information including the wading probability distribution, the wading risk score, and the wading clustering result is determined based on the acquired multi-dimensional target state information, so as to determine the target wading risk level of the vehicle based on the wading-related information. The multi-dimensional target state information can be used to determine multiple wading-related information, and the multi-source information fusion and multi-level risk assessment can be used to accurately detect the wading risk level of the vehicle, thereby significantly improving the comprehensiveness of the vehicle risk perception. In addition, the complex scenarios caused by the failure of a single sensor or incomplete data can be effectively dealt with, and the robustness and adaptability of the vehicle in extreme weather are enhanced, thereby providing accurate safety warning for the wading driving of the vehicle.

[0007] In combination with the first aspect, in some possible implementation manners, the determining of the target wading risk level of the vehicle based on the wading-related information comprises: acquiring a target weight coefficient corresponding to the wading-related information; performing correction processing on the wading-related information based on the target weight coefficient to obtain at least one wading weighted information; and determining the target wading risk level based on the wading weighted information.

[0008] In this implementation, the wading-related information is corrected based on the target weight coefficient corresponding to the wading-related information to obtain at least one wading weighted information, and the target wading risk level is determined based on the wading weighted information. This can improve the intervention degree of the key wading-related information in the current scene on the target wading risk level, and suppress the interference of secondary or noise information, thereby improving the accuracy and scene adaptability of the risk assessment.

[0009] In combination with the first aspect and the above implementation manners, in some possible implementation manners, the method further comprises: acquiring at least one historical wading risk level and at least one historical state information; and adjusting the target weight coefficient based on the historical wading risk level and the historical state information.

[0010] In this implementation, the target weight coefficient is adjusted based on the acquired historical wading risk level and historical state information, so that the target weight coefficient can be adaptively adjusted to be more consistent with the risk law in the actual road environment, and the target weight coefficient can be flexibly adjusted to adapt to the changeable driving environment, especially for the vehicle group that runs in a complex wading environment all year round.

[0011] In some possible implementation manners, the determining, based on the target state information, the water-related information corresponding to the vehicle comprises at least one of the following: determining, by the first model, the water probability distribution based on first state information, the first state information comprising at least one of the following: the water level information, the image information, the speed information, and the acceleration information; determining, by the second model, the water risk score based on second state information and / or the water probability distribution, the second state information comprising at least one of the following: the temperature information, the humidity information, and the air pressure information; and determining the water clustering result based on the water risk score.

[0012] In this implementation manner, the water-related information is flexibly determined based on multi-dimensional information by determining, by the first model, the water probability distribution based on first state information, determining, by the second model, the water risk score based on second state information and / or the water probability distribution, and determining the water clustering result based on the water risk score, and different water-related information can influence each other, thereby solving the problems of slow response and frequent false alarms of a traditional method based on a fixed threshold or a single judgment model in a complex water environment, and laying a solid data foundation for reliable, fine, and real-time water protection control.

[0013] In some possible implementation manners, the determining, based on the first state information, the water probability distribution comprises: determining a first feature vector corresponding to the first state information; and determining the water probability distribution based on the first feature vector.

[0014] In this implementation manner, the water probability distribution is determined based on a first feature vector corresponding to first state information by the first model, the accuracy of the water probability distribution is improved, the probability modeling of the water probability distribution is implemented, and the occurrence probability of different water depth intervals can be quantified to provide input information with uncertainty for other models.

[0015] In some possible implementation manners, the determining, based on the second state information and / or the water probability distribution, the water risk score comprises: performing fuzzy processing on the second state information and / or the water probability distribution to obtain at least one fuzzy language variable; processing the fuzzy language variable based on a fuzzy rule to obtain a fuzzy set; and determining the water risk score based on the fuzzy set.

[0016] In this implementation, the second state information and / or the wading probability distribution are processed based on the fuzzy rules by the second model to obtain fuzzy language variables, the fuzzy language variables are processed to obtain a fuzzy set, and the wading risk score is determined based on the fuzzy set, thereby improving the accuracy of the wading risk score, converting input information with uncertainty into language variables conforming to human cognition, and significantly improving the intuitiveness and interpretability of the wading level evaluation.

[0017] With reference to the first aspect and the above implementation, in some possible implementation, the method further includes: determining a target control instruction corresponding to the target wading risk level; and controlling the vehicle based on the target control instruction in a case where the vehicle meets a triggering condition.

[0018] In this implementation, the second feature vector corresponding to the wading risk score is clustered to obtain at least one cluster center feature, and the wading clustering result is determined based on the cluster center feature, which takes the cluster center feature as a typical representative of each risk category, improves the accuracy of the wading clustering result, and is more adaptive and self-adjusting than the traditional fixed threshold division method, and can more accurately reflect the current wading risk level of the vehicle.

[0019] With reference to the first aspect and the above implementation, in some possible implementation, the method further includes: determining a target control instruction corresponding to the target wading risk level; and controlling the vehicle based on the target control instruction in a case where the vehicle meets a triggering condition.

[0020] In this implementation, the target control instruction corresponding to the target wading risk level is determined, and the vehicle is controlled based on the target control instruction in a case where the vehicle meets a triggering condition, which can implement hierarchical and accurate intervention according to the target wading risk level, provide protection for the vehicle in the wading driving scenario, and improve the safety of the vehicle in a complex environment.

[0021] The second aspect provides a wading risk identification device, which is applied to a vehicle, and includes: The acquisition module is configured to acquire target state information of the vehicle, the target state information including at least one of water level information, image information, temperature information, humidity information, air pressure information, speed information, and acceleration information. determining module, configured to determine water-related information corresponding to the vehicle based on the target state information, the water-related information comprising a water probability distribution, a water risk score, and a water clustering result, the water probability distribution being determined by a first model, the first model being configured to quantize input data into a probability density function, the water risk score being determined by a second model, the second model being configured to perform fuzzy processing on input data; and determine a target water risk level corresponding to the vehicle based on the water-related information.

[0022] With reference to the second aspect, in some possible implementation manners, the obtaining module is configured to obtain a target weight coefficient corresponding to the water-related information; and the determining module is configured to perform correction processing on the water-related information based on the target weight coefficient to obtain at least one water weighting information, and determine the target water risk level based on the water weighting information.

[0023] With reference to the second aspect and the foregoing implementation manners, in some possible implementation manners, the obtaining module is configured to obtain at least one historical water risk level and at least one historical state information; and the apparatus further includes an adjusting module configured to adjust the target weight coefficient based on the historical water risk level and the historical state information.

[0024] With reference to the second aspect and the foregoing implementation manners, in some possible implementation manners, the determining module is configured to determine the water probability distribution based on first state information by using the first model, the first state information comprising at least one of the following: the water level information, the image information, the speed information, and the acceleration information; determine the water risk score based on second state information and / or the water probability distribution by using the second model, the second state information comprising at least one of the following: the temperature information, the humidity information, and the air pressure information; and determine the water clustering result based on the water risk score.

[0025] With reference to the second aspect and the foregoing implementation manners, in some possible implementation manners, the determining module is configured to determine a first feature vector corresponding to the first state information, and determine the water probability distribution based on the first feature vector.

[0026] With reference to the second aspect and the foregoing implementation manners, in some possible implementation manners, the determining module is configured to perform fuzzy processing on the second state information and / or the water probability distribution to obtain at least one fuzzy language variable, perform processing on the fuzzy language variable based on a fuzzy rule to obtain a fuzzy set, and determine the water risk score based on the fuzzy set.

