Intelligent electric commercial vehicle mass estimation method and device based on pavement conditions

Abstract

The invention discloses an intelligent electric commercial vehicle quality estimation method and device based on road conditions, and the method comprises the steps: carrying out the data collection of an image in front of a vehicle, a clock source signal and a vehicle-mounted CAN signal, carrying out the signal preprocessing of a vehicle-mounted CAN bus original signal, carrying out the recognition of the road conditions through a lightweight deep learning algorithm, and carrying out the recognition of the road conditions. And mass estimation triggering discrimination under different road surface conditions is performed, and dynamic-based vehicle mass estimation is performed. According to the method, different road conditions can be effectively identified, a mass estimation triggering mechanism can be dynamically adjusted, and the mass estimation precision of the electric commercial vehicle under the complex road condition is remarkably improved by combining a dynamic model and road slope correction.

Description

Technical Field

[0001] This invention relates to the field of quality estimation for intelligent electric commercial vehicles, and more particularly to a method and apparatus for quality estimation of intelligent electric commercial vehicles based on road conditions. Background Technology

[0002] With the continuous development of intelligent electric commercial vehicle technology, real-time vehicle mass estimation has become a key aspect of improving motion control accuracy and energy efficiency management. In the commercial vehicle sector, significant load variations, especially the frequent switching between empty and fully loaded conditions, place higher demands on the accuracy and real-time performance of mass estimation. Among related technologies, mass estimation methods based on dynamic models achieve low-cost mass estimation without additional sensors through onboard CAN bus signals. Simultaneously, by combining inertial navigation systems and image perception technology, a multi-technology collaborative system is constructed, encompassing key aspects such as data acquisition, signal processing, road surface recognition, and dynamic modeling.

[0003] However, existing dynamics-based mass estimation methods typically do not adequately consider the impact of road adhesion conditions on vehicle dynamic parameters. This is particularly true on low-adhesion surfaces such as ice, snow, and wet surfaces, where wheel slippage is common, leading to significant discrepancies between the torque signals collected by the vehicle's CAN bus and the actual ground forces. Specifically, key input parameters such as longitudinal acceleration, driving torque, and braking torque are particularly distorted during the start-up phase, significantly reducing the accuracy of mass estimation. Traditional methods can achieve mass estimation errors exceeding 30% under complex conditions, severely impacting the formulation and execution of vehicle control strategies, especially in special application scenarios such as winter transportation, posing significant safety hazards and energy efficiency losses. Summary of the Invention

[0004] The present invention aims to at least partially solve one of the technical problems in the related art.

[0005] This invention proposes a method for estimating the quality of intelligent electric commercial vehicles based on road conditions.

[0006] Another objective of this invention is to provide an intelligent electric commercial vehicle quality estimation device based on road conditions.

[0007] To achieve the above objectives, this invention proposes a method for estimating the quality of intelligent electric commercial vehicles based on road conditions, comprising: S1 synchronously acquires images of the vehicle's front, clock signals from the inertial navigation system, and signals from the vehicle's CAN bus, and adds timestamps to each data point to achieve data alignment; it also performs signal preprocessing on the raw signals from the vehicle's CAN bus. S2, based on a lightweight deep learning model, identifies the road surface type of the acquired images and maps the identification results to the road surface state of easy or difficult to slip. S3 dynamically sets the quality estimation trigger conditions based on the identified road conditions and vehicle signals, and starts the quality estimation process when the conditions are met; S4, after triggering mass estimation, uses the iterative least squares method to estimate the vehicle mass based on the driving force-driving resistance dynamic equation including road slope correction.

[0008] The intelligent electric commercial vehicle quality estimation method based on road conditions in this embodiment of the invention may also have the following additional technical features: In one embodiment of the present invention, S1 includes: Images of the front of the vehicle are captured via a USB interface at a frequency of 30 frames per second. The camera is positioned below the rearview mirror inside the windshield. The camera angle is adjusted so that the lowest point of the field of view is at the edge of the hood, and the road surface occupies more than three-fifths of the entire image. The inertial navigation system uses the GPRMC protocol to acquire clock source signals and pitch angular velocity at a frequency of 100Hz, and uses UTC time as the time reference to add timestamps to each frame of image and CAN bus signal to achieve synchronous alignment of multi-source data.

