Vehicle flammable material disposal method, device, and vehicle

Abstract

The application relates to the technical field of vehicles and discloses a vehicle flammable substance processing method and device and a vehicle. The method comprises the following steps: acquiring multi-modal environment data and cabin state data of a vehicle; inputting the multi-modal environment data and the cabin state data into an expert model to obtain a target flammable substance in the vehicle. The flammable risk decision score of the target flammable substance is higher than a first preset score. The expert model comprises a plurality of sub-expert models corresponding to a plurality of flammable scenes in a one-to-one manner. The flammable risk decision score is determined based on a plurality of flammable risk evaluation results of the target flammable substance in the plurality of flammable scenes learned by the plurality of sub-expert models and the cabin state data. The target flammable substance is processed based on the type and the flammable risk decision score of the target flammable substance. The technical scheme provided in the application embodiment improves the efficiency and accuracy of vehicle flammable substance processing and is suitable for complex and various scenes.

Description

Technical Field

[0001] This invention relates to the field of vehicle technology, and more specifically to a method, apparatus, and vehicle for handling flammable materials in a vehicle. Background Technology

[0002] Vehicle flammable material handling refers to the identification, control, removal, or compliant disposal of various flammable materials present on a vehicle throughout its entire lifecycle, from production and use to maintenance and scrapping, in order to prevent potential risks such as fires and explosions.

[0003] In related technologies, the methods for identifying flammable materials inside vehicles are too simplistic, such as using a single smoke sensor or camera, resulting in low accuracy. Furthermore, the handling of flammable materials primarily relies on fixed decision-making rules, such as determining if the temperature exceeds a certain threshold, making it difficult to apply to more complex and variable scenarios involving flammable materials inside vehicles. Summary of the Invention

[0004] In view of the shortcomings of the prior art, the purpose of this application is to provide a method, apparatus and vehicle for handling flammable materials in vehicles, which improves the efficiency and accuracy of handling flammable materials in vehicles and is applicable to complex and diverse scenarios.

[0005] In a first aspect, embodiments of this application provide a method for handling flammable materials in a vehicle. The method includes: acquiring multimodal environmental data and cabin state data of the vehicle; inputting the multimodal environmental data and cabin state data into an expert model to obtain a target flammable material in the vehicle; wherein the flammability risk decision score of the target flammable material is higher than a first preset score; the expert model includes multiple sub-expert models corresponding one-to-one with multiple flammable scenarios; the flammability risk decision score is determined based on multiple flammability risk assessment results of the target flammable material in multiple flammable scenarios learned by the multiple sub-expert models and the cabin state data; and processing the target flammable material based on its type and flammability risk decision score.

[0006] The technical solution provided in this application uses multiple independent sub-expert models to assess the flammability risk of flammable materials in different flammable scenarios. The data is isolated and reasoning is synchronized between the sub-models, effectively improving the efficiency and accuracy of flammable material risk assessment, and thus enhancing the timeliness and accuracy of handling flammable materials. Simultaneously, the expert models combine multimodal environmental data and cabin status data for reasoning, accurately analyzing and determining the flammability risk of flammable materials from multiple dimensions, further improving the accuracy of flammable material risk assessment and making it suitable for complex and ever-changing scenarios.

[0007] One possible implementation involves using cabin status data including the number of occupants, seating arrangement, and window opening / closing status. The process of determining the flammability risk decision score can be specifically implemented as follows: based on the number of occupants, seating arrangement, window opening / closing status, and the risk assessment performance of multiple sub-expert models, determine the risk assessment weights corresponding to each of the multiple flammable scenarios. Based on the multiple flammability risk assessment results and risk assessment weights, determine the flammability risk decision score. By measuring the importance of the sub-expert models in flammability risk prediction using cabin status data such as the number of occupants, seating arrangement, and window opening / closing status, the model processing results more closely reflect actual vehicle cabin scenarios, further improving the accuracy of flammability risk assessment results in complex scenarios.

[0008] One possible implementation involves defining flammable scenarios as including: flammable scenarios caused by vehicle collisions, flammable scenarios caused by aging electrical equipment, and flammable scenarios caused by the natural environment. The process of determining multiple flammable risk assessment results can be specifically implemented as follows: First, determine the sub-expert models corresponding to the flammable scenarios caused by vehicle collisions, flammable scenarios caused by aging electrical equipment, and flammable scenarios caused by the natural environment. Second, classify and identify multimodal environmental data to determine the input data for the sub-expert models corresponding to the flammable scenarios caused by vehicle collisions, flammable scenarios caused by aging electrical equipment, and flammable scenarios caused by the natural environment. Third, based on the sub-expert models corresponding to the flammable scenarios caused by vehicle collisions, flammable scenarios caused by aging electrical equipment, and flammable scenarios caused by the natural environment, and the input data of the sub-expert models corresponding to these scenarios, determine multiple flammable risk assessment results. The above technical solution subdivides flammable scenarios into flammable scenarios caused by vehicle collisions, flammable scenarios caused by aging electrical equipment, and flammable scenarios caused by the natural environment. For different flammable scenarios, different sub-expert models and multimodal data are used to assess the flammability risk of each scenario. The flammability risk assessment results of different sub-expert models are more in line with the characteristics of the corresponding flammable scenarios, thereby improving the accuracy of the risk scores determined under different flammable scenarios.

[0009] One possible implementation is that the expert model also includes a classification sub-model, which is used to determine the type of the target flammable material, including solid and liquid states. Identifying flammable material types through a separate sub-model is more efficient, and combining flammable material type with risk decision scores allows for targeted processing of flammable materials, better meeting the needs of real-world scenarios.

[0010] One possible implementation involves processing the target flammable material based on its type and flammability risk decision score. Specifically, if the material is solid and its flammability risk decision score is higher than a second preset score, the vehicle's fire extinguishing device is used to extinguish the fire. The second preset score is higher than the first preset score. When a solid flammable material has a high flammability risk decision score, the combustion risk is high, and the flammable material can be extinguished using the vehicle's own fire extinguishing device.

