Lateral Force Prediction using Gaussian Process Regression for Intelligent Tire Systems

Amir Khajepour; Bruno Henrique Groenner Barbosa; Hassan Askari; Nan Xu

REVIEW 4 major objections 5 minor 41 references

Lateral Force Prediction using Gaussian Process Regression for Intelligent Tire Systems

T0 review · 4 major / 5 minor · reviewed 2026-08-27 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A Gaussian-process regressor trained on one tri-axial accelerometer's contact-patch readings predicts tire lateral force with 7.91% NRMSE and R = 0.99 on held-out maneuvers, and returns uncertainty bands that cover the measured forces.

desk verdict Solid empirical GPR soft sensor for lateral tire force, but the abstract oversells high-slip accuracy; still worth a serious referee. read the letter →

arxiv 2009.12463 v1 pith:2RXNSNWZ submitted 2020-09-25 eess.SP cs.LG

Bruno Henrique Groenner Barbosa , Nan Xu , Hassan Askari , Amir Khajepour This is my paper · ORCID

classification eess.SPcs.LG

keywords intelligenttirelateralforceGaussianprocessregressionmachinelearningcontactpatchaccelerometeruncertaintyquantificationvehiclecontrol

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper claims that the lateral force a tire generates — the quantity vehicle-stability controllers most need but usually must estimate indirectly — can be read directly from how the contact patch deforms as it rolls past a single three-axis accelerometer glued to the inner liner of the tire. Feeding a 42-point sample of those accelerations into a Gaussian process regressor, the authors report 7.91% normalized root-mean-square error and a 0.99 correlation on maneuvers the model never trained on. The reason for choosing a Gaussian process over a plain black-box regressor is that it returns a stated uncertainty alongside each prediction, and the paper shows that uncertainty widening precisely where the task is hardest: slip angles above six degrees and high vertical loads. If the result holds, an instrumented tire can serve as its own lateral-force sensor for control design, sidestepping indirect vehicle-model-based estimation that typically struggles at high slip angles.

What carries the argument

The load-bearing object is the Gaussian process regressor with the Matérn covariance kernel $\kappa_{v=3/2}(\tau) = \sigma_f^2\left(1 + \frac{\sqrt{3}\tau}{\ell}\right)\exp\left(-\frac{\sqrt{3}\tau}{\ell}\right)$ with automatic relevance determination: the posterior mean, $K(x_*, X)[K(X, X) + \sigma_\epsilon I]^{-1}y$, produces the force prediction and the posterior variance produces the uncertainty band, with hyperparameters $\{\sigma_f^2, \ell\}$ fitted by gradient ascent on the log marginal likelihood. The regressor's input is a 42-point feature vector: 14 angular samples per acceleration axis, taken at 5-degree intervals across the 70-degree contact-patch window and made independent of rolling speed by encoder-based resampling. Two empirical findings make the machinery work: using all three acceleration axes beats using the lateral axis alone, even though the lateral axis has the highest marginal correlation with the force; and the coarse 5-degree grid is statistically indistinguishable from the fine 0.5-degree grid, collapsing the model from 420 to 42 variables without loss of accuracy. The uncertainty output is the payoff: variance grows where the signal is noisy, so the band can serve as a reliability measure for a downstream controller.

What would settle it

Retrain the same regressor on the same data, mount the accelerometer at a different position on the inner liner (or run a different tire at a different inflation pressure), and measure the held-out NRMSE: if the error climbs well past the reported ~8%, the specific sensor location and window, not a general deformation-to-force relation, are what the model learned. A complementary check is to count how often the measured force falls outside the reported confidence band on the held-out data: coverage far below the band's nominal level would invalidate the uncertainty claim that is the paper's main advertised advantage.