[0027] With reference to the second aspect and the foregoing implementation manners, in some possible implementation manners, the determining module is configured to determine a second feature vector corresponding to the wading risk score; perform clustering processing on the second feature vector to obtain at least one cluster center feature; and determine the wading clustering result based on the cluster center feature.

[0028] With reference to the second aspect and the foregoing implementation manners, in some possible implementation manners, the determining module is configured to determine a target control instruction corresponding to the target wading risk level; and the apparatus further includes a control module configured to control the vehicle based on the target control instruction in a case where the vehicle meets a triggering condition.

[0029] In a third aspect, a vehicle is provided, which includes one or more processors and one or more memories, the one or more memories storing at least one program code, the program code being loaded and executed by the one or more processors to implement operations performed by the wading risk identification method.

[0030] In a fourth aspect, a computer-readable storage medium is provided, which stores at least one program code, the program code being loaded and executed by a processor to implement operations performed by the wading risk identification method.

[0031] By using the technical solutions provided in the embodiments of the present application, the wading-related information including the wading probability distribution determined by the first model, the wading risk score determined by the second model, and the wading clustering result is determined based on the obtained multi-dimensional target state information, the target wading risk level of the vehicle is determined based on the wading-related information, the multi-dimensional target state information is used to determine multiple wading-related information, the multi-source information fusion and multi-level risk assessment are used to accurately detect the wading risk level of the vehicle, the comprehensiveness of the vehicle risk perception is significantly improved, meanwhile, the complex scenarios of single sensor failure or incomplete data can be effectively dealt with, and the robustness and adaptability of the vehicle in extreme weather are enhanced, thereby providing accurate safety warning for wading driving of the vehicle. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 is a flowchart of a method for determining a wading risk level provided in the embodiments of the present application; Figure 2 is a flowchart of a wading risk identification method provided in the embodiments of the present application; Figure 3 is a flowchart of another wading risk identification method provided in the embodiments of the present application; Figure 4 is a structural schematic diagram of a wading risk identification apparatus provided in the embodiments of the present application; Figure 5 is a structural schematic diagram of a vehicle provided by an embodiment of the present application. DETAILED DESCRIPTION

[0033] The technical solutions in the present application will be described clearly and exhaustively below with reference to the drawings. In the description of the embodiments of the present application, unless otherwise specified, " / " represents the meaning of or, for example, A / B can represent A or B: "and / or" in the text only describes the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B, which means that there are three cases of A alone, A and B together, and B alone. In addition, in the description of the embodiments of the present application, "multiple" means two or more than two.

[0034] Hereinafter, the terms "first" and "second" are only used for descriptive purposes, and cannot be understood as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined with "first" and "second" can explicitly or implicitly include one or more features.

[0035] In order to describe the technical solutions provided by the embodiments of the present application, some terms related to the embodiments of the present application will be described below.

[0036] Vehicle wading: vehicle wading refers to the behavior of a motor vehicle passing through a low-lying water section, which often occurs in the rainy season or in areas with insufficient drainage facilities, and there is a risk of engine water intake, circuit short circuit, etc.

[0037] Gaussian Mixture Model (Gaussian Mixture Model, GMM): a soft clustering algorithm based on probability, suitable for modeling complex data. It assumes that the data distribution is a weighted sum of multiple Gaussian distributions, and iteratively optimizes the parameters through the expectation maximization algorithm, and finally obtains the probability of each sample point belonging to each cluster.

[0038] Fuzzy Neural Networks (Fuzzy Neural Networks, FNN): FNN combines fuzzy systems and neural networks, and its essence is to process the input of neural networks through fuzzy systems to become fuzzy input signals and fuzzy weights, and to anti-fuzz the output of neural networks, which is called intuitive effective value. Specifically, in the fuzzy neural network, the input and output of the neural network represent the input and output of the fuzzy system, and the membership function and fuzzy rule of the fuzzy system are added to the hidden nodes of the neural network, which fully utilizes the parallel processing capability of the neural network and the reasoning capability of the fuzzy system.

[0039] Clustering analysis: a technique for finding the underlying structure of data, which can organize all data instances into similar groups, and these similar groups are called clusters, data instances in the same cluster are the same, and instances in different clusters are different.

[0040] With the development of vehicle technology, more and more vehicles are equipped with intelligent systems to cope with various complex road environments.

[0041] In the related art, the intelligent system of the vehicle mainly relies on sensors such as cameras, radars and lidars to realize lane line recognition, obstacle detection, adaptive cruise and automatic emergency braking functions. However, in the scene of the water section, due to the unpredictability of the road, high risk level and dynamic change, etc., the current water risk level cannot be accurately detected, and there are problems of low accuracy and high driving risk.

[0042] Therefore, how to accurately detect the water risk level of the vehicle is a research hotspot.

[0043] Figure 1 is a flowchart for determining the water risk level provided by the embodiments of the present application.

[0044] Exemplarily, first, S11 is executed to collect target state information. Then, S12 is executed to determine a water probability distribution based on first state information through a first model, further S13 is executed to determine a water risk score based on second state information and the water probability distribution through a second model, thereby S14 is executed to determine a water clustering result based on the water risk score, and finally S15 is executed to determine a target water risk level corresponding to the vehicle based on the water probability distribution, the water risk score and the water clustering result.

[0045] Among them, the target state information includes at least one of the following: water level information, image information, temperature information, humidity information, air pressure information, speed information, acceleration information. The first model is used to quantify the input data into a probability density function. The first state information includes at least one of the following: water level information, image information, speed information and acceleration information. The second model is used for fuzzy processing of the input data. The second state information includes at least one of the following: temperature information, humidity information and air pressure information.

[0046] In the embodiments of the present application, the target state information of the vehicle is obtained, and the water-related information including the water probability distribution determined by the first model, the water risk score determined by the second model, and the water clustering result is determined based on the water-related information to determine the target water risk level of the vehicle. The water-related information can be determined based on the multi-dimensional target state information, and the water risk level of the vehicle can be accurately detected based on the multi-source information fusion and multi-level risk assessment. The comprehensive risk perception of the vehicle is improved, and the robustness and adaptability of the vehicle in extreme weather conditions are improved. The precise safety warning for the water driving of the vehicle is provided.

[0047] The application scenarios of the technical solutions provided in the embodiments of the present application are introduced as follows. The technical solutions provided in the embodiments of the present application can be applied to different types of vehicles, for example, hybrid vehicles, electric vehicles, pure fuel vehicles, and the like. Of course, with the development of science and technology, other types of vehicles can also appear. The technical solutions provided in the embodiments of the present application are also applicable to other types of vehicles.

[0048] After introducing the application scenarios of the embodiments of the present application, the technical solutions provided in the embodiments of the present application are introduced as follows. Figure 2 Taking the vehicle as an example, the method includes the following steps.

[0049] 201, obtaining target state information of the vehicle, the target state information including at least one of the following: water level information, image information, temperature information, humidity information, air pressure information, speed information, acceleration information.

[0050] The target state information is information used to represent the current state of the vehicle. In some embodiments, the target state information can include but is not limited to at least one of the following: water level information, image information, temperature information, humidity information, air pressure information, speed information, acceleration information, and the like. The water level information is used to represent the depth of water in the environment where the vehicle is located. The image information is used to represent the visual data of the road in front of the vehicle or the surrounding environment. The temperature information is used to represent the temperature of the battery of the vehicle. The humidity information is used to represent the humidity inside the battery compartment of the vehicle. The air pressure information is used to represent the air pressure inside the battery compartment of the vehicle. The speed information is used to represent the instantaneous driving speed of the vehicle relative to the ground. The acceleration information is used to represent the rate of change of the vehicle in the driving direction and perpendicular to the driving direction. In some embodiments, different target state information can be obtained by different types of sensors or data acquisition devices, and the target state information can be processed synchronously according to the acquisition time of the target state information.

