Vehicle Mass Estimation Using Modified Kalman Filter
Solution Overview
Problem
Conventional vehicle mass estimation methods using wheel speed signals are prone to instability and inaccuracies due to outliers and varying vehicle conditions, leading to false alarms in tire pressure monitoring systems, especially when the vehicle is under non-reference conditions such as turning or carrying heavy loads.
Innovation Solution
A robust vehicle mass estimation method that adjusts the observation process to treat deviating values as outliers by increasing dispersion, thereby reducing their contribution to parameter updates, using a modified Kalman filter with gamma distribution-based Kalman gain calculation and confidence interval adjustments to improve estimation robustness and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a conventional Kalman filter is used for real-time vehicle mass estimation, then the estimation can be processed in real-time, but the estimation value becomes unstable when outliers occur frequently
Solution Approach 1:
The patent implements a dynamic outlier detection mechanism that adapts the estimation process based on residual analysis. When outliers are detected through statistical testing of prediction residuals, the system dynamically adjusts by rejecting suspicious data points before they corrupt the Kalman filter state estimates, thereby maintaining stability while preserving real-time processing capability
Solution Approach 2:
The system incorporates feedback through residual analysis where the difference between predicted and actual measurements is continuously monitored. When residuals exceed threshold values indicating potential outliers, the system feeds back this information to reject the corresponding measurements, preventing outlier propagation and maintaining estimation reliability in real-time
2Quantity of substance
If all wheel speed data is used for mass estimation, then more data is available for calculation, but the estimation accuracy decreases due to outliers and varying vehicle conditions
Solution Approach 1:
The patent applies local quality by implementing condition-specific data selection criteria. Different running conditions (acceleration, deceleration, turning, straight-line motion) are identified and evaluated individually. Only data meeting quality thresholds for the current condition are used in estimation, ensuring high precision by selecting locally appropriate data subsets rather than uniformly processing all available data
Solution Approach 2:
The system changes parameters dynamically by adjusting data acceptance criteria based on vehicle operating conditions. Thresholds for acceptable wheel speed deviations, acceleration rates, and other parameters are modified according to the detected running state, allowing the system to maintain high estimation accuracy across varying conditions by adapting which data points qualify for inclusion
3Device complexity
If the DLR method is used for tire pressure detection, then simple calculation processing is required, but false alarms occur during vehicle turning or acceleration
Solution Approach 1:
The patent dynamically adjusts the detection threshold and data selection criteria based on real-time identification of vehicle running conditions. When turning or acceleration is detected through wheel speed differential analysis, the system adapts by excluding data from affected wheels or adjusting threshold values, thereby preventing false alarms while maintaining the relatively simple DLR calculation framework
Solution Approach 2:
The system changes detection parameters adaptively by modifying threshold values and data inclusion criteria according to the detected running state. During turning maneuvers or acceleration events, parameters such as acceptable wheel speed differences and minimum confidence thresholds are adjusted to account for expected variations, preventing false pressure loss detections while preserving the simplicity of the DLR method
Data Source
AI summary
A vehicle mass estimation apparatus for estimating the mass of a vehicle based on rotation information of tires attached to the vehicle. The apparatus includes a rotation speed information detection means for periodically detecting tire rotation speed information of the respective wheels of the vehicle; a rotation acceleration information calculation means for calculating tire rotation acceleration information based on the rotation speed information obtained by the rotation speed information detection means; a driving force calculation means for calculating a driving force of the vehicle based on axle shaft torque information of the vehicle; and a vehicle mass estimation means for estimating the mass of the vehicle as a regression coefficient when the rotation acceleration information and the driving force information are subjected to linear regression. The vehicle mass estimation means calculates, with regard to an error between driving force information calculated based on the estimated mass and the rotation acceleration information and the driving force information calculated by the driving force calculation means, a ratio between the dispersion of the error obtained one time before and an error at the current moment; and calculates an update adjustment parameter by adding a stabilization parameter to thus obtained ratio, and uses the calculated update adjustment parameter to calculate a Kalman gain so that the mass is sequentially estimated so as to reduce an update width of the regression coefficient when the error ratio is higher.