Watch

Extended reading notes

Core claim

A single tri-axial accelerometer fixed to the inner liner of a tire, sampled at the points where the contact patch deforms the carcass, contains enough information to reconstruct the total lateral tire force. The paper establishes this experimentally: over a 70-degree window centred on the contact patch, the longitudinal, lateral, and radial accelerations are resampled by encoder angle onto a speed-independent grid, and a 5-degree grid (42 input points) proves statistically as accurate as a 0.5-degree grid (420 input points). A Gaussian process regressor with a Matérn-3/2 kernel and automatic relevance determination predicts lateral force with 9.13% ± 1.30% NRMSE in five-fold cross-validation and 7.91% NRMSE (R = 0.99) on a separate data set with a different slip-angle protocol. The claim is not only the accuracy but the companion uncertainty: posterior variance is larger for precisely the high-slip, high-load cases where the acceleration signal is noisiest, and the reported confidence bands contain the measured forces. On this evidence the authors propose the estimator as a candidate input for vehicle control strategies, and present the application as the first use of Gaussian process regression for this tire-sensing problem.

Load-bearing premise

The method stands on the premise that 42 acceleration samples from one inner-liner location, inside a fixed 70-degree window around the contact patch, carry all the force-relevant deformation information across every load, speed, and slip angle tested, and that this learned mapping transfers beyond the single tire, sensor placement, and inflation pressure studied.

Editorial extensions

If this is right

A tire with a single tri-axial accelerometer can act as a lateral-force sensor across vertical loads of 2080–6240 N, speeds of 30–60 km/h, and slip angles up to ±8 degrees — an operating range in which estimation techniques are acknowledged to struggle.
Every prediction carries an uncertainty band that widens exactly where the signal is noisy (high slip, high load), so a controller can treat the posterior variance as a reliability signal rather than a fixed error margin.
Resampling the contact patch to a 5-degree grid (42 inputs) gives statistically identical accuracy to a 0.5-degree grid (420 inputs), cutting model complexity by a factor of ten.
All three acceleration axes are needed: lateral-only input is the worst configuration, while the longitudinal and radial channels, which carry information about rolling resistance and normal load, materially improve the prediction.
The same preprocessing-and-GPR pipeline is a template for the outputs the paper names as future work: longitudinal force, radial force, and tire moments.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If the learned mapping transfers across sensor placements and tire models, instrumented tires could eventually replace wheel-hub force transducers and instrumented test rigs as the force reference, which would change how tire-force models are parameterised in practice.
The paper reports the widening of the uncertainty bands qualitatively; a natural next step is to count how often the measured force actually leaves the stated confidence band on held-out data, converting a visual claim into a coverage guarantee a control loop could rely on.
Coarser input grids than 5 degrees are untested, so the true information floor of the contact-patch signal is unknown; probing 10- or 15-degree grids, or a window not centred on the patch, would reveal whether the patch location or the deformation pattern itself carries the force information.
Because the largest prediction errors coincide with the widest bands, the variance could be used online as a trust gate: a stability controller could blend the Gaussian-process force estimate with a conservative model in proportion to the reported uncertainty.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Request a human review

A listed scientist reviews the paper for a fee and the review publishes here regardless of verdict. See the reviewers or get listed.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Short version: the paper reports a genuine experimental result - GPR trained on one maneuver set predicts lateral tire force on a different held-out maneuver set with NRMSE 7.91% and R=0.99, using 42 features from a single tri-axial accelerometer. The high-slip claim in the abstract is stronger than the evidence; the paper's own Fig. 15 shows error rising sharply above 6 degrees slip at high load, so the aggregate numbers are likely carried by the low-slip majority. That caveat is acknowledged in Section 5.2, which is to the authors' credit, but the abstract and conclusions still overstate.

What's new: the combination of experimental intelligent-tire acceleration data with Gaussian process regression for lateral force prediction. It is an application of a standard method rather than a methodological advance, but a useful one for the intelligent tire / vehicle state estimation community. The design is mostly careful: Data set 2 is truly held out and not used for model selection; input selection is done via cross-validation on Data set 1; the reduction from 420 to 42 inputs costs little accuracy; uncertainty bands are reported and behave sensibly (wider at high slip).

Soft spots, in proportion. (1) No baseline comparison. Without a neural network, Kalman filter, or even a simple linear regression, we have no evidence GPR outperforms cheaper tools. (2) The NRMSE equation (Eq. 13) appears to be missing the square on the residual term - likely a typesetting slip, but it should be fixed. (3) Uncertainty calibration is asserted qualitatively ("confidence bars comprised the measured forces") but never quantified; coverage should be computed over the test set. (4) Stratified metrics by slip angle and load are missing. The stress-test note is correct: the abstract's "even high slip angles" is not supported by the aggregate test-set numbers alone. (5) A single tire, a single sensor location, and one inflation pressure limit generalization; this is a reasonable scope limitation, but it should be stated plainly. No public data or code is provided, so the result is not independently reproducible.