[0009] In one embodiment of the present invention, the step of identifying road surface type based on a lightweight deep learning model of the acquired image and mapping the identification result to a slippery or non-slippery road surface state includes: The DeeplabV3+ lightweight deep learning model was used for road surface type recognition. The training data of the model covered different weather conditions such as sunny, rainy and snowy days, different lighting conditions such as daytime, nighttime and tunnel strong light, and different road surface conditions such as dry, wet and icy. Pixel-level annotation was performed using the CVAT tool. The labeled images are divided into training, validation and test sets according to a preset ratio. Each image is a batch and is input into the DeeplabV3+ model for training. The training is iterated until the accuracy requirement is met.

[0010] In one embodiment of the present invention, the step of dynamically setting quality estimation trigger conditions based on the identified road surface conditions and vehicle signals, and initiating the quality estimation process when the conditions are met, further includes: If the identification result indicates a non-slip surface, and the gear is forward with longitudinal acceleration greater than [value missing], then [the following is a possible interpretation of the original text:] Mass estimation is triggered when the vehicle speed is greater than 5 km / h, the absolute value of the steering wheel angle is less than 5°, and the ABS and TCS are disabled. If the identification result indicates a slippery road surface, and the gear is forward with longitudinal acceleration greater than... Mass estimation is triggered when the vehicle speed is greater than 20 km / h, the absolute value of the steering wheel angle is less than 5°, and the ABS and TCS are disabled.

[0011] In one embodiment of the present invention, it further includes: If the mass estimation is not triggered, the previous mass estimation result remains unchanged; if the vehicle fails to complete a valid mass estimation after power-on, the vehicle's unloaded mass is used as the mass estimation result. Set a quality estimation counter. The counter counts at 10ms intervals. The minimum number of counts for quality estimation is 200, and the maximum number of counts is 500. The counter is reset to zero after one quality estimation is completed.

[0012] To achieve the above objectives, another aspect of the present invention provides an intelligent electric commercial vehicle quality estimation device based on road conditions, comprising: The data synchronization acquisition module is used to synchronously acquire images of the vehicle's front, clock signals from the inertial navigation system, and signals from the vehicle's CAN bus, and to add timestamps to each data to achieve data alignment; and to perform signal preprocessing on the raw signals from the vehicle's CAN bus. The road surface condition recognition module is used to identify the road surface type of the acquired images based on a lightweight deep learning model, and to map the recognition results to the road surface condition of easy or difficult to slip. The quality estimation triggering module is used to dynamically set the quality estimation triggering conditions based on the identified road conditions and vehicle signals, and to start the quality estimation process when the conditions are met. The mass estimation calculation module is used to estimate the vehicle mass using the iterative least squares method based on the driving force-driving resistance dynamic equation that includes road slope correction after mass estimation is triggered.

[0013] The intelligent electric commercial vehicle quality estimation method and device based on road conditions of this invention effectively improves the quality estimation accuracy of electric commercial vehicles on complex road surfaces (especially low adhesion coefficient road surfaces) by integrating forward image recognition and on-board signal processing, and combining an adaptive quality estimation triggering mechanism under different road surface conditions.

[0014] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0015] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1This is a flowchart of a method for estimating the quality of intelligent electric commercial vehicles based on road conditions, according to an embodiment of the present invention. Figure 2 This is a structural diagram of an intelligent electric commercial vehicle quality estimation device based on road conditions according to an embodiment of the present invention. Detailed Implementation

[0016] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0017] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0018] The following description, with reference to the accompanying drawings, outlines a method and apparatus for estimating the quality of intelligent electric commercial vehicles based on road conditions, according to embodiments of the present invention.

[0019] Figure 1 This is a flowchart of a method for estimating the quality of intelligent electric commercial vehicles based on road conditions according to an embodiment of the present invention, such as... Figure 1 As shown, it includes: S1 synchronously acquires images of the vehicle's front, clock signals from the inertial navigation system, and signals from the vehicle's CAN bus, and adds timestamps to each data point to achieve data alignment; it also performs signal preprocessing on the raw signals from the vehicle's CAN bus.

[0020] Understandably, this invention acquires images of the vehicle's front at a frequency of 30 frames per second via a USB interface. The camera is positioned below the rearview mirror inside the windshield, and the camera's viewing angle is adjusted so that the lowest point of the field of view is at the edge of the hood, and the road surface occupies more than three-fifths of the entire image. An inertial navigation system is used to acquire clock source signals and pitch angular velocity at a frequency of 100Hz via the GPRMC protocol, and UTC time is used as the time reference to add timestamps to each frame of image and CAN bus signal to achieve synchronous alignment of multi-source data.