[0051] 202、determine the water-related information corresponding to the vehicle based on the target state information, the water-related information including a water probability distribution, a water risk score, and a water clustering result, the water probability distribution being determined by a first model, the first model being configured to quantize input data into a probability density function, and the water risk score being determined by a second model, the second model being configured to perform fuzzy processing on the input data.

[0052] The water-related information is information used to represent the current water level of the vehicle. In some embodiments, the water level of the vehicle includes, but is not limited to, no water, light water, moderate water, heavy water, etc. The water-related information can include, but is not limited to, at least one of the following: water probability distribution, water risk score, water clustering result, etc. The water probability distribution is used to represent the probability distribution of different water levels of the vehicle. In some embodiments, the water probability distribution is determined by a first model, and the first model is configured to quantize input data into a probability density function. The first model can include, but is not limited to, at least one of the following: Gaussian mixture model, Bayesian probability model, etc. The water risk score is a continuous numerical indicator used to represent different water levels of the vehicle. In some embodiments, the water risk score is determined by a second model, and the second model is configured to perform fuzzy processing on the input data. The second model can include, but is not limited to, at least one of the following: fuzzy neural network, fuzzy reasoning system, fuzzy decision tree, etc. The water clustering result is used to represent the clustering situation of different water levels of the vehicle. Different water-related information is determined by selecting different information from the target state information.

[0053] 203、determine the target water risk level corresponding to the vehicle based on the water-related information.

[0054] The target water risk level refers to the current water risk level of the vehicle. In some embodiments, the water risk level can include, but is not limited to, no water risk, low water risk, medium water risk, high water risk, etc. Different water risk levels represent different water levels of the vehicle. The target water risk level is one of the water risk levels. The target water risk level can include, but is not limited to, one of the following: no water risk, low water risk, medium water risk, high water risk, etc. In some embodiments, there is a corresponding relationship between the water level of the vehicle and the water risk level, for example, no water corresponds to no water risk, light water corresponds to low water risk, moderate water corresponds to medium water risk, and heavy water corresponds to high water risk.

[0055] By means of the technical solutions provided in the embodiments of the present application, the target water-related information including the water contact probability distribution determined by the first model, the water contact risk score determined by the second model, and the water contact clustering result is determined based on the acquired multi-dimensional target state information, the target water contact risk level corresponding to the vehicle is determined based on the water-related information, the multi-dimensional target state information can be used to determine various water-related information, the water contact risk level of the vehicle can be accurately detected through multi-source information fusion and multi-level risk assessment, the comprehensiveness of vehicle risk perception is significantly improved, meanwhile, the complex scenarios of single sensor failure or incomplete data can be effectively coped with, the robustness and adaptability of the vehicle in extreme weather are enhanced, and accurate safety warning is provided for water contact driving of the vehicle.

[0056] It should be noted that the steps 201-203 are a simple description of the water contact risk identification method provided in the embodiments of the present application, and the water contact risk identification method provided in the embodiments of the present application will be described in more detail below in combination with some examples, see Figure 3 Taking the execution subject as a vehicle for example, the method comprises the following steps.

[0057] 301、acquire target state information of the vehicle, the target state information comprising at least one of the following: water level information, image information, temperature information, humidity information, air pressure information, speed information, acceleration information.

[0058] The target state information is information used to represent the current state of the vehicle. In some embodiments, the target state information is acquired from sensors or data acquisition devices of the vehicle in the case where the target water contact risk level of the vehicle needs to be detected. In some embodiments, the target state information can include but is not limited to at least one of the following: water level information, image information, temperature information, humidity information, air pressure information, speed information, acceleration information, etc. The water level information is used to represent the depth of accumulated water in the environment where the vehicle is located. The image information is used to represent the visual data of the road ahead of the vehicle or the surrounding environment. The temperature information is used to represent the temperature of the battery of the vehicle. The humidity information is used to represent the humidity inside the battery compartment of the vehicle. The air pressure information is used to represent the air pressure inside the battery compartment of the vehicle. The speed information is used to represent the instantaneous driving speed of the vehicle relative to the ground. The acceleration information is used to represent the rate of change of motion of the vehicle in the driving direction and perpendicular to the driving direction.

[0059] In some embodiments, different target state information can be acquired by different types of sensors or data acquisition devices. Different types of sensors are deployed at multiple key positions of the vehicle, for example, outside, inside, and regions directly interacting with the environment of the battery pack compartment, etc.

[0060] In some embodiments, for water level information, three water level sensors are arranged on the front, middle and rear of the chassis of the vehicle, respectively, to monitor the front water depth, middle water depth and rear water depth during dynamic driving of the vehicle. Further, the three water level measuring points cooperatively constitute a spatial distribution map of the current water situation of the vehicle, and the spatial distribution map of the current water situation of the vehicle is taken as the water level information, which can capture the complex water depth changes on the non-flat water road surface. The selection of the water level sensor meets the vehicle-grade waterproof and shockproof standard, and the precision can reach millimeter level, which can maintain stable reading when passing through the water pit and bump terrain at high speed.

[0061] In some embodiments, for temperature information and humidity information, a temperature and humidity integrated sensor is arranged inside the battery pack cabin of the vehicle to sense the sealed environment state inside the battery in real time, and then obtain the temperature information and humidity information. The increase of cabin humidity is a significant manifestation of early water ingress, and the change of temperature can also reflect the abnormality of environmental heat exchange, so the temperature information and humidity information are one of the key indicators for predicting water ingress of the battery pack. For air pressure information, a pressure sensor is arranged on the vehicle to monitor the small changes of cabin air pressure, and then obtain the air pressure information.

[0062] In some embodiments, the image data of the front-view camera of the whole vehicle is obtained as image information, which can be used to perceive the visual information of the water environment, such as water surface texture reflection, mud concentration change, etc. At the same time, considering the dynamic blur problem of the vehicle-mounted camera under high-speed motion and water reflection, a dynamic region of interest (RoI) is set, and only the intersection region of the image data with the road surface in front of the vehicle head is taken as the image information, to avoid the interference of invalid information such as sky and water surface reflection in the judgment.

[0063] In some embodiments, the data sampling period of the sensor is set to any suitable time length, for example, 100 milliseconds, 200 milliseconds, etc. At the same time, since various sensors are distributed on different electronic control units (ECU) and buses, there may be time sequence offset, therefore, a unified timestamp mechanism is adopted for data synchronization, and the Controller Area Network (CAN) and Ethernet time synchronization protocol are used for timestamp alignment, to ensure that the target state information can be fused and processed in a unified time dimension.

[0064] In some embodiments, after obtaining the target state information, preprocessing is required to filter out noise, repair missing values, and eliminate obviously abnormal data points. For example, water level information, temperature information, and humidity information may have signal mutation problems in the fast wading state. A sliding window filtering algorithm can be used to smooth them, with a window length of 1 second, which can effectively suppress the instantaneous noise signal caused by the tire impacting the water surface. At the same time, if the humidity inside the cabin suddenly rises by more than a certain threshold for three consecutive frames, it is identified as hardware interference or false triggering, and this section of data is repaired by linear interpolation to ensure the smooth continuity of the temperature information and humidity information.