None of this sinks the contribution. The central claim - that a GPR soft sensor can predict lateral force from inner-liner acceleration - holds for moderate slip conditions, and the paper is unusually honest about where it degrades.

Who this is for: researchers working on intelligent tires, vehicle state estimation, or ML-based tire force prediction. It deserves a serious referee, but the referee should ask for stratified metrics, a baseline, quantified uncertainty, and the corrected NRMSE formula before acceptance.

Referee Report

4 major / 5 minor

Summary. The paper develops a Gaussian Process Regression (GPR) model to predict tire lateral force from a single tri-axial accelerometer mounted on the inner liner of a tire. Data were collected on an MTS Flat-Trac rig over a range of vertical loads, velocities, and slip angles up to ±8 degrees. The accelerometer signals are filtered, the contact patch is extracted, and a 70-degree angular window is resampled at 5-degree resolution, yielding 42 input features. The GPR uses a Matérn 3/2 kernel with ARD, trained on Data set 1 with k-fold cross-validation, and tested on the held-out Data set 2. The reported test accuracy is NRMSE 7.91% with R = 0.99, and the posterior confidence bands are said to cover the measured forces. The paper claims this demonstrates reliable prediction even at high slip angles, useful for vehicle control.

Significance. If the claims hold, the paper makes a useful contribution to intelligent-tire soft sensing: it shows that a relatively compact GPR model can map accelerometer signals to lateral force with moderate error on a held-out test set, and that the model provides uncertainty information that could inform control strategies. Strengths include the use of a test set that was not used for model selection, a clear description of the experimental platform and preprocessing pipeline, and the reporting of wall-clock prediction time for real-time feasibility. The main weaknesses are the lack of baseline comparisons, the absence of stratified error analysis for the high-slip/high-load regime highlighted in the abstract, an incorrectly typeset NRMSE formula, and the purely qualitative treatment of uncertainty calibration. These issues are fixable and do not invalidate the core approach, but they must be addressed before the central claims can be accepted.

major comments (4)

[Abstract, Section 5.2, Fig. 15] The abstract claims the system provides reliable information 'even in the case of high slip angles,' but Section 5.2 and Fig. 15 show that errors increase sharply for slip angles above about 6 degrees, particularly at high vertical load. Because Data set 2 contains many rotations at lower slip angles (Table 1, Fig. 6), the aggregate test NRMSE of 7.91% and R=0.99 can mask poor performance in exactly the regime the abstract emphasizes. Please report NRMSE (and ideally mean absolute error) stratified by slip-angle bins and vertical load, and revise the abstract and conclusions to match what the stratified results actually show.
[Eq. (13), Section 5.2] The normalized root mean square error is typeset as NRMSE(%) = 100 sqrt(1/N sum_{i=1}^N y_i - hat{y}_i / y_max), which is dimensionally inconsistent and ambiguous: the residuals should be squared inside the square root, with the normalization applied correctly (e.g., 100*sqrt((1/N)*sum((y_i - hat{y}_i)^2)) / y_max, or the equivalent with the denominator inside the root). As printed, the formula cannot reproduce the reported 7.91%, so the quantitative accuracy claim is not reproducible until this is corrected.
[Section 5.2 (no numbered baseline section)] The claim that GPR predicts lateral forces with 'acceptable accuracy' is not supported by any comparison to alternative methods. Please include baseline models on the same train/test split—for example, linear regression, ridge regression, a neural network, or a physics-based tire model such as the Magic Formula—so that the reader can judge whether the GPR accuracy and its uncertainty output justify the added complexity. Without a baseline, the 7.91% NRMSE has no reference point.
[Section 5.2, Figs. 12-13] The uncertainty estimates are described only qualitatively ('the GPR confidence bars comprised the measured lateral forces'). For vehicle control applications, the calibration of the predictive variance should be quantified: report, for example, the empirical coverage of the 95% credible interval on the test set, or the negative log predictive density, ideally stratified by slip angle. This is important because Fig. 15 shows error increasing with slip angle while the text states that uncertainty also widens there; a quantitative calibration check would show whether the intervals are useful or merely wide.

