KR-20260064851-A - Method for estimating commercial vehicle weight based on driving data and apparatus therefor

Abstract

The present invention relates to a method for estimating the weight of a commercial vehicle based on driving data and an apparatus for the same. A computing device provided in a commercial vehicle according to one embodiment of the present invention includes a processor that executes commands and a memory that stores said commands. The commands may be implemented to collect real-time vehicle driving data based on the commercial vehicle being ignited (IGNITION ON), identify a starting acceleration trip section by comparing the real-time vehicle driving data with a predetermined starting acceleration trip condition, generate an estimation dataset by performing data preprocessing on the identified starting acceleration trip section, and apply the estimation dataset to a weight estimation calculation logic to estimate the weight corresponding to the identified starting acceleration trip section.

Inventors

• 임은채

Assignees

• 현대자동차주식회사
• 기아 주식회사

Dates

Publication Date

20260508

Application Date

20241029

Claims (20)

1. In a computing device equipped in a commercial vehicle, A processor that executes instructions; It includes memory for storing the above commands, A computing device implemented such that the processor collects real-time vehicle driving data based on the commercial vehicle being ignited (IGNITION ON), identifies a starting acceleration trip section by comparing the real-time vehicle driving data with a predetermined starting acceleration trip condition, generates an estimated dataset by performing data preprocessing on the identified starting acceleration trip section, and estimates a weight corresponding to the identified starting acceleration trip section by applying the estimated dataset to a weight estimation calculation logic.
2. In paragraph 1, A computing device in which the processor calculates a first-order derived variable value based on the real-time vehicle driving data, extracts mean and quantile-based statistical figures based on the first-order derived variable value, calculates a second-order derived variable value based on the statistical figures, and generates the estimated dataset based on the second-order derived variable value.
3. In paragraph 2, The above-mentioned first-order derived variable is a computing device that is a variable capable of tracking the correlation between at least two physical quantities included in the above-mentioned real-time vehicle driving data and the above-mentioned weight.
4. In paragraph 3, A computing device that configures the weight estimation operation logic based on the weights for each first-order derived variable extracted from the pre-trained final weight estimation model.
5. In paragraph 4, A computing device in which the processor collects vehicle driving data by weight corresponding to the commercial vehicle, preprocesses the vehicle data by weight to generate a training dataset by weight, performs machine learning based on the training dataset by weight to generate a weight estimation model, and performs performance verification of the weight estimation model based on a pre-configured test dataset to determine the final weight estimation model.
6. In paragraph 5, A computing device that, the processor calculates a first-order derived variable value by weight based on the vehicle driving data by weight, extracts a statistical value based on the average and quantile by weight based on the first-order derived variable value by weight, calculates a second-order derived variable value by weight based on the statistical value by weight, and generates a learning dataset by weight based on at least one of the first-order derived variable value by weight and the second-order derived variable value by weight.
7. In paragraph 1, A computing device in which the processor updates the estimated weight corresponding to the identified oscillation acceleration trip interval by moving average of at least one weight previously estimated within a predetermined window interval and the estimated weight.
8. In Paragraph 7, A computing device in which the processor provides information regarding the updated weight to a drive and braking system equipped in the commercial vehicle.
9. In paragraph 1, The above real-time vehicle driving data is a computing device comprising at least one of wheel-based speed data, lateral acceleration data, longitudinal acceleration data, yaw rate data, brake pedal position data, electric motor speed data, electric motor torque data, inverter current link data, inverter DC link voltage data, pinch angle data, and steering wheel angle data.
10. In paragraph 2, The above-mentioned first-order derived variable is a computing device comprising at least one of acceleration based on vehicle wheelbase speed (Acc_cal), instantaneous motor output (Mot_pwr_kw), vehicle speed*acceleration index (VA_index), torque/acceleration ratio (Rate_acc_tq), VA index/output ratio (Rate_VA_pwr), acceleration performance index (VAP_Index), acceleration efficiency index (VAE_Index), and driving efficiency index (DE_Index).
11. In a method for estimating weight in commercial vehicles, A step of collecting real-time vehicle driving data based on the fact that the commercial vehicle is ignited (IGNITION ON); A step of identifying a launch acceleration trip section by comparing the above real-time vehicle driving data with a predetermined launch acceleration trip condition; A step of generating an estimated dataset by performing data preprocessing on the identified oscillation acceleration trip intervals; and A step of applying the above estimated dataset to the weight estimation calculation logic to estimate the weight corresponding to the identified oscillation acceleration trip section. A method including
12. In Paragraph 11, The step of generating an estimated dataset by performing the above data preprocessing is: A step of calculating a first derived variable value based on the above real-time vehicle driving data; A step of extracting mean and quantile-based statistical figures based on the above-mentioned first-order derived variable values; Step of calculating the value of a secondary derived variable based on the above statistical figures A method comprising, wherein the estimated dataset is generated based on the secondary derived variable values.
13. In Paragraph 12, A method in which the above-mentioned first-order derived variable is a variable capable of tracking the correlation between at least two physical quantities included in the above-mentioned real-time vehicle driving data and the above-mentioned weight.
14. In Paragraph 13, The above weight estimation operation logic is a method configured based on the weights for each first-order derived variable extracted from the pre-trained final weight estimation model.
15. In Paragraph 14, The above final weight estimation model is, A step of collecting vehicle driving data by weight corresponding to the vehicle type of the commercial vehicle above; A step of generating a training dataset by weight by preprocessing the vehicle data by weight as described above; A step of generating a weight estimation model by performing machine learning based on the above-mentioned weight-specific learning dataset; and Step of determining the final weight estimation model by performing performance verification of the weight estimation model based on a pre-configured test dataset. A method generated through a learning process that includes
16. In paragraph 15, The step of generating a weight-specific training dataset by preprocessing the above weight-specific vehicle data is: A step of calculating a primary derived variable value for each weight based on the vehicle driving data for each weight above; A step of extracting statistical figures based on the average and quantile values for each weight based on the primary derived variable values for each weight; Step of calculating secondary derived variable values by weight based on the above weight-specific statistical figures A method comprising, wherein the weight-specific learning dataset is generated based on at least one of the weight-specific first-order derived variable value and the weight-specific second-order derived variable value.
17. In paragraph 1, A method further comprising the step of updating the estimated weight corresponding to the identified oscillation acceleration trip interval by moving average of at least one weight previously estimated within a predetermined window interval and the estimated weight.
18. In Paragraph 17, A method further comprising the step of transmitting information regarding the updated weight to a drive and braking system equipped in the commercial vehicle.
19. In Paragraph 11, A method in which the above real-time vehicle driving data includes at least one of wheel-based speed data, lateral acceleration data, longitudinal acceleration data, yaw rate data, brake pedal position data, electric motor speed data, electric motor torque data, inverter current link data, inverter DC link voltage data, pinch angle data, and steering wheel angle data.
20. In Paragraph 12, A method in which the above-mentioned first-order derived variable includes at least one of acceleration based on vehicle wheelbase speed (Acc_cal), instantaneous motor output (Mot_pwr_kw), vehicle speed * acceleration index (VA_index), torque/acceleration ratio (Rate_acc_tq), VA index/output ratio (Rate_VA_pwr), acceleration performance index (VAP_Index), acceleration efficiency index (VAE_Index), and driving efficiency index (DE_Index).