[0065] 302、based on the target state information, determine the wading related information corresponding to the vehicle, the wading related information including: wading probability distribution, wading risk score, wading clustering result, the wading probability distribution being determined by a first model, the first model being used for quantifying input data into a probability density function, the wading risk score being determined by a second model, the second model being used for fuzzy processing of input data.

[0066] The wading related information is information for representing the current wading degree of the vehicle. In some embodiments, the wading degree of the vehicle includes but is not limited to: no wading, light wading, moderate wading, severe wading, etc. The wading related information can include but is not limited to at least one of the following: wading probability distribution, wading risk score, wading clustering result, etc. The wading probability distribution is used to represent the probability distribution of different wading degrees of the vehicle. In some embodiments, the wading probability distribution is determined by a first model, and the first model is used to quantify the input data into a probability density function. The first model can include but is not limited to at least one of the following: Gaussian mixture model, Bayesian probability model, etc. The wading risk score is a continuous numerical index for representing different wading degrees of the vehicle. In some embodiments, the wading risk score is determined by a second model, and the second model is used for fuzzy processing of input data. The second model can include but is not limited to at least one of the following: fuzzy neural network, fuzzy reasoning system, fuzzy decision tree, etc. The wading clustering result is used to represent the clustering situation of different wading degrees of the vehicle. Different wading related information is determined by selecting different information from the target state information.

[0067] In one possible implementation, by the first model, the wading probability distribution is determined based on the first state information, and the first state information includes at least one of the following: water level information, image information, speed information, and acceleration information. By the second model, the wading risk score is determined based on the second state information and / or the wading probability distribution, and the second state information includes at least one of the following: temperature information, humidity information, and air pressure information. Based on the wading risk score, the wading clustering result is determined.

[0068] To make the above-mentioned embodiments clearer, the process of determining the water-related information corresponding to the vehicle in the above-mentioned embodiments is described in three parts as follows.

[0069] The first part is to determine a water probability distribution based on first state information by a first model.

[0070] The first state information can include but is not limited to at least one of the following: water level information, image information, speed information, acceleration information, etc. The first model can include but is not limited to at least one of the following: Gaussian mixture model, Bayesian probability model, etc. The speed information can be obtained from a normal wheel speed sensor or an inertial measurement unit. The acceleration information can be obtained from an accelerometer of the vehicle. The water probability distribution is a probability distribution for characterizing different degrees of vehicle water immersion. In some embodiments, the water probability distribution is a continuous distribution, which can characterize the probability distribution of different degrees of vehicle water immersion. The fusion and reasoning of the first state information by the first model can determine the water probability distribution.

[0071] In one possible implementation, a first feature vector corresponding to the first state information is determined. Based on the first feature vector, the water probability distribution is determined.

[0072] The first feature vector is a vector obtained by combining the first state information. In some embodiments, the first feature vector is a multi-dimensional feature vector. It should be noted that the first state information can comprehensively reflect the dynamic interaction between the vehicle and the water environment, wherein the water level information is used to characterize the spatial characteristics of the water depth distribution around the vehicle, the speed information and the acceleration information reflect the motion response characteristics of the vehicle in the water environment, and the image information is used to characterize the visual information around the vehicle. Therefore, the first feature vector obtained by combining the first state information can contain multi-dimensional information of the current environment.

[0073] In some embodiments, in the case where the first model is a Gaussian mixture model, the water probability distribution is determined based on the first state information by the Gaussian mixture model. The Gaussian mixture model is a typical unsupervised probability modeling algorithm, which is suitable for classifying modeling of complex continuous variables and has the natural advantage of soft classification of environmental state, and can provide high-credibility pre-input for fuzzy neural networks. For example, when the vehicle enters an area with medium water depth and low speed, the system may determine that the probability of moderate water immersion is the highest, while the probabilities of light water immersion, heavy water immersion and non-water immersion are relatively low. This multi-level, non-binary soft discrimination method has higher precision and fault tolerance than traditional threshold judgment, and performs more stably in continuous working condition transition state.

[0074] It should be noted that, in order to ensure the generalization ability and online calculation efficiency of the Gaussian mixture model, a multinomial model optimization mechanism is introduced. In the training process of the Gaussian mixture model, the optimal number of Gaussian components is selected by using cross-validation method to avoid overfitting or underfitting, and at the same time, the complexity of the Gaussian mixture model is evaluated and screened according to the Akaike information criterion and the Bayesian information criterion. At the same time, a large amount of historical test and actual measurement wading data are used to construct the Gaussian mixture model. Among them, the data come from actual road test, including but not limited to: urban waterlogging road, cross-country river crossing scene, high-speed water passing and other typical working conditions, covering various conditions from dry and waterless to deep water strong impact. Each data sequence is automatically labeled as a specific wading level by the system, and the label distribution is used to guide the setting of the number of components of the Gaussian mixture model. And the expectation maximization algorithm is used to optimize the estimation of the mean, covariance and weight parameters of the mixed components, and finally a plurality of Gaussian sub-distributions are formed, each sub-distribution representing a probability density of the vehicle wading degree.

[0075] At the same time, in the reasoning stage, the Gaussian mixture model is deployed in the central gateway of the whole vehicle with a lightweight running version, only the main parameters are reserved and the lookup table method is used to accelerate the calculation, to ensure that the probability output is completed within milliseconds, meeting the real-time control requirements in high dynamic scenes.

[0076] In some embodiments, in the long-term operation of the vehicle, the sensor response data in the actual wading working condition is continuously collected, and the model parameters of the Gaussian mixture model are adjusted regularly to adapt to the changes in data distribution caused by sensor aging, installation deviation or user driving habit difference. Through the over-the-air (OTA) technology, the incremental model weight update package is periodically issued, and the adaptive adjustment of the Gaussian mixture model can be completed without user intervention.

[0077] In this implementation, by the first model, the wading probability distribution is determined based on the first feature vector corresponding to the first state information, which improves the accuracy of the wading probability distribution and realizes the probability modeling of the wading probability distribution, and can quantify the occurrence probability of different water depth intervals, providing input information with uncertainty for other models.

[0078] The second part, by the second model, determines the wading risk score based on the second state information and / or the wading probability distribution.

[0079] The second state information can include, but is not limited to, at least one of the following: temperature information, humidity information, air pressure information, and the like. The second model can include, but is not limited to, at least one of the following: a fuzzy neural network, a fuzzy inference system, a fuzzy decision tree, and the like. The wading probability distribution is a probability distribution used to represent different degrees of vehicle wading. The wading risk score is a continuous numerical indicator used to represent different degrees of vehicle wading. In some embodiments, the wading risk score ranges from zero to one, indicating the strength of the degree of vehicle wading. The wading risk score can be determined by the second state information, or by the second state information and the wading probability distribution, or by the wading probability distribution. In some embodiments, based on the second state information and / or the wading probability distribution, the second model is used to quantitatively and fusionally calculate the second state information and / or the wading probability distribution, and finally output a continuous wading risk score to determine the wading risk score.

[0080] In a possible implementation, the second state information and / or the wading probability distribution are subjected to fuzzy processing to obtain at least one fuzzy linguistic variable. The fuzzy linguistic variable is processed based on fuzzy rules to obtain a fuzzy set. Based on the fuzzy set, the wading risk score is determined.