minor comments (5)

[Section 2.2, Eq. (4)] Equation (4) has a typographical error: the covariance definition should be E[(f(x)-mu(x))(f(x')-mu(x'))], with the closing parenthesis placed correctly; as typeset, the expectation is ambiguous.
[Section 5.1, Eq. (12)] The log-marginal-likelihood expression in Eq. (12) uses |K_y| without explicitly stating that this is the determinant; please define the notation and ensure the dimensions of K_y are consistent with the earlier equations.
[Section 3.2, Section 5.1] The choice of the 70-degree contact-patch window and the 5-degree resampling is justified empirically on Data set 1, but the paper should note explicitly that the model has been validated only for one tire, one sensor location, one inflation pressure, and one test rig, so transfer to other tires or sensor placements remains untested.
[Section 1, Introduction] The statement that this is 'probably the first attempt to use GPR' for tire force prediction from intelligent tire data should be either supported by a careful literature search or softened to avoid making an unverifiable priority claim.
[Figure 11(b)] Figure 11(b) is discussed in the text as showing that 5-degree resolution is not statistically worse than 0.5-degree resolution, but no statistical significance test is reported; consider adding a brief statement of the test used or interpret the box plots more cautiously.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the GPR lateral-force predictor is trained and selected on Data set 1 and tested on held-out Data set 2, with inputs (accelerations) and outputs (measured lateral force) distinct physical quantities.

full rationale

The paper's derivation chain is empirically self-contained rather than circular. The model inputs are 42 acceleration features extracted from a single tri-axial accelerometer (Section 5.1), and the target is the lateral tire force measured by the MTS Flat-Trac load cell (Section 3.1). There is no equation-level identity between the input accelerations and the output force; the mapping is learned nonparametrically by GPR. The GPR hyperparameters are obtained by maximizing the marginal likelihood on Data set 1 only (Eq. 12), and the input selection (which acceleration components and which contact-patch resolution) is performed with a holdout cross-validation on Data set 1. Data set 2 is explicitly reserved for testing: the paper states in Section 5.2 that it 'was not used before neither to set nor to select the GPR parameters and functions.' The reported test NRMSE of 7.91% and R = 0.99 are therefore genuine out-of-sample generalization metrics, not fitted values renamed as predictions. The posterior uncertainty bands are computed from the GPR predictive variance (Eq. 10) and are not fitted to the measured forces. The paper's own admission that accuracy degrades at high slip angles and high loads (Section 5.2, Fig. 15) is a limitation on the strength of the headline claim, but it is not circularity. The few self-citations (e.g., Refs. [3], [34], [39]) are used for background or for prior vehicle-estimation methods, not as the load-bearing justification for the GPR prediction result. Therefore no circular step can be identified, and the appropriate score is 0.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The central claim rests on an empirical mapping between a single accelerometer's contact-patch acceleration profile and total lateral force. The paper introduces no mechanistic link, so the assumptions above carry the load. All GPR hyperparameters are fitted from training data, and the 5-degree input resolution is a validation-driven design choice.

free parameters (5)

GPR output scale sigma_f = not reported
Kernel hyperparameter optimized by maximizing log marginal likelihood (Eq. 12) on the training set; affects the magnitude of predictive variance.
GPR length scale(s) l = not reported
ARD kernel has a length scale per input dimension, fit on training data; controls smoothness and generalization.
GPR noise variance sigma_epsilon^2 = not reported
Included in Eqs. 6 through 10 and optimized with the other hyperparameters; absorbs output noise.
Contact patch angular resolution = 5 degrees
Selected among tested resolutions based on cross-validation box plots (Fig. 11b); a design choice that trades input count (42 versus 420) against accuracy.
Contact patch extraction window = 70 degrees
Chosen in Section 3.2 to cover the contact patch across conditions; determines which acceleration samples populate the model inputs.

