Estimating Tire Grip During Gentle Driving

The challenge is to estimate continuously how much more braking, cornering or acceleration force each tire can deliver before sliding, using existing wheel-speed and vehicle-control signals without adding sensors. Current model-based methods generally need tires to use more than 30% of their available grip before they can estimate the limit reliably, but everyday and autonomous driving usually stays below that level. During gentle driving, surfaces with very different grip limits can produce similar signals, while tire wear, temperature and changes in load further blur the differences. The software must separate these effects in noisy data, distinguish at least four grip levels and update each wheel’s remaining grip estimate fast enough for autonomous braking and path planning, before a wheel starts to slide.

Idea 01 of 03Borrowed from Continuous glucose monitoring

Borrowing factory-calibrated drift correction from glucose monitors, we predict and subtract each tire’s contribution to wheel-speed signals to estimate road grip during gentle, everyday driving.

How it works

Wheel-speed signals change with road grip, but also with tire wear, temperature, pressure and load. During gentle driving, those tire effects make grip hard to distinguish. Glucose-monitor makers solve a related problem by correcting for predictable changes in the sensor itself. We apply that approach to tires: factory drum tests establish each tire model’s wheel-speed signature on a reference surface across different tire conditions. In the vehicle, existing software estimates each tire’s current condition, and we subtract the matching predicted signature from the measured signal. A small machine-learning model translates the remaining pattern into a grip estimate for each wheel. This separates tire correction from grip estimation instead of trying to infer both together, using the vehicle’s existing anti-lock braking wheel-speed sensors without adding hardware.

Where have we seen this before?

Medical sensingDexcom G6 continuous glucose monitorThe G6 uses a factory-calibrated function that adjusts sensor gain and offset according to time since insertion. This corrects predictable sensor drift over its 10-day wear period without requiring routine fingerstick calibration.Shared idea Factory calibration that corrects for predictable changes in the sensing element before estimating the target quantity.Factory-Calibrated Continuous Glucose Monitoring: How and Why It Works, and the Dangers of Reuse Beyond Approved Duration of Wear (opens in a new tab)
Astronomical imaging and spectroscopyJames Webb Space Telescope data-processing pipelineThe JWST pipeline subtracts background images obtained from dedicated background observations or observations with the target shifted to another position. Astronomers can then measure the target’s light with less contamination from the surrounding background.Shared idea Estimate an unwanted contribution separately, then subtract it from the measurement to isolate the target signal.JWST Pipeline Documentation: Background Subtraction Methods — Introduction (opens in a new tab)

Idea 02 of 03Borrowed from Nuclear reactor physics

Borrowing nuclear reactor monitoring’s variance-to-mean counting method, we use fluctuations in existing wheel-speed signals to estimate available tire grip during gentle driving, when average wheel-speed responses reveal little.

How it works

A car needs to know how much grip remains before it brakes hard, but conventional wheel-speed methods struggle when the tires use only a small fraction of available grip. Nuclear engineers face a similar measurement problem: they assess neutron chain reactions without pushing a reactor to its limit by measuring how detector counts cluster over time. We propose applying their Feynman-Y variance-to-mean statistic to tiny timing fluctuations in the car’s existing ABS wheel-speed pulses. Even during gentle driving, small areas of tread slip and transfer stress to their neighbors. The proposed grip estimate comes from how strongly these slips trigger further slips: more propagation indicates less available friction. The software compares fluctuations across several time windows and wheels, then uses tire models to separate the cascade signal from road texture, wear, pressure and temperature. This adds a grip-sensing channel without extra sensors or deliberate hard braking.

Where have we seen this before?

Nuclear reactor monitoringKyoto University Critical Assembly subcriticality experimentsResearchers applied Feynman variance-to-mean and Rossi-α timing measurements to neutron detections at the Kyoto University Critical Assembly. The measured decay parameters changed with subcriticality, providing information about how far the assembly was below a self-sustaining chain reaction.Shared idea Inferring chain-reaction behavior from the variance and timing of detected events.Application of variance-to-mean technique to subcriticality monitoring for accelerator-driven subcritical reactor (opens in a new tab)
Earthquake seismologyYosihiko Ogata’s Epidemic-Type Aftershock Sequence (ETAS) modelOgata’s ETAS model represents earthquakes as events that temporarily raise the likelihood of more earthquakes, including aftershocks that trigger their own aftershocks. Fitting these clustered sequences provides the cascade-modeling component borrowed for the proposed analysis of tiny tread slips.Shared idea Modeling local slips as cascades in which each event can trigger further events.Statistical Models for Earthquake Occurrences and Residual Analysis for Point Processes (opens in a new tab)

Idea 03 of 03Borrowed from Bridge structural health monitoring

Borrowing bridge-monitoring methods for measuring vibration decay and cancelling shared drift, we compare two kinds of tire vibration in existing wheel-speed data to estimate available grip during gentle driving.

How it works

Drivers need to know how much grip remains before they brake hard, but conventional slip-based estimates need substantial tire-road force to reveal the limit. We propose measuring how quickly tire vibrations fade instead, using methods engineers apply to bridges under everyday traffic. Even when a tire rolls without an obvious skid, tiny areas at the edges of its road contact slip and lose energy. On a slippery surface, this slipping region widens and damps twisting vibrations more than radial vibrations, which move the tread inward and outward. Software extracts both damping values from existing ABS wheel-speed signals and compares their ratio. That comparison reduces changes shared by both vibrations, such as those caused by temperature, pressure and wear. The software also accounts for how strongly road roughness shakes the tire and uses cointegration, a bridge-monitoring technique, to cancel remaining shared drift over time. The result is a per-wheel grip estimate from ordinary driving, without extra sensors or a maneuver near the grip limit.

Where have we seen this before?

Bridge vibration monitoringZ24 Bridge vibration-identification benchmarkResearchers identified the Z24 Bridge's vibration frequencies, damping ratios and mode shapes with stochastic subspace identification. We borrow that signal-processing method to extract tire damping from wheel-speed measurements.Shared idea Extracting vibration decay from measured structural motion.Z24 Bridge benchmark — KU Leuven Structural Mechanics (opens in a new tab)
Structural health monitoringCointegration of the Z24 Bridge's natural frequenciesResearchers combined four bridge-frequency time series using cointegration to remove environmental and operational effects, including temperature changes. We apply the same principle to separate slow tire-condition drift from friction-related changes.Shared idea Cancelling environmental drift shared across several vibration signals.A Topological Analysis of Cointegrated Data: A Z24 Bridge Case Study (opens in a new tab)
Machining fixture designBrun and Melkote's spherical-workpiece fixture experimentsBrun and Melkote modeled energy loss where a spherical workpiece touched flat fixture surfaces under oscillating loads. Their experiments agreed favorably with the partial-slip damping model, demonstrating the contact-level mechanism we use as a tire-grip signal.Shared idea Vibration damping caused by tiny slipping regions within an otherwise gripping contact.Modeling and experimental verification of partial slip for multiple frictional contact problems (opens in a new tab)

Estimating Tire Grip During Gentle Driving

Subtract the Tire’s Signal to Read Road Grip

How it works

Find Road Grip in Tiny Wheel-Speed Fluctuations

How it works

Read Tire Grip from How Vibrations Fade

How it works