# Understeer gradient and steady-state cornering (bicycle model)

> The linear bicycle model defines understeer gradient, neutral steer, characteristic and critical speed. In data it becomes the understeer-angle channel, a powerful balance statistic, provided you remember it is only strictly valid in the linear range.

Area: Weight Transfer & Balance · Starting level: L2 Practitioner · Updated 2026-11-16
Source page: https://race-engineer-academy.vercel.app/lesson/wt-understeer-gradient-steady-state

## L1 · Foundation

Collapse the car into a **bicycle**: one front and one rear tyre on the centreline. At very low speed, with no slip angles, the steer angle needed for a corner of radius $R$ is the **Ackermann (kinematic) angle** $\delta_{kin} = L/R$.

At speed, both axles run slip angles. If the front needs more slip angle than the rear, you need extra steering beyond Ackermann. That extra is **understeer**:

$$\delta = \frac{L}{R} + K\,a_y$$

$K$ is the **understeer gradient** (deg of road-wheel angle per g):
- $K > 0$: **understeer**. More steer needed as lateral g rises.
- $K = 0$: **neutral steer**. Steer stays at the kinematic value.
- $K < 0$: **oversteer**. Less steer needed; above a critical speed the car is unstable without driver correction.

In the bicycle model $K$ comes from axle loads and axle cornering stiffnesses:

$$K = \frac{m_f}{C_f} - \frac{m_r}{C_r}$$

A heavier front or a less stiff front axle gives understeer. Race cars in the linear range run small positive gradients, of order a few tenths of a degree to ~1° per g at the road wheels. Road cars run more, typically 1.5–4°/g, for stability.

**Using $L/R = L\,a_y/V^2$** you can compute the kinematic angle from logged speed and lateral G. Steering-wheel angle divided by steering ratio, minus that, is the **understeer angle** (Segers §7.5). It is a key balance channel.

## L2 · Practitioner

**The understeer-angle math channel (Segers §7.5):**

$$\delta_u = \delta_{wheel} - 57.3\,\frac{L\,a_y}{V^2}$$

with $a_y$ in m/s², $V$ in m/s and angles in degrees. Steps:
1. Measure steering-wheel → road-wheel angle on turn plates (outside wheel) to get a steering-ratio function. It may be non-linear.
2. Build the kinematic-angle channel from $L$, logged $a_y$ and speed.
3. Take absolute values to remove left/right sign conventions.

**Example (GT3, $L = 2.70$ m):** 150 km/h (41.7 m/s), 1.6 g (15.7 m/s²): kinematic angle $= 57.3 \times 2.70 \times 15.7/41.7^2 = 1.40°$. If the logged road-wheel angle is 2.6°, the understeer angle is **1.2°**.

**How to use it:**
- **Track map overlay**: colour the understeer angle by position to see where the car pushes (entry/mid/exit).
- **Lap statistic**: average understeer angle per lap (gated for $|a_y| >$ some threshold). Segers' GT1 example: average falling from ~1.0° to ~0.7° over five laps as the rears went off, and the fastest lap had the highest average understeer angle (Segers Table 7.2).
- **Stint run chart**: trends towards oversteer as rears degrade on RWD cars.
- **Driver comparison**: lap time vs average understeer angle shows each driver's preferred balance window (Segers Fig. 7.28).

**Complementary channels (know what they really measure):**
- **Attitude velocity** $r - a_y/V$ (Segers §7.6). Kinematics give $a_y = V(r + \dot\beta)$, so this channel equals $-\dot\beta$. It is **zero in any steady corner**, understeering or oversteering, and spikes when body slip is changing: positive when the rear is stepping out, negative when the car is straightening up. It is a transient oversteer/understeer *event* detector, not a steady-state balance measure.
- **Front minus rear lateral acceleration** from two accelerometers on the axle centrelines equals $L\,\dot r$ (yaw acceleration). Segers §7.7 reads it as a balance indicator; physically it flags rotation building (entry) or dying (exit) and reads zero mid-corner whatever the balance.
- For **steady-state** balance use the understeer angle, or measured yaw rate vs a bicycle-model reference $r_{ref} = V\delta/(L + K V^2)$.