[0081] The fuzzy linguistic variable is data obtained by fuzzy processing of the second state information and / or the wading probability distribution. Fuzzy processing can effectively process uncertain and fuzzy data, thereby improving data quality and reliability. Fuzzy rules are a form of reasoning based on fuzzy logic, and their essence is a binary fuzzy relation defined on two universes. The fuzzy set is a set used to express the concept of fuzziness.

[0082] In some embodiments, when the second model is a fuzzy neural network, the second state information and / or the wading probability distribution are subjected to fuzzy processing by the fuzzy neural network, and the fuzzy linguistic variable obtained is processed based on fuzzy rules to obtain a fuzzy set, and then the wading risk score is determined based on the fuzzy set. The fuzzy neural network combines the rule explainability of the fuzzy logic system and the parameter adaptive ability of the neural network, and can process nonlinear, multivariate, and uncertain sensor data inputs, and output continuous wading risk ratings.

[0083] In some embodiments, the structure of the fuzzy neural network is divided into an input layer, a fuzzification layer, a rule layer, a normalization layer, and an output layer. The input layer is responsible for receiving input values, the fuzzification layer performs fuzzy quantization processing on each input value according to a set membership function to obtain fuzzy linguistic variables, the rule layer performs reasoning on the fuzzy linguistic variables according to predefined fuzzy rules to obtain multi-rule outputs, the normalization layer performs weighted normalization processing on the multi-rule outputs, and the output layer generates a continuous risk score value, which is the wading risk score.

[0084] In some embodiments, the number of fuzzy rules is at least one, covering the correspondence between various combinations of input conditions and output risk levels. For example, the fuzzy rules include complex logical combinations such as outputting severe wading under the condition that the water level information represents a high probability of water depth and the humidity information represents high humidity in the cabin, or outputting moderate wading in the case of water level information representing a low probability of water depth but air pressure information representing a sharp pressure fluctuation.

[0085] In some embodiments, all fuzzy rules and membership function parameters can be iteratively optimized through data training, with certain self-learning capabilities. When training the fuzzy neural network, a set of measured wading scene data is constructed internally by the vehicle manufacturer. The data sources can include but are not limited to: closed test site wading experiments, urban road rainwater accumulation section driving tests, and real vehicle data under typical off-road wading conditions, etc. Each set of data contains complete input and manually labeled wading risk level labels. The training process uses the error back propagation algorithm to optimize the membership function center and width parameters in the network, minimizing the error between the output risk score and the label. Early stopping mechanism and regularization methods are introduced in the training process to prevent model overfitting, while data augmentation techniques are used to expand the diversity and boundary distribution of input samples, improving the fuzzy neural network's ability to recognize extreme wading conditions.

[0086] In this implementation, the fuzzy language variable obtained by fuzzy processing the second state information and / or the wading probability distribution based on the fuzzy rules is processed by the second model to obtain a fuzzy set, so as to determine the wading risk score based on the fuzzy set, thereby improving the accuracy of the wading risk score, converting the input information with uncertainty expression into a language variable consistent with human cognition, and significantly improving the intuitiveness and interpretability of the wading level evaluation.

[0087] The third part, based on the wading risk score, determines the wading clustering result.

[0088] The wading risk score is a continuous numerical index used to represent different levels of vehicle wading. The wading clustering result is used to represent the clustering of different levels of vehicle wading. In some embodiments, the wading clustering result is determined by clustering the wading risk score. The wading clustering result includes but is not limited to at least one of the following: no wading, light wading, moderate wading, severe wading, etc.

[0089] In this implementation, the water involvement probability distribution is determined by the first state information, the water involvement risk score is determined by the second state information and / or the water involvement probability distribution, and the water involvement clustering result is determined by the water involvement risk score. The water involvement related information can be flexibly determined by multi-dimensional information, and different water involvement related information can influence each other, solving the problem of slow response and frequent false alarms of traditional fixed threshold or single judgment model in complex water involvement environment, and laying a solid data foundation for reliable, fine and real-time water involvement prevention control.

[0090] In a possible implementation, a second feature vector corresponding to the water involvement risk score is determined. The second feature vector is subjected to clustering processing to obtain at least one clustering center feature. Based on the clustering center feature, a water involvement clustering result is determined.

[0091] The second feature vector is a vector converted from the water involvement risk score. In some embodiments, the number of second feature vectors is at least one. At least one representative point is selected from the continuous water involvement risk scores, the weight distributed on the representative point is calculated, and then the weight on the representative point is arranged to obtain the second feature vector. Clustering processing is a statistical learning method, which can classify data objects with similar features into the same class. The clustering center feature is obtained by clustering processing. In some embodiments, the number of clustering center features is at least one, and the number of clustering center features is the same as the number of vehicle water involvement degrees.

[0092] In some embodiments, the method of clustering processing the second feature vector can include but is not limited to: clustering processing the second feature vector by K-Means clustering algorithm, clustering processing the second feature vector by mean shift clustering algorithm, etc. For example, K-Means clustering can set at least one category, and divide each second feature vector into the category represented by the nearest cluster center through iteration to obtain at least one clustering center feature. For another example, mean shift clustering can determine a window radius, and start sliding from a randomly selected center point. Each time the window slides to a new area, the mean value in the window is calculated as the center point. When multiple sliding windows overlap, the sliding is stopped and at least one center point is obtained. Finally, each second feature vector is divided based on the center point to obtain at least one clustering center feature.

[0093] In some embodiments, a hierarchical clustering analysis method is used to divide the risk level of the second feature vector. The similarity distance between the second feature vectors is calculated, and the second feature vectors are gradually aggregated into multiple categories to form a clustering tree structure. In the aggregation process, the minimum variance is selected as the criterion, and the score value is classified step by step. The final clustering result is divided into four categories, representing no water involvement, light water involvement, moderate water involvement, and severe water involvement, respectively.

[0094] It should be noted that, in order to enhance the stability of the clustering boundary, edge smoothing processing is performed on the clustering result, the second feature vectors close to the critical value are merged or corrected, and the problem of frequent jump of control caused by score fluctuation is avoided. At the same time, a unique number is assigned to each clustering result, and a mapping table is established, and all subsequent modules only use discrete level numbers for control strategy index and trigger logic execution, so as to improve the response consistency and decision controllability of the system in the actual running process.

[0095] In this embodiment, by clustering the second feature vectors corresponding to the wading risk scores, at least one cluster center feature is obtained, and then the wading clustering result is determined based on the cluster center feature. The cluster center feature is taken as the typical representative of each risk category, which improves the accuracy of the wading clustering result. Compared with the traditional fixed threshold division method, it has better adaptability and self-adjusting ability, and can more accurately reflect the current wading risk level of the vehicle.

[0096] 303、Obtain the target weight coefficient corresponding to the wading-related information.

[0097] The target weight coefficient is a parameter for representing the importance of the wading-related information. In some embodiments, the target weight coefficient can be any suitable size, for example, 0.3, 0.5, etc. The sizes of the target weight coefficients corresponding to different wading-related information can be the same or different. The greater the target weight coefficient corresponding to the wading-related information, the greater the influence of the wading-related information on the target wading risk level.

[0098] In a possible implementation, at least one historical wading risk level and at least one historical state information are obtained; and the target weight coefficient is adjusted based on the historical wading risk level and the historical state information.

[0099] The historical wading risk level is a historical output wading risk level of the vehicle. The historical state information is historical state information collected by the vehicle. In some embodiments, the historical wading risk level and the historical state information can be obtained from other vehicles of the same type as the current vehicle to improve the amount of data.