[0011] One possible implementation involves controlling a vehicle's fire extinguishing device to extinguish a target flammable material. Specifically, this can be achieved by: determining the spray direction of the fire extinguishing device based on its positional relationship with the target flammable material; determining the spray flow rate of the fire extinguishing device based on a flammability risk decision score; and controlling the fire extinguishing device to extinguish the fire on the target flammable material based on the spray direction and flow rate. By accurately locating the combustion site and extinguishing requirements based on the flammability risk decision score and the location of the flammable material, targeted fire suppression can be achieved, resulting in better fire suppression outcomes.

[0012] One possible implementation involves processing the target flammable material based on its type and flammability risk decision score. This further includes controlling the opening of vehicle windows and / or doors when the material is solid and the flammability risk decision score is higher than a second preset score. Opening the windows and / or doors while extinguishing the fire effectively reduces the accumulation of smoke and gas inside the vehicle.

[0013] One possible implementation involves processing the target flammable material based on its type and flammability risk decision score. This further includes outputting an alarm notification when the material is in a liquid state, or when it is in a solid state and the flammability risk decision score falls within a closed interval formed by a first preset score and a second preset score. The alarm notification includes a vehicle-mounted voice alarm and a remote alarm from the vehicle owner's mobile terminal. Since some types of fire extinguishing devices are inapplicable when the flammable material is in a liquid state, providing an early warning when the flammability risk decision score is low allows users to take timely measures.

[0014] One possible implementation involves an expert model trained in the cloud and distributed to vehicles within the target area. The vehicle flammable material handling method provided in this application further includes: collecting vehicle owners' evaluations after handling the target flammable material; generating optimized samples based on multimodal environmental data, cabin status data, and the evaluations; and reporting the optimized samples to the cloud. The cloud updates and re-distributes the expert model based on optimized samples reported by vehicles within the target area within a preset period. By updating and training the expert model using a large amount of vehicle flammable material handling data within the target area, the cloud effectively improves the prediction accuracy of the expert model.

[0015] One possible implementation involves optimizing samples including post-treatment evaluation of flammable materials, changes in multimodal data, and changes in cabin status data. The cloud-based system updates the expert model parameters based on these data. By incorporating feedback from flammable material treatment across various dimensions, the cloud-based system further improves the predictive accuracy of the updated expert model. Another possible implementation utilizes multimodal environmental data, including image data, smoke sensor data, temperature sensor data, humidity sensor data, and vehicle seat pressure sensor data.

[0016] Secondly, this application provides a vehicle flammable material handling device, which includes an acquisition module and a processing module.

[0017] The aforementioned acquisition module is used to acquire multimodal environmental data and cabin status data of the vehicle.

[0018] The aforementioned processing module is used to input multimodal environmental data and cabin status data into an expert model to obtain the target flammable material in the vehicle. The flammability risk decision score of the target flammable material is higher than a first preset score. The expert model includes multiple sub-expert models corresponding one-to-one with multiple flammable scenarios. The flammability risk decision score is determined based on multiple flammability risk assessment results of the target flammable material under multiple flammable scenarios learned by the multiple sub-expert models, as well as cabin status data.

[0019] The aforementioned processing module is also used to process the target flammable material based on its type and flammability risk decision score.

[0020] One possible implementation involves using cabin status data including: the number of occupants, seat distribution, and window opening / closing status. The aforementioned processing module is specifically used to: determine the risk assessment weights for multiple flammable scenarios based on the number of occupants, seat distribution, window opening / closing status, and the risk assessment performance of multiple sub-expert models. Based on the multiple flammable risk assessment results and risk assessment weights, a flammable risk decision score is determined.

[0021] One possible implementation is that the expert model also includes a classification sub-model, which is used to determine the type of the target flammable material, including solid and liquid states.

[0022] One possible implementation is that the aforementioned processing module is specifically used to: when the type is solid and the flammability risk decision score is higher than the second preset score, control the fire extinguishing device in the vehicle to extinguish the target flammable material. The second preset score is greater than the first preset score.

[0023] One possible implementation is that the aforementioned processing module is specifically used to: determine the spray direction of the fire extinguishing device based on the positional relationship between the fire extinguishing device and the target flammable material; determine the spray flow rate of the fire extinguishing device based on the flammability risk decision score; and control the fire extinguishing device to extinguish the target flammable material based on the spray direction and spray flow rate.

[0024] In one possible implementation, the aforementioned processing module is further configured to: control the opening of the vehicle's windows and / or doors when the type is solid and the flammability risk decision score is higher than the second preset score.

[0025] In one possible implementation, the aforementioned processing module is further configured to: output an alarm notification when the substance is liquid, or solid and the flammability risk decision score falls within the closed interval formed by the first preset score and the second preset score. The alarm notification includes a vehicle-mounted voice alarm and a remote alarm from the vehicle owner's mobile terminal.

[0026] In one possible implementation, the aforementioned processing module is further configured to: display the target flammable material and its corresponding processing operation on the vehicle's display device when the type is solid and the flammability risk decision score is higher than the second preset score.

[0027] One possible implementation involves the expert model being trained in the cloud and distributed to vehicles within the target area. The aforementioned processing module is further used to: collect vehicle owners' evaluations after processing the target flammable material; generate optimized samples based on multimodal environmental data, cabin status data, and processing evaluations; and report the optimized samples to the cloud. The cloud updates and re-distributes the expert model based on optimized samples reported by vehicles within the target area within a preset period.

[0028] One possible implementation approach is to include flammable scenarios such as those caused by vehicle collisions and those caused by natural environmental factors. Multimodal environmental data includes image data, smoke sensor data, temperature sensor data, humidity sensor data, and pressure sensor data from vehicle seats.