assumptions (5)

domain assumption Local inner-liner accelerations are sufficient to infer total lateral force.
The model is purely data-driven; no physical model connects local strain to global F_y. Section 5.1 uses 42 acceleration features as inputs and F_y as output.
domain assumption MTS Flat-Trac force measurements are unbiased ground truth.
Section 3.1 states forces are measured and recorded by the MTS testbed; any bias or compliance in the rig becomes the target the GPR learns.
domain assumption Matern 3/2 kernel with ARD is an appropriate prior.
Section 5.1 adopts v=3/2 citing [14] as normally appropriate, without a formal justification or comparison to other kernels.
domain assumption The 70-degree angle window and 0.5-degree sampling contain all force-relevant information.
Section 3.2 discards accelerations outside the window, so any force information carried by the rest of the rotation is inaccessible to the model.
domain assumption Output noise is independent and identically distributed Gaussian noise.
Eq. 1 assumes yi=f(xi)+epsilon with Gaussian noise, used in the GPR likelihood and prediction variances.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Lateral Force Prediction using Gaussian Process Regression for Intelligent Tire Systems." pith.science (2026). https://pith.science/paper/2RXNSNWZ

@misc{pith2026200912463,
  author       = {Pith},
  title        = {Pith review of: Lateral Force Prediction using Gaussian Process Regression for Intelligent Tire Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2RXNSNWZ}},
  note         = {Machine review of arXiv:2009.12463}
}

read the original abstract

Understanding the dynamic behavior of tires and their interactions with road plays an important role in designing integrated vehicle control strategies. Accordingly, having access to reliable information about the tire-road interactions through tire embedded sensors is very demanding for developing enhanced vehicle control systems. Thus, the main objectives of the present research work are i. to analyze data from an experimental accelerometer-based intelligent tire acquired over a wide range of maneuvers, with different vertical loads, velocities, and high slip angles; and ii. to develop a lateral force predictor based on a machine learning tool, more specifically the Gaussian Process Regression (GPR) technique. It is delineated that the proposed intelligent tire system can provide reliable information about the tire-road interactions even in the case of high slip angles. Besides, the lateral forces model based on GPR can predict forces with acceptable accuracy and provide level of uncertainties that can be very useful for designing vehicle control strategies.

Figures

Figures reproduced from arXiv: 2009.12463 by the authors.

**Figure 2.** Intelligent tire system. tudinal (Acx), lateral (Acy) and radial (Acz) - are always from the same tire location. To properly study the measured accelerations and to develop the lateral force soft sensor, a wide range of maneuvers were conducted with different vertical loads, velocities and slip angles, as shown in [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** The entire testing system. 4 Tire Acceleration Measurements Analysis In this section, the acceleration signals in different directions (x, y, z)for several testing conditions are presented [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Figures from the paper (11 more)

**Figure 4.** Figure 4: Contact patch and the featured region extrac [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

**Figure 5.** Figure 5: Measurements over tire rotations of Data set 1. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

**Figure 6.** Figure 6: Measurements over tire rotations of Data set 2. [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]

**Figure 7.** Figure 7: Samples of accelerations measured over high tire speed experiments, for different loads and slip angles [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]

**Figure 8.** Figure 8: Mean values of accelerations measurements over low load and high tire speed experiments, for different slip [PITH_FULL_IMAGE:figures/full_fig_p010_8.png]

**Figure 13.** Figure 13: The GPR achieved a good prediction accuracy with [PITH_FULL_IMAGE:figures/full_fig_p010_13.png]

**Figure 9.** Figure 9: Samples of accelerations measurements over [PITH_FULL_IMAGE:figures/full_fig_p011_9.png]

**Figure 12.** Figure 12: Lateral Force prediction using GPR over vali [PITH_FULL_IMAGE:figures/full_fig_p012_12.png]

**Figure 11.** Figure 11: Comparison between GPR performance over Data set 1 (cross-validation): (a) for different acceleration measurements as inputs and (b) different contact patch resolution. situations, besides the emergence of a high temperature environment. Further efforts should be mad…

**Figure 15.** Figure 15: GPR absolute prediction error analysis (mean [PITH_FULL_IMAGE:figures/full_fig_p013_15.png]

**Figure 14.** Figure 14: Samples of Acy measurements over tire rotations for different slip angles, high speed, high load, and for the contact patch center point (Data set 2). for intelligent vehicles and its fpga implementation, IEEE Transactions on Systems, Man, and Cybernetics: Systems 49…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 39 canonical work pages

[1]