[0021] In one embodiment of the present invention, vehicle speed, vehicle longitudinal acceleration, drive motor torque, braking torque, steering wheel angle, gear signal, ABS function enable signal, and TCS function enable signal are acquired in real time via the vehicle CAN bus. Visual images of the front of the vehicle captured by an onboard camera, with a resolution of [resolution missing]. The acquisition frequency is 30 frames per second. The camera is positioned below the rearview mirror inside the windshield. The camera angle is adjusted so that the lowest point of the field of view is at the edge of the hood, and the road surface occupies more than three-fifths of the entire image, ensuring that the road surface area in the acquired image exceeds 50 meters. The inertial navigation system uses the GPRMC protocol to acquire clock source signals and pitch angular velocity at a frequency of 100Hz. The entire system uses the clock source signal as the time reference for recording timestamps. The acquired images are transmitted to the domain controller via USB, and the inertial navigation system transmits them to the domain controller via serial port. The vehicle speed, longitudinal acceleration, drive motor torque, braking torque, steering wheel angle, gear signal, ABS function enable signal, and TCS function enable signal are transmitted to the domain controller via the CAN bus interface. The domain controller assigns the UTC time acquired by the inertial navigation system to each frame of image and records the timestamp of the CAN bus signal, providing a reference for aligning the subsequent road recognition results with the CAN bus signal.

[0022] In one embodiment of the present invention, the original signals of the vehicle CAN bus are preprocessed. A second-order Butterworth low-pass filter is used to measure vehicle speed. Data filtering is performed with a cutoff frequency of 2Hz. , , , , speed v The filtering formula is: (1) Longitudinal acceleration is measured using a Kalman filter. Perform data filtering, let , , dt Take 0.01, , , I The identity matrix represents the longitudinal acceleration. The filtering formula is: (2) (3) (4) A sliding window weighted average filter is used to measure the torque of the drive motor. Perform data filtering to drive motor torque. The filtering formula is: (5) Where the weight is , , , , ; Using a second-order filter to adjust the braking torque Perform data filtering, braking torque The filtering formula is: (6) in, , , The frequency corresponding to the maximum amplitude in the range of 5-20Hz, i.e. .

[0023] S2 uses a lightweight deep learning model to identify road surface types in the acquired images and maps the identification results to road surface conditions that are either slippery or non-slippery.

[0024] Specifically, the DeeplabV3+ lightweight deep learning model is used for road surface type recognition. The training data of the model covers different weather conditions such as sunny, rainy, and snowy days, different lighting conditions such as daytime, nighttime, and tunnel strong light, and different road surface conditions such as dry, wet, and icy / snowy. CVAT tool is used for pixel-level annotation. The annotated images are divided into training set, validation set, and test set according to a preset ratio. Each image is a batch and is input into the DeeplabV3+ model for training. The training is iterated until the accuracy requirement is met.

[0025] The training data collection and processing included collecting images covering different weather conditions (sunny, rainy, snowy), different lighting conditions (daytime, nighttime, tunnel light), and different road surface conditions (dry, wet, icy, snowy), ensuring at least 500 valid images for each road surface condition. Pixel-level annotations were performed using CVAT tools, with eight road surface types: dry asphalt road surface, wet asphalt road surface, dry cement road surface, wet cement road surface, loose snow road surface, compacted snow road surface, ice film road surface, and background. The annotated images were then divided into training, validation, and test sets in a 7:2:1 ratio. The DeeplabV3+ model was trained; each image in the training set was processed. Zhang was in a batch, among which ,and n The value is an integer. Each batch is input into the network for training, and the process is repeated 1000 times. After training, the images in the test set are used as input. If the recognition accuracy of each image is greater than 90%, the model is successfully trained. Otherwise, the process is repeated 1000 times based on the previous training, and so on until the accuracy requirement is met. The trained model is invoked on the domain controller to identify road conditions; the trained DeeplabV3+ lightweight deep learning network model is invoked on the domain controller to classify road types, and the obtained road types, excluding the background, are mapped and output according to the correspondence provided in Table 1.

[0026] Table 1

[0027] S3 dynamically sets quality estimation trigger conditions based on the identified road conditions and vehicle signals, and starts the quality estimation process when the conditions are met.