[0029] Thirdly, this application provides a vehicle. The vehicle includes the vehicle flammable material treatment device according to any embodiment of the second aspect above, or the vehicle treats flammable materials using the vehicle flammable material treatment method according to any embodiment of the first aspect above.

[0030] Fourthly, this application provides a computer-readable storage medium storing at least one computer program, which is loaded and executed by a processor to implement the vehicle flammable material handling method in any of the embodiments of the first aspect described above.

[0031] Fifthly, this application provides a computer program product, which includes a computer program or instructions that, when executed by a processor, implement the vehicle flammable material handling method in any of the embodiments of the first aspect described above.

[0032] The solutions provided in the third to fifth aspects above can realize the vehicle flammable material handling method in any embodiment of the first aspect above, and their specific implementations will not be described in detail here. The technical effects corresponding to any implementation of the solutions provided in the third to fifth aspects above can be found in the technical effects corresponding to any implementation of the first aspect above, and will not be described in detail here.

[0033] It should be noted that any of the possible implementations of any of the above aspects can be combined, provided that the solutions do not contradict each other. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of this application or the background art, the accompanying drawings used in the embodiments of this application will be described below.

[0035] Figure 1 This is a schematic diagram of the structure of a vehicle flammable material handling device disclosed in an embodiment of this application; Figure 2 This is a schematic flowchart of a method for handling flammable materials from a vehicle, as disclosed in an embodiment of this application. Figure 3 This is a schematic flowchart of another method for handling flammable materials in a vehicle disclosed in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of a vehicle flammable material handling device disclosed in an embodiment of this application.

[0036] Explanation of reference numerals in the attached figures: 100 - Sensor; 200 - Processor; 300 - Actuator. Detailed Implementation

[0037] To enable those skilled in the art to better understand the technical solutions of this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings.

[0038] It should be noted that the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0039] In the embodiments of this application, the words "exemplary," "for example," or "e.g.," are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "exemplary," "for example," or "e.g.," in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the words "exemplary," "for example," or "e.g.," is intended to present the relevant concepts in a specific manner.

[0040] The embodiments of this application are described below with reference to the accompanying drawings.

[0041] This application provides a vehicle.

[0042] Alternatively, a vehicle may also be referred to as a vehicle, mobile carrier, electric vehicle (EV), hybrid electric vehicle (HEV), plug-in hybrid electric vehicle (PHEV), fuel cell vehicle (FCV), autonomous vehicle, intelligent and connected vehicle (ICV), driverless vehicle, or new energy vehicle. In this application embodiment, the vehicle may be a sedan, sport utility vehicle (SUV), truck, electric vehicle, motorcycle, tricycle, special vehicle (such as ambulance, fire truck, police car, etc.), driverless taxi, intelligent connected bus, autonomous logistics vehicle, electric truck, etc. The method provided in this application embodiment is also applicable to various special-purpose vehicles, such as agricultural vehicles, mining vehicles, forestry vehicles, airport vehicles, and port vehicles; this application does not impose specific limitations on these.

[0043] In some embodiments, the vehicle includes a flammable material handling device.

[0044] Among them, flammable material handling equipment is used to identify, collect, control or handle flammable materials inside / around a vehicle, reducing the risk of fire or explosion caused by flammable materials in the vehicle.

[0045] For example, please refer to Figure 1 , Figure 1 This is a schematic diagram of a vehicle flammable material handling device provided in an embodiment of this application. The vehicle flammable material handling device provided in this embodiment includes a sensor 100, a processor 200, and an actuator 300.

[0046] The sensor 100 is used to monitor information related to flammable materials, such as combustible gas concentration, temperature, smoke, battery pack pressure, and static electricity. Optionally, the sensor 100 includes a temperature sensor, a combustible gas sensor, a smoke sensor, an infrared camera, and a current sensor.

[0047] The processor 200 analyzes and processes the information monitored by the sensor 100, assesses the flammability risk of flammable materials, and sends control commands to the actuator 300 based on the risk assessment results, controlling the actuator 300 to execute corresponding processing strategies. Optionally, the processor 200 includes a vehicle controller, a body domain controller, a cabin domain controller, an electronic control unit, and a dedicated fire safety controller.

[0048] Actuator 300 receives control commands from processor 200 and takes physical measures to intervene in potential flammability risks. Optionally, actuator 300 includes a fire extinguisher, window control unit, power cut-off unit, fuel cut-off valve, warning light, etc.

[0049] This application also provides a method for handling flammable materials from a vehicle. This method can be applied to the flammable material handling equipment described above, or to other equipment capable of performing flammable material handling functions. For example, please refer to... Figure 2 The present application provides a method for handling flammable materials from vehicles, which includes: Step S201: The flammable material handling equipment acquires multimodal environmental data and cabin status data of the vehicle.

[0050] Multimodal environmental data refers to environmental data of different dimensions associated with flammable materials, collected through various types of sensors. Optionally, multimodal environmental data includes image data, smoke sensing data, temperature sensing data, humidity sensing data, circuit current and voltage data, and vehicle seat pressure sensing data, etc.

[0051] Cockpit status data reflects the vehicle's interior (cabin) environment and occupant status. Optionally, cockpit status data includes the number of occupants, seating arrangements, window opening / closing status, door locking status, driver operation status, and the status of the vehicle's electrical equipment and drive circuits.

[0052] Multimodal environmental data and cabin status data can directly influence or reflect the flammability risk of flammable materials inside a vehicle. For example, high temperatures inside the cabin (such as temperatures exceeding 65°C after exposure to direct sunlight) can accelerate the volatilization or thermal runaway of flammable materials such as perfumes, power banks, and lighters. When temperature sensors detect that the cabin temperature is close to the ignition point of flammable materials, the flammability risk is high. Another example is that when electrical appliances such as onboard chargers and seat heaters are operating in the cabin, and flammable materials leak, electrical sparks can easily cause circuit scorching and fire.