Pacejka, Tire and vehicle dynamics, Elsevier, 2005

H. Pacejka, Tire and vehicle dynamics, Elsevier, 2005

2005
[2]

Nepote, P

A. Nepote, P. De La Pierre, T. Varpula, A. Rautiainen, K. Jaakkola, The intelligent tire: Acceleration sensors data acquisition, Tech. rep., SAE Technical Paper (2005)

work page 2005
[3]

Askari, E

H. Askari, E. Hashemi, A. Khajepour, M. B. Khamesee, Z. L. Wang, Tire condition monitoring and intelligent tires using nanogenerators based on piezoelectric, electromag- netic, and triboelectric effects, Advanced Materials Tech- nologies 4 (1) (2019) 1800105

work page 2019
[4]

K. Guo, D. Lu, Unitire: uniﬁed tire model for vehicle dynamic simulation, Vehicle System Dynamics 45 (S1) (2007) 79–99

work page 2007
[5]

H. B. Pacejka, E. Bakker, The magic formula tyre model, Vehicle system dynamics 21 (S1) (1992) 1–18

1992
[6]

W. Zhao, C. Zhang, J. Zhang, Continuous measurement of tire deformation using long-gauge strain sensors, Mechan- ical Systems and Signal Processing 142 (2020) 106782

work page 2020
[7]

Pottinger, The three-dimensional contact patch stress ﬁeld of solid and pneumatic tires, Tire Science and Tech- nology 20 (1) (1992) 3–32

M. Pottinger, The three-dimensional contact patch stress ﬁeld of solid and pneumatic tires, Tire Science and Tech- nology 20 (1) (1992) 3–32

work page 1992
[8]

H. Lee, S. Taheri, A novel approach to tire parameter identiﬁcation, Proceedings of the Institution of Mechani- cal Engineers, Part D: Journal of Automobile Engineering 233 (1) (2019) 55–72

work page 2019

Show all 41 references

[9]

H. Guo, F. Liu, F. Xu, H. Chen, D. Cao, Y . Ji, Nonlinear model predictive lateral stability control of active chassis 12 Figure 13: Lateral Force prediction using GPR over the test data set - Data set 2. Figure 14: Samples of Acy measurements over tire rota- tions for differe...

2019
[10]

Mendoza-Petit, D

M. Mendoza-Petit, D. Garc ´ıa-Pozuelo, V . D´ıaz, O. Olatun- bosun, A strain-based intelligent tire to detect contact patch features for complex maneuvers, Sensors 20 (6) (2020) 1750

2020
[11]

Acosta, M

M. Acosta, M. Gladstone, J. Prins, C. Rayo, M. Furlan, J. Guti ´errez, A. O’Neill, On tyre force virtual sensing for future automated vehicle-based objective tyre testing (avbott), Vehicle System Dynamics 57 (12) (2019) 1795– 1821

2019
[12]

L. A. Aguirre, B. O. S. Teixeira, B. H. G. Barbosa, A. F. Figure 15: GPR absolute prediction error analysis (mean and standard deviation) over Data set 2, according to dif- ferent values of slip angles and with vertical loads from 2080 N to 6240 N. Teixeira, M. C. M. M. Campos...

2017
[13]

B. H. G. Barbosa, L. A. Aguirre, A. P. Braga, Piecewise afﬁne identiﬁcation of a hydraulic pumping system using evolutionary computation, IET Control Theory Applica- tions 13 (9) (2019) 1394–1403

2019
[14]

Rasmussen, C

C. Rasmussen, C. Williams, Gaussian Processes for Machine Learning, Adaptive Computation and Machine Learning, MIT Press, Cambridge, MA, USA, 2006

2006
[15]

Schulz, M

E. Schulz, M. Speekenbrink, A. Krause, A tutorial on gaussian process regression: Modelling, exploring, and ex- ploiting functions, Journal of Mathematical Psychology 85 (2018) 1 – 16

2018
[16]

E. D. Klenske, M. N. Zeilinger, B. Schoelkopf, P. Hen- nig, Gaussian Process-Based Predictive Control for Peri- odic Error Correction, IEEE Transactions on Control Sys- tem Technology 24 (1) (2016) 110–121

2016
[17]

