Vehicle makers: tyre forces and track data models cannot explain

Tyre forces are unknown, so safety margins stay wide

What happens. Every driving-dynamics function depends on how much force the tyre can transmit now. Nobody measures that. Grip changes with road surface, wetness, temperature and wear, and the control works with a conservative assumption.

What it costs. Stability and braking functions that intervene too early in good conditions, comfort and sportiness given up for safety, and little room for functions that need to know the limit.

Why the usual fix fails. A tyre sensor adds cost and does not tell what the road does. Fixed tables for dry and wet are crude. A function that reacts after the tyre has started to slide has no time left to prevent it.

How a fixed-price remote diagnosis checks it. From vehicle signals you already record (wheel speeds, accelerations, yaw rate, steering angle, brake and drive torque) I look at how well the tyre forces can be estimated and under which conditions the estimate fails. The fixed-price remote diagnosis tells you whether a computed tyre force can let the margins shrink, and what it takes.

Background: Computing unmeasured states for control

Vehicle dynamics models do not match track data

What happens. A vehicle dynamics model is built from design data and simulation. On the track, with real tyres and real set-ups, the car behaves differently. The team compares simulation with measured data, finds gaps, and corrects parameters one by one.

What it costs. Set-up and control decisions made on a model that is off, track days used for finding out what the model gets wrong, and engineers busy with correlation instead of development.

Why the usual fix fails. Adjusting parameters by hand until one lap fits breaks the next one. A more detailed model brings more parameters to fit. Data from different days, tyres and temperatures blur each other.

How a fixed-price remote diagnosis checks it. From track recordings and the current model I look at which parameters the data really identify, and where the model structure itself is missing something. The fixed-price remote diagnosis tells you whether identification against the data can give a model that holds across conditions.