[0028] S4, after triggering mass estimation, uses the iterative least squares method to estimate the vehicle mass based on the driving force-driving resistance dynamic equation including road slope correction.

[0029] The method further includes: when no quality estimation is triggered, keeping the previous quality estimation result unchanged; if the vehicle fails to complete a valid quality estimation after power-on, using the vehicle's unloaded quality as the quality estimation result; setting a quality estimation counter, which counts at 10ms intervals, with a minimum of 200 counts and a maximum of 500 counts, and clearing the counter after completing one quality estimation.

[0030] Specifically, the system uses timestamp information to select road surface recognition results, filtered longitudinal acceleration, vehicle speed, and gear information from the same moment, and then uses this information to trigger the quality estimation judgment. If the road surface recognition result indicates a non-slip surface, and the gear is forward with longitudinal acceleration greater than... Vehicle speed greater than 5 km / h, absolute value of steering wheel angle less than When the ABS and TCS enable bits are disabled, quality estimation is triggered. If the road surface identification result indicates a slippery road surface, and the gear is forward with longitudinal acceleration greater than... Vehicle speed greater than 20km / h, absolute value of steering wheel angle less than When the ABS and TCS enable bits are disabled, quality estimation is triggered. After triggering quality estimation, a counter with an interval of 10ms is used to count. The minimum number of counts for quality estimation is 200, and the maximum number of counts is 500. The counter is reset to zero after one quality estimation is completed. In other cases, the previous mass estimation results are retained. If a valid estimation cannot be completed after the vehicle is powered on, the vehicle's unloaded mass is used as the mass estimation result.

[0031] When the conditions for quality estimation in the working condition judgment are met, quality estimation is performed; The mass estimation model obtains an accurate estimate of the vehicle's mass through the driving force-driving resistance dynamic equation.

[0032] Considering the vehicle's motion during the initial stage as a single longitudinal acceleration / deceleration motion, according to Newton's second law, the dynamic equation for the longitudinal motion driving process is: (7) In the formula, m For vehicle quality, For the longitudinal acceleration of the vehicle, For the torque of each drive motor, The transmission ratio of each electric drive system, This is an equivalent coefficient that includes system transmission losses and tire characteristics. For the braking torque of each wheel, Let the rolling radius of each wheel be , f This is the rolling resistance coefficient between the tire and the ground. g It is the acceleration due to gravity. This refers to the vehicle's air resistance coefficient. A w The vehicle's frontal area. The road gradient is obtained by integrating the pitch angular velocity of the inertial navigation system. v The speed of the vehicle.

[0033] According to the above formula, we can obtain: (8) Let the estimated parameter vector be: (9) Let the coefficient vector be: (10) Simultaneously set: (11) The iterative least squares method is used to estimate the mass of the vehicle during the start-up process, letting... I The identity matrix is ​​represented by the iterative formula: (12) (13) (14) In summary, the estimated value of the vehicle mass is obtained through iterative calculation. .

[0034] To achieve the above embodiments, such as Figure 2 As shown, this embodiment also provides an intelligent electric commercial vehicle quality estimation device 10 based on road conditions, including: The data synchronization acquisition module 100 is used to synchronously acquire the image in front of the vehicle, the clock source signal of the inertial navigation system and the vehicle CAN bus signal, and add timestamps to each data to achieve data alignment. The road surface condition recognition module 200 is used to identify the road surface type of the acquired images based on a lightweight deep learning model, and map the recognition results to the road surface condition of easy-to-skid or non-slip-prone surfaces. The quality estimation trigger module 300 is used to dynamically set the quality estimation trigger conditions based on the identified road surface conditions and vehicle signals, and to start the quality estimation process when the conditions are met. The mass estimation calculation module 400 is used to estimate the vehicle mass using the iterative least squares method based on the driving force-driving resistance dynamic equation including road slope correction after the mass estimation is triggered.

[0035] Furthermore, the data synchronization acquisition module is also used for: Images of the front of the vehicle are captured via a USB interface at a frequency of 30 frames per second. The camera is positioned below the rearview mirror inside the windshield. The camera angle is adjusted so that the lowest point of the field of view is at the edge of the hood, and the road surface occupies more than three-fifths of the entire image. The inertial navigation system uses the GPRMC protocol to acquire clock source signals and pitch angular velocity at a frequency of 100Hz, and uses UTC time as the time reference to add timestamps to each frame of image and CAN bus signal to achieve synchronous alignment of multi-source data.