**Pitfall:** the gradient is defined in **steady state, linear range**. At the limit, under braking or with large body slip, the numbers are still useful for comparison but not for absolute claims ("the car has 1.2°/g of understeer gradient at 2 g" is meaningless).

## L3 · Advanced

**Bicycle-model results (RCVD Ch. 5).** With axle cornering stiffnesses $C_f, C_r$ (N/rad), CG distances $a, b$ to the front and rear axles, and $m_f = m b/L$, $m_r = m a/L$:

- Understeer gradient: $K = m_f/C_f - m_r/C_r$ (rad per m/s²).
- Steady-state yaw-rate gain: $\dfrac{r}{\delta} = \dfrac{V}{L + K V^2}$.
- **Characteristic speed** ($K > 0$): $V_{ch} = \sqrt{L/K}$. Yaw gain peaks there at $V_{ch}/(2L)$, and above it more speed gives *less* yaw response per degree of steer.
- **Critical speed** ($K < 0$): $V_{crit} = \sqrt{-L/K}$. The yaw gain goes to infinity and the car becomes divergently unstable without correction.
- **Static margin**: $SM = (x_{NSP} - a)/L$, where the neutral steer point is where a side force produces no yaw. Positive SM means understeer. For race cars typically a few percent of wheelbase.

**Worked example.** $m = 1300$ kg, 45% front, $L = 2.70$ m, effective axle stiffness $C_f = 180$ kN/rad, $C_r = 300$ kN/rad:
$K = 585/180\,000 - 715/300\,000 = 3.25\times10^{-3} - 2.38\times10^{-3} = 0.87\times10^{-3}$ rad/(m/s²) $\approx 0.49$ °/g.
$V_{ch} = \sqrt{2.70/0.00087} \approx 56$ m/s ≈ **200 km/h**.

Note that the rear axle needs a much higher cornering stiffness (wider rear tyres, rear compliance and toe-in) to keep $K$ small but positive on a car with 55% rear weight.

**What changes $K$ in the linear range:** anything that changes **effective** axle cornering stiffness: tyre size, pressure, camber, compliance steer, toe, roll steer, LLTD (through load-sensitive $C_\alpha$), aero load at speed. Dynamically, front roll-steer (bump-steer in roll) is a powerful and often neglected contributor.

**Limit balance is a different question.** At the limit, balance is determined by which axle saturates first (peak forces, LLTD via load-sensitive μ, aero balance), not by $K$. RCVD Ch. 7 pair analysis and Ch. 8 moment method treat the non-linear case: a car can be mildly understeering in the linear range and limit-oversteering, or the reverse.

## L4 · Expert

**Model limits and data subtleties.**
- **Steering compliance** and power-steering assistance mean steering-wheel angle ÷ ratio ≠ road-wheel angle under load. In high-$F_y$ corners a GT steering system can lose several tenths of a degree to compliance. Measure the ratio under load (or with a column torque sensor plus a compliance model) for absolute understeer-angle work.
- **Ackermann and large steer.** The bicycle model's $L/R$ ignores track-width effects. In hairpins ($R < 20$ m) inner and outer wheels steer differently (Ackermann geometry), and the "outside wheel angle" convention (Segers) matters.
- **Body slip angle** $\beta$: in steady state $\beta = b/R - m_r V^2/(C_r R)$. At low speed $\beta > 0$ (the car's nose points outside the path tangent at the CG); above $V = \sqrt{b\,C_r/m_r}$ it turns negative (nose points inside, rear drifts out). Optical slip sensors measure $\beta$ directly and let you compute axle slip angles, a much richer balance picture than one understeer number.
- **Transient**: yaw response has a natural frequency and damping that depend on $C_f, C_r, I_z, V$ (RCVD Ch. 6). Two cars with equal $K$ can feel very different on entry if their yaw damping differs (rear cornering stiffness and wheelbase dominate). Understeer gradient alone does not capture "nervousness".
- **Statistics, not snapshots.** Segers' strength is using the understeer angle as a gated, averaged statistic. Bin by speed (aero vs mechanical), by phase (entry: brake > 0; mid: |a_x| < 0.2 g; exit: throttle > 20%) and by $a_y$. A clean matrix of mean understeer angle per (phase × speed band), run-charted per stint, is one of the most useful balance tools a race engineer can own.