[0053] Examples of ways to acquire multimodal data include, but are not limited to: Example 1: Deploy in-vehicle cameras in front of the seats and in the roof to cover the areas inside the cabin. The cameras continuously capture images of the cabin interior at a frame rate f (e.g., 30Hz) and output an RGB format matrix. Where t represents the sampling time, and the image dimension is the width. high 3. Supports real-time video stream processing.

[0054] Example 2: The smoke sensor is installed at the center of the top of the cockpit to detect the smoke concentration. The smoke sensor has a sampling rate of 10Hz, a data range of 0 to 1000ppm, and outputs a digital signal through an analog-to-digital converter.

[0055] Example 3: Temperature sensors are distributed in the dashboard and seat area to measure ambient temperature (in degrees Celsius). The sampling rate of the temperature sensors is 5Hz, and thermocouples can be used to ensure that the sampling accuracy error is less than [value missing]. Celsius.

[0056] Example 4: The seat pressure sensor is embedded in the seat cushion and backrest to sense distributed pressure values ​​or total pressure. The sampling rate of the seat pressure sensor is 20Hz, and the output is an array of force values.

[0057] Since the sampling rates of the various sensors are different, it is necessary to use global synchronization to unify the timestamps of the multimodal environmental data obtained from the detection. For example, the sampling interval of each sensor is calculated. Where f represents the sampling rate. The timestamp sequence is unified using the minimum sampling interval; specifically, the sampling interval is the time when the latest data from each sensor is collected. Multiples of. In cases of sampling mismatch, data is supplemented through linear interpolation. For example: smoke sensors in The data at time t can be expressed as the following formula:

[0058] in, For smoke sensors Data monitored in real time For smoke sensors Data monitored in real time exist Time and In a short period of time, This represents the sampling interval for the smoke sensor.

[0059] After multimodal data acquisition is completed, the data from different modes are converted into data vectors and transmitted in real time to the processor (such as the central processing unit) via the controller area network (CAN) bus or Ethernet protocol, with the transmission delay controlled within 5ms.

[0060] For example, cabin status data can be directly collected through devices such as window / door position sensors, cameras, and pressure sensors. Vehicle electrical data (such as the operating status of the central control screen, airbag status, and on-board charger power) can be read through vehicle bus data. The interaction between the driver and the vehicle (such as steering wheel angle, accelerator / decelerator pedal opening, gear shift lever position, manual control commands for windows / doors, Bluetooth connection status, and gesture control actions) can be recorded through operation logs.

[0061] Step S202: The flammable material handling device inputs multimodal environmental data and cabin status data into the expert model to obtain the target flammable material in the vehicle. The flammability risk decision score of the target flammable material is higher than the first preset score.

[0062] The expert model includes multiple sub-expert models that correspond one-to-one with multiple flammable scenarios. The flammability risk decision score is determined based on the flammability risk assessment results of the target flammable material in multiple flammable scenarios learned by the multiple sub-expert models, as well as the cabin status data.

[0063] A flammable scenario refers to a scenario in which flammable materials ignite. Optionally, flammable scenarios include flammable scenarios caused by vehicle collisions (such as scenarios where a vehicle collision leads to fuel tank leakage and circuit damage), flammable scenarios caused by aging electrical equipment (such as scenarios where short circuits or electric sparks generated when a charging pile is overloaded ignite flammable materials), and flammable scenarios caused by the natural environment (such as scenarios where a vehicle is exposed to high temperatures for a long time). Each of the two flammable scenarios corresponds to a sub-expert model.

[0064] In some embodiments, the expert model's assessment of the flammability risk of flammable materials includes: Step 1: Classify and identify the multimodal environmental data to determine the multimodal data corresponding to each sub-expert model. For example, determine the sub-expert models corresponding to flammable scenarios caused by vehicle collisions, flammable scenarios caused by aging electrical equipment, and flammable scenarios caused by the natural environment. Classify and identify the multimodal data of the sub-expert models corresponding to these scenarios as the input data for each sub-expert model.

[0065] For example, a data source category is identified using a data type detector, utilizing metadata or preprocessing analysis (such as dimensionality checking). Once identified, a routing mechanism based on hard-coded rules transmits multimodal environmental data to the corresponding sub-expert model, ensuring input queue isolation to avoid cross-interference. For instance, temperature data is transmitted to the sub-expert model corresponding to a flammable scenario caused by the natural environment.

[0066] Step 2: Each sub-expert model performs risk assessment based on the corresponding multimodal data and outputs the risk score corresponding to the flammable material as the flammability risk assessment result.

[0067] As an example, the flammability risk assessment process for sub-expert models corresponding to different flammable scenarios includes the following, but is not limited to these.

[0068] Scenario 1: Flammable scenario caused by aging electrical equipment.

[0069] Optionally, flammable scenarios caused by aging electrical equipment include scenarios where electrical sparks / high temperatures ignite surrounding flammable materials when abnormalities occur in the vehicle's electrical system (such as wiring harnesses, batteries, central control screens, charging piles, on-board chargers, and other charging circuits) (such as short circuits, overloads, aging, or poor disconnection). This is different from circuit damage caused by vehicle collisions. Flammable scenarios caused by aging electrical equipment are mainly caused by electrical hazards during routine vehicle operation or when the vehicle is parked quietly.

[0070] For example, the cabin state data input to the sub-expert model corresponding to the flammable scenario caused by aging electrical equipment includes the distance and quantity between flammable materials and electrical components, as well as the operating parameters of electrical equipment (such as voltage, current, and power). Multimodal environmental data includes temperature, humidity, and vehicle parking time. During the training phase, the sub-expert model corresponding to the flammable scenario caused by aging electrical equipment learns the feature database corresponding to various abnormal operating conditions of vehicle electrical equipment (such as the current mutation threshold during short circuits, the resistance change of aging wiring harnesses, and the temperature peak of charging pile overload). Combining the feature database and the flammability characteristics of flammable materials (such as ignition point and energy required for ignition), it assesses the risk score (quantified as 0-10 points) of "aging electrical equipment igniting surrounding flammable materials".