[0100] In some embodiments, the construction parameter adjustment module adjusts the target weight coefficient based on the obtained at least one historical wading risk level and at least one historical state information. The input of the parameter adjustment module can include but is not limited to: historical wading risk level, historical state information, control quality instruction execution record, user's region geographical environment label, vehicle running time, vehicle fault code data, system exception reporting record and historical wading log file, etc. The parameter adjustment module preliminarily caches and classifies the input data through the local operation and control domain controller, uploads to the remote server or the local data evaluation unit according to the period, and performs strategy deviation analysis and parameter correction by the exclusive model performance monitoring module.

[0101] In some embodiments, the output of the parameter adjustment module includes two dimensions of results: first, the target weight coefficient is adjusted to obtain an updated target weight coefficient, which can directly act on the input correction logic of the current risk level modeling, realize the dynamic adjustment ability in the short term, and improve the adaptation degree of the model under the current environment. Second, a strategy optimization package or a parameter template that can be issued is provided for the remote OTA parameter scheduling platform, which is used to issue to the user group of the same type of vehicle and the same type of environment after centralized optimization, forming a model strategy distribution ability across vehicles and regions.

[0102] In some embodiments, the parameter adjustment module internal structure includes three parts, namely the judgment deviation detection module, the parameter generation module, and the OTA adaptation module. The judgment deviation detection module is used to compare the effectiveness consistency difference between the model judgment output and the control response in the actual vehicle water crossing scene. For example, if the phenomenon of frequent over-sensitive triggering, delayed response, or execution failure frequently occurs in multiple water crossing scenes, the judgment deviation detection module will identify the strategy deviation and enter the parameter adjustment channel. The judgment deviation detection module runs locally and has real-time response and short-term adjustment capability. The parameter generation module will generate parameter correction suggestions for the current vehicle according to the judgment deviation performance, combined with regional environmental factors and state information, including but not limited to: grade threshold fine-tuning, sensor weight adjustment, strategy trigger boundary buffer time, and other dimensions. The parameter generation module is configured with a lightweight adjustment network based on historical data learning and construction, which can generate weight disturbance suggestions according to the performance of the model under different input combinations, and automatically integrate the original strategy to form an intermediate version. All generated parameters complete safety verification and risk simulation evaluation locally to ensure that parameter updates do not cause system control conflicts. The OTA adaptation module is responsible for converting the locally evaluated parameter adjustment suggestions into a parameter package format that meets the version control and remote delivery requirements, and matching the corresponding parameter model according to the vehicle's partition, hardware version, sensor layout, etc. to avoid strategy conflicts caused by vehicle configuration differences. This module also supports obtaining the latest optimized parameters from the vehicle manufacturer's server and completing the overlay deployment or incremental correction, and completing online deployment through the OTA scheduling mechanism. The deployment process uses security signature verification, version comparison verification, dynamic hot loading, and other methods to ensure that the update process does not interrupt the current control flow or cause runtime errors.

[0103] In this implementation, by adjusting the target weight coefficient based on the obtained historical water crossing risk level and historical state information, the adaptive adjustment of the target weight coefficient can be realized, so that the target weight coefficient is more in line with the risk law in the actual road environment, and can flexibly respond to the changing driving environment, especially suitable for vehicle groups that run in complex water crossing environments all year round.

[0104] 304、Based on the target weight coefficient, the water crossing related information is corrected and processed to obtain at least one water crossing weighted information.

[0105] The target weight coefficient is a parameter used to represent the importance of the water crossing related information. In some embodiments, the target weight coefficient can be any suitable size, such as 0.3, 0.5, etc. The sizes of the target weight coefficients corresponding to different water crossing related information can be the same or different. The water crossing weighted information is information obtained by correcting and processing the water crossing related information. In some embodiments, the number of water crossing weighted information is the same as the number of water crossing related information.

[0106] 305、determine a target wading risk level based on the wading-related information.

[0107] The target wading risk level refers to the current wading risk level of the vehicle. In some embodiments, the wading risk level can include but is not limited to: no wading risk, low wading risk, medium wading risk, high wading risk, etc. Different wading risk levels represent different degrees of wading of the vehicle. The target wading risk level is one of the wading risk levels. The target wading risk level can include but is not limited to one of the following: no wading risk, low wading risk, medium wading risk, high wading risk, etc.

[0108] In some embodiments, the output consistency scheduling module is constructed to modify the wading-related information based on the target weight coefficient to obtain at least one wading-weighted information, and further determine the target wading risk level based on the wading-weighted information. At the same time, the output consistency scheduling module can realize signal standardization processing, convert data of different formats and different update periods (including but not limited to wading-related information, control instructions, etc.) into a unified structure of state vector, and perform timestamp alignment and synchronous delay compensation, to ensure that the fusion logic makes judgments in the latest state.

[0109] In some embodiments, the internal structure of the output consistency scheduling module consists of three parts, namely the model interface module, the state scheduling module, and the arbitration fusion module. The model interface module acts as a buffer between the models and the fusion mechanism, responsible for receiving, caching, and converting data from each model, while recording the model running state and the last output timestamp for the state scheduling module to determine whether to enter the fusion process. The state scheduling module dynamically monitors the update frequency, change amplitude, and model health state of all incoming signals, identifies whether there are any abnormal phenomena such as long-time no output, abnormal fluctuations, or signal conflicts, and can suspend a model access, reset the model cache, or switch to a backup model based on this. The arbitration fusion module operates based on a multi-model output consistency strategy for dynamic fusion.

[0110] It should be noted that the output consistency scheduling module is deployed in the high-reliability real-time computing core of the vehicle central gateway, with a running period of no more than fifty milliseconds, sufficient computing power and resource isolation mechanism, to ensure that the model output consistency is maintained in complex dynamic environments such as high-speed wading, impact and jolt, supporting the stable operation of the control system and building a multi-algorithm collaborative infrastructure for complex scenarios.

[0111] Meanwhile, the vehicle wading degree corresponding to the vehicle wading risk level given by the majority model is directly used as the target vehicle wading risk level when the vehicle wading probability distribution, the vehicle wading risk score, and the vehicle wading clustering result are consistent with each other. When there is a slight disagreement, the wading-related information is weighted and fused according to the target weight coefficient. When there is a serious conflict or a large difference in signals, the arbitration state is automatically entered, the control output is temporarily frozen, the delay determination mechanism is started, and the next cycle model is waited for to output again, so as to avoid that the abnormal data directly triggers the key control command. In addition, different processing parts have hot plug capability, and when a model fails or needs to be offline due to limited hardware resources, the interface module automatically exits the fusion link, and the system can rely on the remaining models to complete the determination, so as to ensure that the control chain is complete and uninterrupted. The target vehicle wading risk level can be provided with a consistency identifier and a fusion confidence score, which are important basis for subsequent safety strategies and log analysis. The target vehicle wading risk level and the actual wading feedback result are compared and analyzed, which are closed-loop basis for model evaluation and optimization, and the long-term operation accuracy of the system is improved.

[0112] In this implementation, the target wading-related information is corrected by the target weight coefficient corresponding to the obtained wading-related information, at least one wading-weighted information is obtained, and the intervention degree of the key wading-related information in the current scene on the target wading risk level is improved based on the wading-weighted information, while the interference of secondary or noise information is suppressed, so that the accuracy and scene adaptability of risk assessment are improved.