## Key formulas

**Steady-state steer**

$$\delta = \dfrac{L}{R} + K\,a_y$$

- \delta: road-wheel steer angle
- L: wheelbase (m)
- R: path radius (m)
- K: understeer gradient

**Understeer gradient (bicycle model)**

$$K = \dfrac{m_f}{C_f} - \dfrac{m_r}{C_r}$$

- m_f, m_r: axle masses (kg)
- C_f, C_r: effective axle cornering stiffness (N/rad)

**Understeer angle channel (Segers)**

$$\delta_u = \delta_{wheel} - 57.3\,\dfrac{L\,a_y}{V^2}$$

- \delta_{wheel}: road-wheel angle (deg)
- a_y: lateral acceleration (m/s²)
- V: speed (m/s)

**Characteristic and critical speed**

$$V_{ch} = \sqrt{L/K}\;(K>0), \qquad V_{crit} = \sqrt{-L/K}\;(K<0)$$


**Yaw-rate gain**

$$\dfrac{r}{\delta} = \dfrac{V}{L + K V^2}$$


## Trackside takeaways

- δ = L/R + K·a_y. K > 0 understeer, 0 neutral, < 0 oversteer.
- Understeer angle = road-wheel angle − 57.3·L·a_y/V². Use it as a gated, lap-averaged statistic.
- Characteristic speed √(L/K): above it, yaw response per degree of steer falls.
- K is a linear-range quantity; limit balance is decided by which axle saturates first.
- Effective axle stiffness includes compliance, toe, roll steer and load transfer, not just tyres.

## By car category

### GT3

Power steering and compliance mean steering-wheel angle needs a calibrated ratio. Average understeer angle per lap trending down over a stint is the classic signature of rear degradation on RWD GT cars.

### FIA F3

Stiff, low-compliance cars make the understeer angle channel quite accurate. Speed-binned understeer angle separates aero (high-speed) from mechanical (low-speed) balance cleanly.

### One-make Cup

Even with limited data channels (steer, speed, lat G are normally available), the understeer angle math channel is computable and is a great objective tool for driver coaching on one-make cars.

### TCR

FWD cars show understeer angle rising strongly on throttle (traction eats front lateral capacity). Gate by throttle position to separate steady-state balance from power understeer.

### Hypercar (LMH/LMDh)

Front-axle hybrid deployment on LMH cars changes effective front cornering stiffness and limit balance above the deployment speed threshold, so balance channels should be split by deployment state.

## References

- Milliken & Milliken, Race Car Vehicle Dynamics (SAE 1995): Ch. 5 Simplified Steady-State Stability and Control (Bicycle model, understeer gradient, characteristic/critical speed)
- Milliken & Milliken, Race Car Vehicle Dynamics (SAE 1995): Ch. 6 Simplified Transient Stability and Control
- Milliken & Milliken, Race Car Vehicle Dynamics (SAE 1995): Ch. 7 Steady-State Pair Analysis
- Segers, Analysis Techniques for Racecar Data Acquisition, 2nd ed. (SAE 2014): §7.5 The Understeer Angle
- Segers, Analysis Techniques for Racecar Data Acquisition, 2nd ed. (SAE 2014): §7.6 Vehicle Balance with a Yaw Rate Sensor
- Segers, Analysis Techniques for Racecar Data Acquisition, 2nd ed. (SAE 2014): §7.7 Front and Rear Lateral Acceleration (Front−rear difference is L·(yaw acceleration): a transient indicator)