[0071] Scenario 2: Flammable scenarios caused by the natural environment.

[0072] Optionally, flammable scenarios caused by the natural environment include flammable materials stored in the vehicle cabin for a long time, and scenarios where the flammability increases due to aging, deterioration, leakage, etc. caused by time and environmental factors, thus causing flammability risks. Examples include expired perfumes evaporating faster, lithium batteries aging and bulging, and flammable liquids leaking due to seal failure. This is different from flammable materials stored for a short time, and focuses on flammable scenarios caused by flammable materials stored for a long time.

[0073] For example, the cabin state data input to the sub-expert model corresponding to the flammable scenario caused by the natural environment includes vehicle driving frequency (flammable materials in long-term idle vehicles age faster), storage time of flammable materials, initial production date / shelf life, and appearance condition, such as the degree of lithium battery bulging and the integrity of flammable liquid seals. Multimodal environmental data includes cabin temperature and humidity, light intensity (long-term exposure to sunlight accelerates the aging of flammable materials), seasonal temperature and humidity changes and sunshine duration of the vehicle's external environment (high temperature and high humidity environments accelerate the deterioration of flammable materials). During the training phase, the sub-expert model corresponding to the flammable scenario caused by the natural environment learns the aging / deterioration pattern database of various types of vehicle-mounted flammable materials (such as the flammability risk increase of lithium batteries for one year of storage, the volatility change of perfume after expiration, and the time threshold for flammable liquid seal failure). Combining the aging / deterioration pattern database, multimodal environmental data, and cabin turntable data, it assesses the risk score (quantified as 0-10 points) of flammable materials caused by aging / deterioration, thereby realizing the monitoring of flammability risks caused by changes in the state of flammable materials themselves.

[0074] Scenario 3: A flammable scenario caused by the coupling of extreme natural environment and vehicle operating conditions.

[0075] Optionally, extreme natural environments include severe thunderstorms, blizzards, sandstorms, and sharp temperature drops, while vehicle operating conditions include high-speed driving, idling, and charging. When extreme natural environments are coupled with vehicle operating conditions, the risk of flammability of flammable materials is likely to increase. Moreover, unlike single environmental factors or single vehicle factors, the flammability risk generated when extreme natural environments are coupled with vehicle operating conditions is higher and more concealed.

[0076] For example, the cabin state data input to the sub-expert model corresponding to the flammable scenario caused by the coupling of extreme natural environment and vehicle operating conditions includes the type of flammable material, its storage location, flammability characteristics, and vehicle operating parameters. The vehicle operating parameters include engine speed when the vehicle is driving at high speed, charging power when charging, and exhaust emission temperature when idling. The multimodal environmental data includes lightning intensity during thunderstorms, ambient temperature during blizzards, particulate matter concentration during sandstorms, ambient wind speed, and air pressure (strong winds accelerate the spread of combustion). During the training phase, the sub-expert model for flammable scenarios caused by the coupling of extreme natural environments and vehicle operating conditions learns a database of flammable risk characteristics when various extreme natural environments and vehicle operating conditions are coupled. For example, the risk of ignition of flammable and explosive materials in the cabin when lightning strikes a vehicle during thunderstorms; the risk of flammable materials being ignited by high temperatures near the air vents when a vehicle is idling and heating during blizzards; and the risk of flammable materials being ignited by short circuits caused by particulate matter entering electrical components during sandstorms. Combining the flammable risk characteristic database and the flammability characteristics of flammable materials, the model assesses the risk score (quantified as 0-10 points) under flammable scenarios caused by the coupling of extreme natural environments and vehicle operating conditions. This covers the assessment of flammable hazards in extreme environments and improves the model's ability and adaptability to assess flammable risks in complex scenarios.

[0077] In some embodiments, the process of determining a flammability risk decision score includes, but is not limited to, the following steps: Step 1: Based on the number of people in the vehicle, the distribution of seating positions, the opening and closing status of the windows, and the risk assessment performance of multiple sub-expert models, determine the risk assessment weights corresponding to each of the multiple flammable scenarios.

[0078] Among them, the risk assessment weight represents the importance or confidence level of the judgment result of each sub-expert model under the current input cabin status data.

[0079] For example, real-time cabin status data is acquired via a CAN bus or sensor interface. This data includes the number of occupants, seat distribution, and window opening / closing status. The cabin status data is then normalized and its features extracted to obtain a state vector. Where R represents the real number field, This represents the feature dimension, such as the one-hot encoding of the number of people and the Boolean value of the car window status. The state vector s is input into the gating network, and the hidden layer of the gating network performs a non-linear transformation on the state vector s to obtain the hidden layer output: Among them, the gating network is constructed based on a multi-layer sensing mechanism (including an input layer, a hidden layer, and an output layer). This is the weight matrix. Let R be the bias vector, and let R denote the real number field. Representing feature dimension, Indicates the hidden layer dimension. This is the activation function of the hidden layer. The output 'h' of the hidden layer is passed to the output layer to calculate the weight vector. ,in, , and These are trainable parameters, where n represents the number of flammable scenarios, and the softmax function guarantees... , Indicates the first Risk assessment weights corresponding to each flammable scenario.

[0080] Step 2: Determine the flammability risk decision score based on multiple flammability risk assessment results and risk assessment weights.

[0081] For example, a flammable risk decision score is obtained by weighting and fusing multiple flammable risk scores under multiple flammable scenarios output by each sub-expert model, along with the corresponding risk assessment weights under multiple flammable scenarios. The flammable risk decision score is a score that represents the overall flammable risk level by fusing local flammable risk scores through global gating weights.

