Abstract
This paper considers mass estimation of passenger cars from vehicle longitudinal dynamics. The primary novelty lies in the work towards mass estimation without data from the controller area network (CAN-bus), instead using only inertial measurement unit (IMU) and global navigation satellite system (GNSS) data. The work first considers improvements related to the estimation of road gradients from the vehicle pitch angles, by compensating for (1) pitching from vehicle loading and (2) inertial forces during acceleration and braking. (1) is estimated by the ratio of vertical and longitudinal vehicle speeds, computed by a Kalman filter based sensor fusion of IMU and GNSS data. (2) is estimated by two linear models in the longitudinal vehicle acceleration a: one for a > 0 (accelerating) and one for a < 0 (braking). The model coefficients are estimated from drives on a road where the road gradient first is estimated from drives at constant speed, which is then used as target values for other drives, subject to non-zero acceleration levels. The two compensations (1) and (2) each contribute to noticeable increases in vehicle mass. The mass is estimated within of the actual, across drives with three different load levels, estimated using an approach with access to vehicle CAN-bus. Removing the CAN-bus access introduces challenges related to the lack of information on transmission gear and engine torque. Methods for estimating model parameters and the engine torque curve are presented. Gear and mass estimates are carried out on batches of samples during acceleration segments, rather than continuously to simplify data processing. The current gear is estimated from spectral data of the vehicle acceleration during segments, by considering the relative power content of engine frequencies in different gears. The proposed method provides accurate gear estimates during acceleration segments, both with constant and varying gears. Mass estimation is carried out using the estimated gears and with the full load engine torque curve. The result is that accurate mass estimation is achieved during segments where full load is applied. For other segments, however, the mass is overestimated. The proposed method allows the driver to obtain on-demand mass estimates, by applying full engine load when estimates are desired.
Original language | English |
---|---|
Article number | 111907 |
Journal | Mechanical Systems and Signal Processing |
Volume | 223 |
Number of pages | 25 |
ISSN | 0888-3270 |
DOIs | |
Publication status | Published - 2024 |
Keywords
- Experimental validation
- Longitudinal dynamics
- Mass estimation
- Sensor fusion
- Vehicular dynamics