Editorial Type:
Article Category: Research Article
 | 
Online Publication Date: 01 Apr 2018

Robustness and Applicability of a Model-Based Tire State Estimator for an Intelligent Tire

and
Page Range: 105 – 126
DOI: 10.2346/tire.18.460204
Save
Download PDF

ABSTRACT

Tire states can be estimated by measuring the tire contact patch shape as it varies with vertical load, longitudinal and lateral slip, and so on. In this study, a miniature triaxial accelerometer is used to measure the centripetal accelerations at the tire inner liner. A tire state estimator (TSE) algorithm is developed to transform the measured accelerations to actual tire states, in this case vertical load. The approach used for the TSE is the extended Kalman filter (EKF), but an additional peak detection algorithm is used to synchronize the simulation model with the measurement signal before applying the EKF. The simulation model used in the EKF is an empirical model that describes the basic shape of the centripetal acceleration signal. The applicability of the estimator is assessed by considering the accuracy and robustness for several tire operating conditions: vertical load, velocity, inflation pressure, sideslip, camber, and braking. It is concluded that the TSE exhibits accurate vertical load estimation even in cases of varying load and velocity. Further, it is concluded that the vertical load estimation is robust for (pure) camber changes and (pure) longitudinal force disturbances. For relatively high lateral forces as result of sideslip, the estimation error is larger. The current estimator appears to be not robust for inflation pressure changes, but this can be solved by adding an inflation pressure sensor. Similarly, extension of the estimator to estimate lateral force by adding a second accelerometer not only provides an additional state but also adds the possibility of improving the vertical load estimation. Finally, it is demonstrated that the TSE is able to perform in real time and shows fast convergence capabilities for cases in which the initial vertical load and/or sensor position are unknown or when moving away from situations in which the signal-to-noise ratio is poor.

Keywords: intelligent tire; tire state estimator; vertical force estimator; extended Kalman filter; inner liner centripetal acceleration
FIG. 1
FIG. 1

Slip ring assembly.

FIG. 2
FIG. 2

Mounting positions of the triaxial accelerometers at the tire inner liner. Left: setup for benchmarking different sensors on the center line. Right: setup for acquiring the accelerations at three lateral positions.

FIG. 3
FIG. 3

i-Tire setup on TNO's tire test trailer.

FIG. 4
FIG. 4

Schematic representation of the setup for the vertical load TSE and explanation of the notation.

FIG. 5
FIG. 5

Example of centripetal acceleration signal az at different forward velocities and different vertical loads. Dots represent 10 individual revolutions; the solid line is the averaged shape of these.

FIG. 6
FIG. 6

Schematic representation of the basic shape function (solid), being built up by multiplication of an exponential (dashed) and a tangent hyperbolic (dotted) function.

FIG. 7
FIG. 7

Schematic representation of the vertical load tire state estimator.

FIG. 8
FIG. 8

Schematic representation of the VLO using the EKF method.

FIG. 9
FIG. 9

Schematic representation of the PS algorithm.

FIG. 10
FIG. 10

Estimation result for a measurement starting at standstill; vertical load and forward velocity are varied.

FIG. 11
FIG. 11

Peak detection at low velocities for the experiment of Fig. 10.

FIG. 12
FIG. 12

Fz estimation (black) versus actual vertical force (gray) for large (top) and low (bottom) vertical forces. The inflation pressure ranges from low (left), nominal (middle), to high (right).

FIG. 13
FIG. 13

Averaged az shape for different inflation pressures at one fixed vertical load.

FIG. 14
FIG. 14

Brake force (top left) during the test. The resulting Fz estimation error (bottom left) for several vertical loads and the average Fz estimation error as a function of brake force (bottom right).

FIG. 15
FIG. 15

Fz estimation (black) and reference (gray) for different camber angles, −3° (left), 1° (middle), and 5° (right) at different vertical loads of 7.2 kN (top) and 4.8 kN (bottom).

FIG. 16
FIG. 16

Fz estimation error versus the lateral force for several vertical loads.

FIG. 17
FIG. 17

Results of the real-time implementation of the TSE on a dSPACE MicroAutoBox. (Top) Computational load of the real-time process. (Middle) Estimated centripetal acceleration (black) versus the measured centripetal acceleration (gray). (Bottom) Estimated vertical force (black) versus the actual vertical force (gray).

FIG. 18
FIG. 18

Shape of the centripetal tire acceleration az for different values of the lateral force Fy; the measurements are performed with a vertical load of Fz = 7.2 kN.

FIG. 19
FIG. 19

Shape of the centripetal tire acceleration az of the inner (left) and outer (right) accelerometer for different values of lateral force Fy. The measurements are performed with a vertical load of 4.8 kN.

FIG. 20
FIG. 20

Example of extended asymmetrical shape function (black) fitted on averaged centripetal acceleration az (thin gray) of the inner sensor (Fy = −3.5 kN, Fz = 4.8 kN).

Contributor Notes

1 Corresponding author. Integrated Vehicle Safety Department, TNO, Automotive Campus 30, Helmond, 5708 JZ, the Netherlands. Email: antoine.schmeitz@tno.nl
2 Integrated Vehicle Safety Department, TNO, Automotive Campus 30, Helmond, 5708 JZ, the Netherlands. Email: arjan.teerhuis@tno.nl
  • Download PDF