Description

Method for estimating commercial vehicle weight based on driving data and apparatus therefor The present invention relates to a vehicle weight estimation technology, and more specifically, to a commercial vehicle weight estimation technology based on driving data utilizing a machine learning model. Commercial vehicles, such as trucks and buses, are characterized by significant weight variations depending on the number of passengers or cargo load. If accurate weight estimation is possible for commercial vehicles, it can be utilized in various ways for vehicle control. For example, efficiency can be improved by optimally controlling drivetrain torque and regenerative torque according to the weight of the commercial vehicle, and braking performance and stability can be ensured by minimizing braking distance through the adjustment of the brake system based on the vehicle's weight. Furthermore, by utilizing the vehicle's weight information, tire wear can be predicted to determine the appropriate replacement time, and the thermal management system can be optimized through the optimal control of the PE (Power Electronic) and BTMS (Battery Thermal Management System) based on the vehicle's weight. Conventionally, vehicle weight has been largely measured using a sensor-based measurement method that uses a separately provided weight-measuring sensor and a physical formula-based estimation method that estimates weight based on a physical formula utilizing physical quantities generated during driving. However, sensor-based measurement methods had limitations in terms of cost because a separate sensor for weight measurement had to be installed on the vehicle, and physical-based estimation methods had the problem that it was impossible to mathematically accurately model the various resistance forces applied to the vehicle while driving. For example, conventional physical-based estimation methods utilize measured values from longitudinal acceleration sensors, which leads to excessively large estimation errors caused by sensor noise in situations such as acceleration, deceleration, and gradients. Furthermore, even when using state estimation algorithms, there were limitations such as unstable initial estimation results depending on the parameter settings required for tuning, or the time required to estimate a stabilized result. For instance, for vehicle types such as city buses, there was a problem in that weight could not be calculated using physical-based estimation methods due to frequent passenger boarding and alighting and driving conditions characterized by acceleration and deceleration profiles. Recently, research applying various deep learning models has been conducted to overcome the limitations of existing vehicle weight measurement or estimation methods; however, general deep learning models have limitations in estimation accuracy because they exclude the physical relationship between driving data and vehicle weight, and they also have the disadvantage that weight estimation performance can significantly degrade for unlearned driving condition data. Furthermore, the need for large-scale training data results in high model complexity, and the excessive demand for computing resources presents limitations in terms of embedding. FIG. 1 is a diagram illustrating the configuration of a commercial vehicle weight estimation system according to one embodiment of the present invention. FIG. 2 is a diagram illustrating the control flow by subject of a commercial vehicle weight estimation system according to one embodiment of the present invention. FIG. 3 is a block diagram illustrating the configuration of a learning engine according to one embodiment of the present invention. FIG. 4 is a block diagram for explaining the detailed structure of a commercial vehicle (10) according to one embodiment of the present invention. FIG. 5 is a flowchart illustrating a method for generating a model for weight estimation in a learner according to an embodiment of the present invention. FIG. 6 is a flowchart illustrating a preprocessing method in a learning device according to an embodiment of the present invention. FIG. 7 is a flowchart illustrating a weight estimation method in a weight estimator according to an embodiment of the present invention. FIG. 8 shows an example of applying a derivative variable of the oscillation acceleration trip reference according to one embodiment of the present invention. FIG. 9 shows an example of applying a statistical-based derived variable according to one embodiment of the present invention. FIG. 10 is a table in which a derived variable and its operation formula are defined according to one embodiment of the present invention. FIG. 11 is a diagram illustrating a method for extracting average/quantile data per trip according to an embodiment of the present invention. FIG. 12 is a diagram illustrating a polynomial linear regression model, which is a machine learning model according to one embodiment