[0113] In a possible implementation, a target control instruction corresponding to the target wading risk level is determined. In a case where the vehicle meets a trigger condition, the vehicle is controlled based on the target control instruction.

[0114] The target control instruction is an instruction corresponding to the target wading risk level. The trigger condition is data used to determine whether to execute the target control instruction. In some embodiments, the trigger condition can be that the vehicle speed is lower than a threshold value, the target wading risk level is higher than a set value, and the battery temperature is normal. The target wading risk level can also be one of a low wading risk, a medium wading risk, and a high wading risk.

[0115] In some embodiments, the target wading risk level is queried in a first relationship table to obtain the target control instruction.

[0116] The first relationship table stores a plurality of target wading risk levels and target control instructions corresponding to each target wading risk level. The target control instruction corresponding to the target wading risk level can be obtained by querying the target wading risk level in the first relationship table. The first relationship table is calibrated by a technician according to actual conditions, and embodiments of the present application are not limited in this regard.

[0117] In some embodiments, a whole vehicle control strategy module is constructed to determine a target control instruction corresponding to the target wading risk level, and to control the vehicle based on the target control instruction in the case where the vehicle meets the triggering condition. The input of the whole vehicle control strategy module is the target wading risk level, and it can also receive key state information from a vehicle speed sensor, a driving motor torque state, a battery pack temperature sensor, and a whole vehicle network feedback, for execution state confirmation, execution condition judgment, and feedback closed-loop verification of the target control instruction. The target control instruction is a standardized sealed execution instruction signal, which includes a sealed valve action trigger, an electromagnetic pump opening, a cabin pressure regulation port closing, and other specific execution actions. The target control instruction is issued to each sealed controller execution unit through an Ethernet interface.

[0118] In some embodiments, the internal structure of the whole vehicle control strategy module is divided into three sub-modules, namely a strategy mapping module, an execution judgment module, and an exception fault tolerance module. The strategy mapping module generates a corresponding target control instruction according to the input target wading risk level by calling a predefined strategy table. Each level of target wading risk level is bound to a specific set of execution action combination, for example, low wading risk only performs sealed port state preloading, medium wading risk triggers valve closure in advance, and high wading risk performs comprehensive sealing and negative pressure regulation linkage action. The execution judgment module is responsible for confirming whether the vehicle meets the triggering condition and judging the timing of issuing the target control instruction. The execution judgment module is embedded with a time window buffer mechanism to avoid strategy jitter or actuator repeated triggering caused by frequent jumps, and to ensure stable and effective execution actions. The exception fault tolerance module is responsible for identifying problems such as issuance failure, controller non-response, and feedback exception. After each execution of the target control instruction, the execution feedback signal is listened to, which can include but is not limited to cabin pressure change, sealed valve position feedback, controller status code, etc. If no confirmation signal is received within a specified time, a redundant control path is triggered to force the target control instruction to be executed in an emergency mode, and if necessary, a fault state is reported to the whole vehicle domain controller for safety processing. At the same time, the exception fault tolerance module supports step-by-step rollback logic, which can automatically downgrade the execution strategy according to the execution feedback signal in the case of high risk level mis-triggering, to avoid system over-protection or function conflict caused by misjudgment.

[0119] It should be noted that the whole vehicle control strategy module is deployed in the whole vehicle high reliability operation and control domain controller, and can realize high-speed data interaction with the vehicle power domain, battery domain and chassis domain controllers. The target control instruction and the execution feedback signal adopt an encryption transmission mechanism to prevent signal disturbance in an extreme electromagnetic environment and ensure the safety and accuracy of the execution action. The design of the whole vehicle control strategy module not only ensures efficient closed-loop execution of risk judgment, but also has integrity of engineering deployment, scalability of algorithm logic and compatibility of software and hardware interfaces, thereby providing the whole vehicle with a highly consistent, fast response and safe and reliable active protection capability in off-road and water-involved environments.

[0120] In this embodiment, the target control instruction corresponding to the target water-involved risk level is determined, so that the vehicle is controlled based on the target control instruction in the case where the vehicle meets the triggering condition, the hierarchical and accurate intervention can be implemented according to the target water-involved risk level, the vehicle is protected in the water-involved driving scene, and the safety of the vehicle in a complex environment is improved.

[0121] Figure 4 is a structural schematic diagram of a water-involved risk identification device provided by an embodiment of the present application, referring to Figure 4 The water-involved risk identification device 400 comprises: The acquisition module 401 is configured to acquire target state information of the vehicle, and the target state information comprises at least one of the following: water level information, image information, temperature information, humidity information, air pressure information, speed information and acceleration information. The determination module 402 is configured to determine water-involved related information corresponding to the vehicle based on the target state information, and the water-involved related information comprises a water-involved probability distribution, a water-involved risk score and a water-involved clustering result; and determine the target water-involved risk level corresponding to the vehicle based on the water-involved related information.

[0122] In a possible implementation, the acquisition module 401 is configured to acquire a target weight coefficient corresponding to the water-involved related information; and the determination module 402 is configured to perform correction processing on the water-involved related information based on the target weight coefficient to obtain at least one water-involved weighted information; and determine the target water-involved risk level based on the water-involved weighted information.

[0123] In a possible implementation, the acquisition module 401 is configured to acquire at least one historical water-involved risk level and at least one historical state information; and the device further comprises an adjustment module configured to adjust the target weight coefficient based on the historical water-involved risk level and the historical state information.

[0124] In a possible implementation, the determining module 402 is configured to determine a wading probability distribution based on the first state information, the first state information comprising at least one of the following: water level information, image information, speed information, and acceleration information; determine a wading risk score based on the second state information and / or the wading probability distribution, the second state information comprising at least one of the following: temperature information, humidity information, and air pressure information; and determine a wading clustering result based on the wading risk score.

[0125] In a possible implementation, the determining module 402 is configured to determine a first feature vector corresponding to the first state information; and determine the wading probability distribution based on the first feature vector by using a Gaussian mixture model.

[0126] In a possible implementation, the determining module 402 is configured to perform fuzzy processing on the second state information and / or the wading probability distribution to obtain at least one fuzzy linguistic variable; process the fuzzy linguistic variable based on fuzzy rules to obtain a fuzzy set; and determine the wading risk score based on the fuzzy set.

[0127] In a possible implementation, the determining module 402 is configured to determine a second feature vector corresponding to the wading risk score; perform clustering processing on the second feature vector to obtain at least one clustering center feature; and determine the wading clustering result based on the clustering center feature.

[0128] In a possible implementation, the determining module 402 is configured to determine a target control instruction corresponding to a target wading risk level; and the apparatus further includes a control module configured to control the vehicle based on the target control instruction when the vehicle meets a triggering condition.

[0129] It should be noted that the wading risk identification apparatus provided in the above embodiments is only used as an example for the division of the above functional modules in controlling the vehicle, and in actual applications, the above functions can be completed by different functional modules according to needs, that is, the internal structure of the computer device is divided into different functional modules to complete all or part of the above described functions. In addition, the wading risk identification apparatus and the wading risk identification method provided in the above embodiments belong to the same concept, and the specific implementation process is described in the method embodiments, which will not be repeated here.