Hewing, J

L. Hewing, J. Kabzan, M. N. Zeilinger, Cautious model predictive control using gaussian process regression, IEEE Transactions on Control Systems Technology (2019) 1–8

2019
[18]

Y . Ito, K. Fujimoto, Y . Tadokoro, Kernel-based hamil- ton–jacobi equations for data-driven optimal and h-inﬁnity control, IEEE Access 8 (2020) 131047–131062

2020
[19]

Y . Wang, R. Gnadler, R. Schieschke, Two-dimensional contact area of a pneumatic tire subjected to a lateral force, Vehicle System Dynamics 23 (1) (1994) 149–163. 13

1994
[20]

Kagami, T

S. Kagami, T. Akasaka, H. Shiobara, A. Hasegawa, Anal- ysis of the contact deformation of a radial tire with camber angle, Tire Science and Technology 23 (1) (1995) 26–51

1995
[21]

K. Guo, J. Sui, A theoretical observation on empirical ex- pression of tire shear forces, Vehicle System Dynamics 25 (sup1) (1996) 263–274

1996
[22]

El-Gindy, H

M. El-Gindy, H. L. Lewis, A. S. Lewis, Development of a tire/pavement contact-stress model based on an artiﬁ- cial neural network, in: Innovations in Vehicle Design and Development-1999 (The ASME International Mechanical Engineering Congress and Exposition), ASME, 1999, pp. 25–34

1999
[23]

Urbina, C

L. Urbina, C. A. Duchanoy, G. Faustino-Gonz ´alez, M. A. Moreno-Armend´ariz, C. A. Cruz-Villar, H. Calvo, A novel tire contact patch soft sensor via neural networks, in: 2015 12th International Conference on Electrical Engineering, Computing Science and Automatic Control (CCE)...

2015
[24]

Cossalter, A

V . Cossalter, A. Doria, The relation between contact patch geometry and the mechanical properties of motorcycle tyres, Vehicle System Dynamics 43 (sup1) (2005) 156– 164

2005
[25]

Ivanov, Analysis of tire contact parameters using visual processing, Advances in Tribology 2010 (491723) (2010) 11

V . Ivanov, Analysis of tire contact parameters using visual processing, Advances in Tribology 2010 (491723) (2010) 11

2010
[26]

Anghelache, R

G. Anghelache, R. Moisescu, Measurement of stress dis- tributions in truck tyre contact patch in real rolling con- ditions, Vehicle System Dynamics 50 (12) (2012) 1747– 1760

2012
[27]

Xiong, A

Y . Xiong, A. Tuononen, A laser-based sensor system for tire tread deformation measurement, Measurement Sci- ence and Technology 25 (11) (2014) 115103

2014
[28]

Khaleghian, O

S. Khaleghian, O. Ghasemalizadeh, S. Taheri, Estimation of the tire contact patch length and normal load using in- telligent tires and its application in small ground robot to estimate the tire-road friction, Tire Science and Technol- ogy 44 (4) (2016) 248–261

2016
[29]

Baffet, A

G. Baffet, A. Charara, D. Lechner, Estimation of vehicle sideslip, tire force and wheel cornering stiffness, Control Engineering Practice 17 (11) (2009) 1255–1264

2009
[30]

Viehweger, C

M. Viehweger, C. Vaseur, S. van Aalst, M. Acosta, E. Re- golin, A. Alatorre, W. Desmet, F. Naets, V . Ivanov, A. Fer- rara, et al., Vehicle state and tyre force estimation: demon- strations and guidelines, Vehicle System Dynamics (2020) 1–28

2020
[31]

K. B. Singh, M. A. Arat, S. Taheri, Literature review and fundamental approaches for vehicle and tire state estima- tion, Vehicle System Dynamics 57 (11) (2019) 1643–1665

2019
[32]

H. Guo, Z. Yin, D. Cao, H. Chen, C. Lv, A review of esti- mation for vehicle tire-road interactions toward automated driving, IEEE Transactions on Systems, Man, and Cyber- netics: Systems 49 (1) (2019) 14–30

2019
[33]

W. Cho, J. Yoon, S. Yim, B. Koo, K. Yi, Estimation of tire forces for application to vehicle stability control, IEEE Transactions on Vehicular Technology 59 (2) (2009) 638– 649