[0036] Furthermore, the road surface condition recognition module is also used for: The DeeplabV3+ lightweight deep learning model was used for road surface type recognition. The training data of the model covered different weather conditions such as sunny, rainy and snowy days, different lighting conditions such as daytime, nighttime and tunnel strong light, and different road surface conditions such as dry, wet and icy. Pixel-level annotation was performed using the CVAT tool. The labeled images are divided into training, validation, and test sets in a 7:2:1 ratio. Each image is a batch and is input into the DeeplabV3+ model for training. The training is iterated until the accuracy requirement is met.

[0037] Furthermore, the quality estimation trigger module is also used for: If the identification result indicates a non-slip surface, and the gear is forward with longitudinal acceleration greater than [value missing], then [the following is a possible interpretation of the original text:] Mass estimation is triggered when the vehicle speed is greater than 5 km / h, the absolute value of the steering wheel angle is less than 5°, and the ABS and TCS are disabled. If the identification result indicates a slippery road surface, and the gear is forward with longitudinal acceleration greater than... Mass estimation is triggered when the vehicle speed is greater than 20 km / h, the absolute value of the steering wheel angle is less than 5°, and the ABS and TCS are disabled.

[0038] Furthermore, it also includes: The quality retention module is used to maintain the previous quality estimation result when the quality estimation is not triggered; if the vehicle fails to complete a valid quality estimation after power-on, the vehicle's unloaded quality is used as the quality estimation result. The counting control module is used to set the quality estimation counter. The counter counts at 10ms intervals, with a minimum of 200 counts and a maximum of 500 counts. The counter is reset to zero after one quality estimation is completed.

[0039] The intelligent electric commercial vehicle quality estimation device based on road conditions in this invention collects visual images of the vehicle's front from the onboard camera, clock signals from the inertial navigation system, and vehicle speed signals from the onboard CAN bus. It preprocesses the raw onboard CAN bus signals using various filters, identifies road conditions using the DeeplabV3+ lightweight deep learning algorithm, performs quality estimation trigger discrimination under different road conditions, and performs dynamic quality estimation. This method effectively improves the accuracy of dynamic quality estimation for electric commercial vehicles by identifying road conditions in advance and setting quality estimation trigger conditions under different road conditions.

[0040] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0041] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

Claims

1. A method for estimating the quality of intelligent electric commercial vehicles based on road conditions, characterized in that, include: S1 synchronously acquires images of the vehicle's front, clock signals from the inertial navigation system, and signals from the vehicle's CAN bus, and adds timestamps to each data point to achieve data alignment. And perform signal preprocessing on the raw signals of the vehicle CAN bus; S2, based on a lightweight deep learning model, identifies the road surface type of the acquired images and maps the identification results to the road surface state of easy or difficult to slip. S3 dynamically sets the quality estimation trigger conditions based on the identified road conditions and vehicle signals, and starts the quality estimation process when the conditions are met; S4, after triggering mass estimation, uses the iterative least squares method to estimate the vehicle mass based on the driving force-driving resistance dynamic equation including road slope correction.

2. The method as described in claim 1, characterized in that, S1 includes: Images of the front of the vehicle are captured via a USB interface at a frequency of 30 frames per second. The camera is positioned below the rearview mirror inside the windshield. The camera angle is adjusted so that the lowest point of the field of view is at the edge of the hood, and the road surface occupies more than three-fifths of the entire image. The inertial navigation system uses the GPRMC protocol to acquire clock source signals and pitch angular velocity at a frequency of 100Hz, and uses UTC time as the time reference to add timestamps to each frame of image and CAN bus signal to achieve synchronous alignment of multi-source data.

3. The method as described in claim 1, characterized in that, The process of identifying road surface types from acquired images using a lightweight deep learning model and mapping the identification results to slippery or non-slippery road surface conditions includes: The DeeplabV3+ lightweight deep learning model was used for road surface type recognition. The training data of the model covered different weather conditions such as sunny, rainy and snowy days, different lighting conditions such as daytime, nighttime and tunnel strong light, and different road surface conditions such as dry, wet and icy. Pixel-level annotation was performed using the CVAT tool. The labeled images are divided into training, validation and test sets according to a preset ratio. Each image is a batch and is input into the DeeplabV3+ model for training. The training is iterated until the accuracy requirement is met.