[0130] By adopting the technical solutions provided in the embodiments of the present application, the water-related information including the water-involved probability distribution determined by the first model, the water-involved risk score determined by the second model, and the water-involved clustering result is determined based on the acquired multi-dimensional target state information, and the target water-involved risk level of the vehicle is determined based on the water-related information, so that the water-involved risk level of the vehicle can be accurately detected by multi-dimensional target state information, multi-source information fusion, and multi-level risk assessment, the comprehensiveness of vehicle risk perception is significantly improved, meanwhile, the complex scenarios of single sensor failure or incomplete data can be effectively coped with, the robustness and adaptability of the vehicle in extreme weather are enhanced, and accurate safety warning is provided for water-involved driving of the vehicle.

[0131] The embodiments of the present application further provide a vehicle, Figure 5 FIG. 1 is a structural schematic diagram of a vehicle provided by the embodiments of the present application.

[0132] Generally, the vehicle 500 includes one or more processors 501 and one or more memories 502.

[0133] The processor 501 can include one or more processing cores, such as a 4-core processor, a 5-core processor, etc. The processor 501 can be implemented in at least one of a hardware form of a DSP (Digital Signal Processing), an FPGA (Field-Programmable Gate Array), and a PLA (Programmable Logic Array). The processor 501 can also include a main processor and a coprocessor. The main processor is a processor for processing data in an awake state, also known as a CPU (Central Processing Unit). The coprocessor is a low-power processor for processing data in a standby state. In some embodiments, the processor 501 can be integrated with a GPU (Graphics Processing Unit) that is responsible for rendering and drawing the content to be displayed by the display screen. In some embodiments, the processor 501 can further include an AI (Artificial Intelligence) processor for processing machine learning-related computing operations.

[0134] The memory 502 can include one or more computer-readable storage media. The computer-readable storage media can be non-transitory. The memory 502 can also include high-speed random access memory and can include non-volatile memory, such as one or more magnetic disk storage devices, optical storage devices, flash memory devices, or other non-volatile solid-state storage devices. In some embodiments, the non-transitory computer-readable storage medium in the memory 502 is used to store at least one computer program for being executed by the processor 501 to implement the method for identifying the water-related risk provided by the method embodiments in the present application.

[0135] Those skilled in the art can understand that, Figure 5 The structure shown in the figure does not constitute a limitation on the vehicle 500, and can include more or fewer components than shown, or combine certain components, or adopt a different arrangement of components.

[0136] In addition, the apparatus provided by the embodiments of the present application can be a chip, a component, or a module, which can include a processor and a memory connected to each other. The memory is used to store instructions, and when the processor invokes and executes the instructions, the chip can execute the method for identifying the water-related risk provided by the above-mentioned embodiments.

[0137] The embodiment also provides a computer-readable storage medium, which stores computer program codes. When the computer program codes run on a computer, the computer can execute the above-mentioned related method steps to implement the method for identifying the water-related risk provided by the above-mentioned embodiments.

[0138] The embodiment also provides a computer program product, which makes the computer execute the above-mentioned related steps to implement the method for identifying the water-related risk provided by the above-mentioned embodiments when the computer program product runs on the computer.

[0139] The apparatus, computer-readable storage medium, computer program product, or chip provided by the embodiments can be used to execute the corresponding method provided above, and thus can achieve the beneficial effects of the corresponding method provided above, which will not be described here.

[0140] From the above description of the embodiments, those skilled in the art can understand that, for the convenience and brevity of description, only the division of the above functional modules is taken as an example for illustration. In actual application, the above functions can be completed by different functional modules according to needs, i.e., the internal structure of the apparatus is divided into different functional modules to complete all or part of the functions described above.

[0141] In the embodiments of the present disclosure, it should be understood that the disclosed apparatus and method can be implemented in other manners. For example, the embodiments of the apparatus described above are merely schematic; for example, the division of the modules or units is merely logical function division; and an actual mapping relationship can be different, for example, a plurality of units or components can be combined or integrated into another device, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections can be indirect couplings or communication connections through some interfaces, devices or units, and can be in electrical, mechanical or other forms.

[0142] The foregoing is merely specific implementation of the present disclosure, but the protection scope of the present disclosure is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present disclosure, which should be covered by the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be subject to the protection scope of the claims.

Claims

1. A method of identifying a risk of wading, characterized in that, The method is applied in a vehicle, and the method includes: Obtain the target status information of the vehicle, which includes at least one of the following: water level information, image information, temperature information, humidity information, air pressure information, speed information, and acceleration information; Based on the target state information, the water-related information corresponding to the vehicle is determined. The water-related information includes: water-wading probability distribution, water-wading risk score, and water-wading clustering results. The water-wading probability distribution is determined by a first model, which is used to quantize the input data into a probability density function. The water-wading risk score is determined by a second model, which is used to perform fuzzy processing on the input data. Based on the water-related information, the target water-related risk level of the vehicle is determined.

2. The method of claim 1, wherein, The step of determining the target water-related risk level of the vehicle based on the water-related information includes: Obtain the target weight coefficients corresponding to the water-related information; Based on the target weighting coefficient, the water-related information is corrected to obtain at least one water-related weighted information. Based on the aforementioned water-related weighted information, the target water-related risk level is determined.

3. The method of claim 2, wherein, The method further includes: Obtain at least one historical water risk level and at least one historical status information; The target weight coefficient is adjusted based on the historical water risk level and the historical status information.

4. The method of claim 1, wherein, The determination of water-related information corresponding to the vehicle based on the target state information includes at least one of the following: Based on the first state information, the water wading probability distribution is determined using the first model. The first state information includes at least one of the following: water level information, image information, velocity information, and acceleration information. The water wading risk score is determined using the second model based on the second state information and / or the water wading probability distribution. The second state information includes at least one of the following: the temperature information, the humidity information, and the air pressure information. Based on the water-related risk score, the water-related clustering result is determined.

5. The method of claim 4, wherein, Determining the water wading probability distribution based on the first state information includes: Determine the first feature vector corresponding to the first state information; Based on the first feature vector, the water wading probability distribution is determined.

6. The method of claim 4, wherein, Determining the water-related risk score based on the second state information and / or the water-related probability distribution includes: The second state information and / or the water wading probability distribution are subjected to fuzzy processing to obtain at least one fuzzy linguistic variable; The fuzzy linguistic variables are processed based on fuzzy rules to obtain a fuzzy set; The water-related risk score is determined based on the fuzzy set.

7. The method of claim 4, wherein, The determination of the water-related clustering results based on the water-related risk score includes: Determine the second feature vector corresponding to the water-related risk score; Clustering is performed on the second feature vector to obtain at least one cluster center feature; Based on the cluster center characteristics, the water-related clustering results are determined.

8. The method according to any one of claims 1 to 7, characterized in that, The method further includes: Determine the target control instructions corresponding to the target water risk level; In a case where the vehicle meets a triggering condition, the vehicle is controlled based on the target control instruction.

9. A wading risk identification device, characterized in that The device is applied to a vehicle, and the device comprises: an acquisition module, configured to acquire target state information of the vehicle, the target state information comprising at least one of water level information, image information, temperature information, humidity information, air pressure information, speed information, and acceleration information; a determination module, configured to determine, based on the target state information, water-related information corresponding to the vehicle, the water-related information comprising a water probability distribution, a water risk score, and a water clustering result, the water probability distribution being determined by a first model, the first model being used to quantize input data into a probability density function, the water risk score being determined by a second model, the second model being used to perform fuzzy processing on input data; and determine a target water risk level corresponding to the vehicle based on the water-related information.

10. A vehicle characterized by comprising: The vehicle comprises: a memory, configured to store executable program code; a processor, configured to call and run the executable program code from the memory, so that the vehicle performs the method according to any one of claims 1 to 8.

Images