[0082] Specifically, the multimodal feature data input to the sub-expert model is represented as a vector. Where n represents the feature dimension. The sub-expert model employs a multilayer perceptron model, achieving nonlinear transformation through multiple hidden layers, with the first hidden unit calculating the output. ,in, This is the weight matrix. This is the bias term. The final regression layer outputs a risk decision score. ,in and These are the parameters for the output layer. The sub-expert model uses a mean squared error loss function during the training phase to ensure the output is aligned with the true labels.

[0083] In some embodiments, the gating network and expert model are optimized using historical driving data during the training phase to minimize decision-making errors. Specifically, the numerical basis of the flammability risk score and risk assessment weights is determined based on trainable parameters within the expert model, which are continuously adjusted during the training phase using historical driving data through an optimization algorithm to minimize prediction errors.

[0084] Step S203: The flammable material handling equipment handles the target flammable material based on its type and flammability risk decision score.

[0085] Among them, the target flammable material is the flammable material whose flammable risk decision score is higher than the first preset score after the flammable risk decision score is calculated. It represents the flammable material that needs to be handled.

[0086] Optionally, the type includes solid and liquid. The first preset score is either a default value or a manually set value, which is not limited in this application.

[0087] In some embodiments, the expert model also includes a classification sub-model for determining the type of the target flammable material.

[0088] For example, a data source category is identified using a data type detector, utilizing metadata or preprocessing analysis (such as dimensionality checking). When the data is determined to be image data, a routing mechanism transmits the image data to the classification sub-model based on hard-coded rules. Specifically, image data is typically represented as a multidimensional tensor, while other multimodal environmental data (such as sensor data) are time series or feature vectors. Upon receiving the corresponding multimodal environmental data, the classification sub-model identifies flammable materials based on the multimodal environmental data to determine the type of flammable material.

[0089] Specifically, the classification sub-model uses a convolutional neural network model to process the input image data. The image data input to the classification sub-model is represented as follows: Where H and W represent the height and width, respectively, the classification sub-model receives image data and extracts features through convolutional layers. The output of the first convolutional layer... ,in, Represents the convolution kernel weight matrix. This represents the bias term. After multiple convolutional and pooling layers, high-level feature vectors are extracted. , to high-level feature vector The input is compressed into a fully connected layer. Finally, a classification layer outputs the probability distribution of flammable substance categories. Where c represents the category index, and These represent the weights and bias parameters, respectively. The classification sub-model is optimized during the training phase using cross-entropy loss, and the class with the highest probability is output as the recognition result.

[0090] In some embodiments, the target flammable material is processed based on its type and flammability risk decision score, including but not limited to: Example 1: When the type is solid and the flammability risk decision score is higher than the second preset score, control the fire extinguishing device in the vehicle to extinguish the target flammable material.

[0091] The second preset score is greater than the first preset score.

[0092] For example, the spray direction of the fire extinguishing device is determined based on the positional relationship between the fire extinguishing device and the target flammable material, the spray flow rate of the fire extinguishing device is determined based on the flammability risk decision score, and the fire extinguishing device is controlled to extinguish the target flammable material based on the spray direction and spray flow rate.

[0093] For example, the flow rate of extinguishing agent released by the fire extinguishing device. The formula can be expressed as:

[0094] in, Indicates the flammability risk decision score. Indicates jet flow rate, The maximum flow constant, This is the attenuation factor.

[0095] Example 2: When the type is solid and the flammability risk decision score is higher than the second preset score, control the opening of the vehicle's windows and / or doors.

[0096] For example, when a car window is opened by a window motor, the formula for the opening angle of the window can be expressed as:

[0097] in, Indicates the opening angle of the car window. This indicates the maximum opening angle of the car window, and t represents time. For response rate.

[0098] Example 3: When the type is liquid, or when the type is solid and the flammability risk decision score is within the closed interval formed by the first preset score and the second preset score, an alarm prompt will be output.

[0099] The alarm prompts include vehicle-mounted voice alarms and remote alarms from the vehicle owner's mobile terminal.

[0100] For example, activating the text-to-speech (TTS) engine to issue a voice alarm, while simultaneously pushing alarm information to the car owner's mobile phone via Bluetooth Low Energy protocol, the success rate of the push operation is... ,in, d represents the connection rate parameter, and d represents time. This parameter can be used to evaluate and ensure the effective transmission of alarm information under specific connection rate parameters.

[0101] Example 4: When the type is solid and the flammability risk decision score is higher than the second preset score, the target flammable material and its corresponding handling operation are displayed on the vehicle's display device.

[0102] For example, the brightness of the warning user interface (UI) icon displayed on the vehicle's central control screen... ,in For maximum brightness, For the sigmoid function, Indicates the flammability risk decision score. This represents the scaling adjustment factor.

[0103] The operation logs for all the above processing methods are asynchronously written to the SQL database, with event timestamps. ,in, The transmission delay is set as an upper limit to enable periodic self-checking of the parameters of the decision expert model to optimize robustness.

[0104] In the above embodiments, the second preset score is greater than the first preset score, and solids and liquids each have different preset scores. For example, risk levels can be classified based on the second and first preset scores. For instance, a flammability risk decision score higher than the second preset score indicates a high-risk level, a flammability risk decision score between the second and first preset scores indicates a medium-risk level, and a flammability risk decision score lower than the first preset score indicates a low-risk level. Different flammability risk decision scores are mapped to different risk levels, and then the decision engine determines action codes based on the type and risk level of the flammable material. Different action codes correspond to different processing methods, such as decision codes. In this case, the flammable material handling method corresponds to that in Example 1 above.

[0105] In some embodiments, the expert model is a model of a vehicle trained in the cloud and distributed to the target area, where the vehicle is located. The vehicle flammable material handling method provided in this application further includes: collecting the vehicle owner's evaluation of the handling after processing the target flammable material; generating an optimized sample based on multimodal environmental data, cabin status data, and the handling evaluation; and reporting the optimized sample to the cloud.

[0106] The cloud platform updates and redistributes the expert model based on optimized samples reported by vehicles within the target area within a preset period.