4. The method as described in claim 1, characterized in that, The step of dynamically setting quality estimation trigger conditions based on the identified road surface conditions and vehicle signals, and initiating the quality estimation process when the conditions are met, further includes: If the identification result indicates a non-slip surface, and the gear is forward with longitudinal acceleration greater than [value missing], then [the following is a possible interpretation of the original text:] Mass estimation is triggered when the vehicle speed is greater than 5 km / h, the absolute value of the steering wheel angle is less than 5°, and the ABS and TCS are disabled. If the identification result indicates a slippery road surface, and the gear is forward with longitudinal acceleration greater than... Mass estimation is triggered when the vehicle speed is greater than 20 km / h, the absolute value of the steering wheel angle is less than 5°, and the ABS and TCS are disabled.

5. The method as described in claim 1, characterized in that, Also includes: When quality estimation is not triggered, the previous quality estimation results remain unchanged. If the vehicle fails to complete an effective mass estimation after being powered on, the vehicle's unloaded mass will be used as the mass estimation result. Set a quality estimation counter. The counter counts at 10ms intervals. The minimum number of counts for quality estimation is 200, and the maximum number of counts is 500. The counter is reset to zero after one quality estimation is completed.

6. A smart electric commercial vehicle quality estimation device based on road conditions, characterized in that, include: The data synchronization acquisition module is used to synchronously acquire images of the vehicle in front, clock signals from the inertial navigation system, and signals from the vehicle's CAN bus, and to add timestamps to each data to achieve data alignment. And perform signal preprocessing on the raw signals of the vehicle CAN bus; The road surface condition recognition module is used to identify the road surface type of the acquired images based on a lightweight deep learning model, and to map the recognition results to the road surface condition of easy or difficult to slip. The quality estimation triggering module is used to dynamically set the quality estimation triggering conditions based on the identified road conditions and vehicle signals, and to start the quality estimation process when the conditions are met. The mass estimation calculation module is used to estimate the vehicle mass using the iterative least squares method based on the driving force-driving resistance dynamic equation that includes road slope correction after mass estimation is triggered.

7. The apparatus as claimed in claim 6, characterized in that, The data synchronization acquisition module is also used for: Images of the front of the vehicle are captured via a USB interface at a frequency of 30 frames per second. The camera is positioned below the rearview mirror inside the windshield. The camera angle is adjusted so that the lowest point of the field of view is at the edge of the hood, and the road surface occupies more than three-fifths of the entire image. The inertial navigation system uses the GPRMC protocol to acquire clock source signals and pitch angular velocity at a frequency of 100Hz, and uses UTC time as the time reference to add timestamps to each frame of image and CAN bus signal to achieve synchronous alignment of multi-source data.

8. The apparatus as claimed in claim 6, characterized in that, The road surface condition recognition module is also used for: The DeeplabV3+ lightweight deep learning model was used for road surface type recognition. The training data of the model covered different weather conditions such as sunny, rainy and snowy days, different lighting conditions such as daytime, nighttime and tunnel strong light, and different road surface conditions such as dry, wet and icy. Pixel-level annotation was performed using the CVAT tool. The labeled images are divided into training, validation and test sets according to a preset ratio. Each image is a batch and is input into the DeeplabV3+ model for training. The training is iterated until the accuracy requirement is met.

9. The apparatus as claimed in claim 6, characterized in that, The quality estimation triggering module is also used for: If the identification result indicates a non-slip surface, and the gear is forward with longitudinal acceleration greater than [value missing], then [the following is a possible interpretation of the original text:] Mass estimation is triggered when the vehicle speed is greater than 5 km / h, the absolute value of the steering wheel angle is less than 5°, and the ABS and TCS are disabled. If the identification result indicates a slippery road surface, and the gear is forward with longitudinal acceleration greater than... Mass estimation is triggered when the vehicle speed is greater than 20 km / h, the absolute value of the steering wheel angle is less than 5°, and the ABS and TCS are disabled.

10. The apparatus as claimed in claim 6, characterized in that, Also includes: The quality preservation module is used to maintain the previous quality estimation results unchanged when the quality estimation is not triggered; If the vehicle fails to complete an effective mass estimation after being powered on, the vehicle's unloaded mass will be used as the mass estimation result. The counting control module is used to set the quality estimation counter. The counter counts at 10ms intervals, with a minimum of 200 counts and a maximum of 500 counts. The counter is reset to zero after one quality estimation is completed.

Images