2009
[34]

Hashemi, M

E. Hashemi, M. Pirani, A. Khajepour, A. Kasaiezadeh, S.-K. Chen, B. Litkouhi, Corner-based estimation of tire forces and vehicle velocities robust to road conditions, Control Engineering Practice 61 (2017) 28–40

2017
[35]

Rezaeian, R

A. Rezaeian, R. Zarringhalam, S. Fallah, W. Melek, A. Khajepour, S.-K. Chen, N. Moshchuck, B. Litkouhi, Novel tire force estimation strategy for real-time imple- mentation on vehicle applications, IEEE Transactions on Vehicular Technology 64 (6) (2014) 2231–2241

2014
[36]

L. R. Ray, Nonlinear tire force estimation and road fric- tion identiﬁcation: Simulation and experiments, Automat- ica 33 (10) (1997) 1819–1833

1997
[37]

Doumiati, A

M. Doumiati, A. Charara, A. Victorino, D. Lechner, Ve- hicle dynamics estimation using Kalman ﬁltering: experi- mental validation, John Wiley & Sons, 2012

2012
[38]

Khaleghian, O

S. Khaleghian, O. Ghasemalizadeh, S. Taheri, G. Flintsch, A combination of intelligent tire and vehicle dynamic based algorithm to estimate the tire-road friction, SAE In- ternational Journal of Passenger Cars-Mechanical Systems 12 (2) (2019) 81–98

2019
[39]

Askari, Towards intelligent tire and self-powered sens- ing systems, Ph.D

H. Askari, Towards intelligent tire and self-powered sens- ing systems, Ph.D. thesis, University of Waterloo, Canada (2019)

2019
[40]

H. Jin, X. Chen, L. Wang, K. Yang, W. Lei, Adaptive soft sensor development based on online ensemble gaussian process regression for nonlinear time-varying batch pro- cesses, Industrial & Engineering Chemistry Research 54 (2015) 150714220845003

2015
[41]

H. Liu, Y . Ong, X. Shen, J. Cai, When gaussian process meets big data: A review of scalable GPs, IEEE Trans- actions on Neural Networks and Learning Systems (2020) 1–19. 14

2020

Pith tools

Reviewed August 27, 2026 · model on record in the stance chip above.

[1] [1]

Pacejka, Tire and vehicle dynamics, Elsevier, 2005

H. Pacejka, Tire and vehicle dynamics, Elsevier, 2005

2005

[2] [2]

Nepote, P

A. Nepote, P. De La Pierre, T. Varpula, A. Rautiainen, K. Jaakkola, The intelligent tire: Acceleration sensors data acquisition, Tech. rep., SAE Technical Paper (2005)

work page 2005

[3] [3]

Askari, E

H. Askari, E. Hashemi, A. Khajepour, M. B. Khamesee, Z. L. Wang, Tire condition monitoring and intelligent tires using nanogenerators based on piezoelectric, electromag- netic, and triboelectric effects, Advanced Materials Tech- nologies 4 (1) (2019) 1800105

work page 2019

[4] [4]

K. Guo, D. Lu, Unitire: uniﬁed tire model for vehicle dynamic simulation, Vehicle System Dynamics 45 (S1) (2007) 79–99

work page 2007

[5] [5]

H. B. Pacejka, E. Bakker, The magic formula tyre model, Vehicle system dynamics 21 (S1) (1992) 1–18

1992

[6] [6]

W. Zhao, C. Zhang, J. Zhang, Continuous measurement of tire deformation using long-gauge strain sensors, Mechan- ical Systems and Signal Processing 142 (2020) 106782

work page 2020

[7] [7]

Pottinger, The three-dimensional contact patch stress ﬁeld of solid and pneumatic tires, Tire Science and Tech- nology 20 (1) (1992) 3–32

M. Pottinger, The three-dimensional contact patch stress ﬁeld of solid and pneumatic tires, Tire Science and Tech- nology 20 (1) (1992) 3–32

work page 1992

[8] [8]

H. Lee, S. Taheri, A novel approach to tire parameter identiﬁcation, Proceedings of the Institution of Mechani- cal Engineers, Part D: Journal of Automobile Engineering 233 (1) (2019) 55–72

work page 2019