[0107] For example, the vehicle collects post-treatment evaluation data of flammable materials, multimodal data, and cabin status data (such as temperature changes, smoke concentration changes, and personnel feedback operations) as optimization samples, which are then sent back to the expert model in the cloud to update the model parameters of the expert module. The model parameters include identification parameters and parameters corresponding to the weight allocation process of the gating network, in order to optimize the accuracy of subsequent decisions.

[0108] For example, real-time acquisition of temperature change values Smoke concentration change rate And the user operation feedback vector f (such as the encoding of key presses or voice commands). This data needs to be preprocessed (such as outlier filtering and standardization) and then integrated into a standardized feature vector x, expressed by the formula:

[0109] Ensure that all input dimensions have a mean of 0 and a variance of 1 to avoid scale mismatch issues affecting the model's convergence.

[0110] Then, the standardized feature vector x is transmitted back to the expert model on the cloud server via a secure encrypted channel (such as Transport Layer Security, TLS). Upon receiving the data, the model first calculates the activation weights of each sub-expert model through a gating network. The formula is expressed as:

[0111] in, These are the weight parameters of the j-th sub-expert model. The weight parameter k of the k-th sub-expert model represents the number of sub-expert models. This represents the standardized feature vector. Meanwhile, each sub-expert model is based on its recognition parameters. Output flammability risk assessment results The final flammability risk decision score will then be obtained. The loss function L during the expert model training process is expressed as:

[0112] in, This represents the flammability risk decision score predicted by the model. The label represents the true flammability risk decision, where N represents the batch data sample size, used to quantify the current decision error. The true flammability risk decision label is based on historical real data or determined manually.

[0113] Next, based on the loss function L, the recognition parameters of the sub-expert model are updated using the backpropagation algorithm. The gradient is calculated using the formula for the gradient of the sub-expert model:

[0114]

[0115] Apply gradient descent to update identification parameters The formula is expressed as:

[0116] Among them, learning rate Set to 0.001 to balance convergence speed and stability, and optimize the ability of each expert model to identify specific states (such as abnormal smoke concentration).

[0117] Simultaneously, the weight allocation strategy parameters of the gating network are updated. The gradient is calculated using the formula for the gradient of a gated network:

[0118]

[0119] Apply gradient descent to update weight allocation strategy parameters The formula is expressed as:

[0120] The contributions of each sub-expert model are more accurately allocated by adjusting weights (e.g., prioritizing the activation of the sub-expert model corresponding to smoke data in cases of high smoke concentration). The updated expert models are then evaluated for decision accuracy on the new dataset, with the loss value reduced. (Preset threshold) Verify the optimization effect and achieve closed-loop adaptive iteration.

[0121] For example, please refer to Figure 3 Another method for handling flammable materials in vehicles provided in this application includes: Step S301: Begin.

[0122] Step S302: Acquire multimodal environment data and cockpit status data.

[0123] Step S303: Expert model processing.

[0124] Step S304: Obtain the flammability risk decision score and the target flammable material type.

[0125] Step S305: The flammable material is solid, and the flammability risk decision score is higher than the second preset score. Execute processing method one.

[0126] For example, one approach includes controlling a fire extinguisher to extinguish the fire and controlling the opening of the vehicle windows.

[0127] Step S306: The flammable material is in liquid form, and the flammability risk decision score is higher than the first preset score. Execute processing method two.

[0128] For example, the second processing method includes remote alarming via Bluetooth or the like.

[0129] Step S307: If the flammability risk decision score is lower than the first preset score, proceed with processing method three.

[0130] For example, the third processing method includes displaying an icon on the central control screen to issue a warning, or not processing at all.

[0131] Step S308: Collect feedback information.

[0132] Optionally, feedback information may include multimodal environmental data, cabin status data, and owner evaluations.

[0133] Step S309: Update the expert model.

[0134] For example, the expert model is updated based on the feedback information, and after the update is completed, step S302 is executed to achieve a continuous flammable material handling process.

[0135] This application also provides a vehicle flammable material handling device; please refer to [link / reference]. Figure 4 This application provides a vehicle flammable material handling device, including an acquisition module 401 and a processing module 402. The acquisition module 401 is used to perform... Figure 2 In the illustrated method, step S201 is executed by processing module 402. Figure 2 The illustrated method includes steps S202 and S203.

[0136] The foregoing mainly describes the solutions provided by the embodiments of this application from the perspective of methods and apparatus. To achieve the above functions, the vehicle flammable material handling device includes hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should readily recognize that, based on the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0137] This application embodiment, based on the aforementioned vehicle flammable material handling method, exemplarily divides the vehicle flammable material handling device into functional modules. For example, the vehicle flammable material handling device may include various functional modules corresponding to each functional division, or two or more functions may be integrated into one processing module. The integrated module can be implemented in hardware or as a software functional module. It should be noted that the module division in this application embodiment is illustrative and only represents one logical functional division; in actual implementation, there may be other division methods.

[0138] This application also provides a computer-readable storage medium storing at least one computer program, which is loaded and executed by a processor to implement the vehicle flammable material handling method provided in the above-described method embodiments.

[0139] Optionally, the computer-readable storage medium may be a non-transitory computer-readable storage medium, such as a read-only memory (ROM), random access memory (RAM), magnetic tape, floppy disk, and optical data storage device.

[0140] This application also provides a computer program product, which includes a computer program or instructions. When the computer program or instructions are executed by a processor, they implement the vehicle flammable material handling method provided in the above-described method embodiments.

[0141] It should be noted that when one or more instructions in the computer-readable storage medium or computer program product are executed by the processor of a computing device, they implement the various processes of the above-described method embodiments and achieve the same technical effects as the above-described methods. To avoid repetition, they will not be described again here.

[0142] Through the above description of the embodiments, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0143] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another apparatus, or some features may be ignored or not executed. Furthermore, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0144] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0145] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, essentially, or the parts that contribute to the prior art, or all or part of the technical solutions, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0146] It should be understood that the application of this application is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims. Those skilled in the art can understand that implementing all or part of the processes of the above embodiments and making equivalent changes according to the claims of this application still fall within the scope of this application.

Claims

1. A method for handling flammable materials in a vehicle, characterized in that, The method for handling flammable materials in vehicles includes: Acquire multimodal environmental data and cabin status data of the vehicle; The multimodal environmental data and the cabin status data are input into an expert model to obtain the target flammable material in the vehicle; the flammability risk decision score of the target flammable material is higher than the first preset score; wherein, the expert model includes multiple sub-expert models corresponding one-to-one with multiple flammable scenarios, and the flammability risk decision score is determined based on the multiple flammability risk assessment results of the target flammable material in multiple flammable scenarios learned by the multiple sub-expert models and the cabin status data; The target flammable material is processed based on its type and flammability risk decision score.

2. The method for handling flammable materials in a vehicle according to claim 1, characterized in that, The cabin status data includes: the number of people in the vehicle, the seating arrangement, and the open / closed status of the windows; The process of determining the flammability risk decision score includes: Based on the number of people in the vehicle, the distribution of seating positions, the opening and closing status of the windows, and the risk assessment performance of the multiple sub-expert models, the risk assessment weights corresponding to the multiple flammable scenarios are determined. Based on the multiple flammability risk assessment results and the risk assessment weights, the flammability risk decision score is determined.

3. The method for handling flammable materials in a vehicle according to claim 2, characterized in that, The flammable scenarios include: flammable scenarios caused by vehicle collisions, flammable scenarios caused by aging electrical equipment, and flammable scenarios caused by the natural environment. The process for determining the multiple flammability risk assessment results includes: Sub-expert models are determined for the flammable scenarios caused by vehicle collisions, flammable scenarios caused by aging electrical equipment, and flammable scenarios caused by the natural environment, respectively. The multimodal environmental data is classified and identified to determine the input data of the sub-expert models corresponding to the flammable scenarios caused by vehicle collisions, the flammable scenarios caused by the aging of electrical equipment, and the flammable scenarios caused by the natural environment. Based on the sub-expert models corresponding to the flammable scenarios caused by vehicle collisions, flammable scenarios caused by electrical equipment aging, and flammable scenarios caused by the natural environment, and the input data of the sub-expert models corresponding to the flammable scenarios caused by vehicle collisions, flammable scenarios caused by electrical equipment aging, and flammable scenarios caused by the natural environment, the multiple flammable risk assessment results are determined.

4. The method for handling flammable materials from a vehicle according to any one of claims 1-3, characterized in that, The expert model also includes a classification sub-model; the classification sub-model is used to determine the type of the target flammable material; the type includes solid and liquid.

5. The method for handling flammable materials from a vehicle according to any one of claims 1-3, characterized in that, The process of processing the target flammable material based on its type and flammability risk decision score includes: If the type is solid and the flammability risk decision score is higher than the second preset score, the fire extinguishing device in the vehicle is controlled to extinguish the target flammable material. The second preset score is greater than the first preset score.

6. The method for handling flammable materials in a vehicle according to claim 5, characterized in that, The method of controlling the fire extinguishing device in the vehicle to extinguish the target flammable material includes: The spray direction of the fire extinguishing device is determined based on the positional relationship between the fire extinguishing device and the target flammable material. Based on the flammability risk decision score, the spray flow rate of the fire extinguishing device is determined; Based on the spray direction and the spray flow rate, the fire extinguishing device is controlled to extinguish the target flammable material.

7. The method for handling flammable materials in a vehicle according to claim 5, characterized in that, The process of processing the target flammable material based on its type and flammability risk decision score further includes: If the type is solid and the flammability risk decision score is higher than the second preset score, control the opening of the vehicle's windows and / or doors.

8. The method for handling flammable materials in a vehicle according to claim 5, characterized in that, The process of processing the target flammable material based on its type and flammability risk decision score further includes: If the type is liquid, or if the type is solid and the flammability risk decision score is within the closed interval formed by the first preset score and the second preset score, an alarm prompt will be output. The alarm notifications include vehicle-mounted voice alarms and remote alarms from the vehicle owner's mobile terminal.

9. The method for handling flammable materials from a vehicle according to any one of claims 1-3, characterized in that, The expert model is a model of vehicles trained in the cloud and distributed to the target area; The vehicle is located in the target area; The method for handling flammable materials in vehicles further includes: After the target flammable material is treated, the vehicle owner's evaluation of the treatment is collected; Based on the multimodal environment data, the cockpit status data, and the processing evaluation, an optimized sample is generated; The optimized samples are reported to the cloud; the cloud updates and re-issues the expert model based on the optimized samples reported by vehicles from the target area within a preset period.

10. The method for handling flammable materials in a vehicle according to claim 9, characterized in that, The optimized samples include the treatment evaluation after flammable material treatment, multimodal data changes, and cabin status data changes; the cloud platform updates the model parameters of the expert model based on the treatment evaluation after flammable material treatment, the multimodal data changes, and the cabin status data changes.

11. A vehicle flammable material handling device, characterized in that, include: The acquisition module is used to acquire multimodal environmental data and cabin status data of the vehicle; The processing module is used to input the multimodal environmental data and the cabin state data into an expert model to obtain the target flammable material in the vehicle; the flammability risk decision score of the target flammable material is higher than a first preset score; wherein, the expert model includes multiple sub-expert models corresponding one-to-one with multiple flammable scenarios, and the flammability risk decision score is determined based on the multiple flammability risk assessment results of the target flammable material in multiple flammable scenarios learned by the multiple sub-expert models and the cabin state data; The processing module is further configured to process the target flammable material based on the type of the target flammable material and the flammability risk decision score.

12. A vehicle, characterized in that, Includes the vehicle flammable material handling device as described in claim 11.

Images