Beyond ESP: How Four-Wheel Steering (4WS) Redefines High-Speed Vehicle Stability

• Author: Johnny Liu, CEO at Dowway
• Co-author/Affiliation: Institute of Smart Carrier Equipment
• Published: June 2, 2026
• Category: Automotive Engineering, Vehicle Dynamics, Control Systems
• Read Time: 11 mins

Technical Summary

What is the stability envelope ($S_c$) of a 4WS vehicle?

The stability envelope$S_c$is a closed region in the Sideslip Angle ($\beta$) vs. Yaw Rate ($r$) plane. It is bounded by six limits: two yaw rate bounds ($L_1, L_2$) defined by peak road adhesion, and four sideslip angle bounds ($L_3, L_4, L_5, L_6$) calculated from the saturation slip angles of the front and rear tires using a Brush Tire Model.

How does 4WS improve upon Front-Wheel Steering (FWS)?

Unlike FWS, which only manages rear tire saturation ($L_7, L_8$), 4WS actively controls both axles to prevent tire saturation, pull the operating point closer to the phase-plane origin, lower lateral tire force demands, and provide active safety intervention before traditional Electronic Stability Programs (ESP) trigger.

Introduction: The Quest for High-Speed Active Safety

When a vehicle turns at high speeds, fractions of a second dictate the boundary between absolute control and spinning out. Traditionally, safety systems have relied on Electronic Stability Programs (ESP). However, ESP is reactive: it intervenes by applying brakes after a loss of traction or side-slip is detected, which slows the vehicle down and disrupts momentum.

Can we prevent instability before it starts?

The answer lies in Four-Wheel Steering (4WS). By actively controlling both the front and rear steering angles, 4WS steers the vehicle pre-emptively to maintain neutral handling and bypass the limits of traditional steering.

In this article, we translate vehicle dynamics theory into an active control framework. We will build a simplified 3-Degree-of-Freedom (3-DOF) vehicle model, estimate unmeasurable states in real-time using an Extended Kalman Filter (EKF), map the physical limits of tire adhesion using a modified Brush Tire Model, and construct the 4WS stability boundary envelope ($S_c$).

Quick FAQ

• Why do we need an EKF to find the sideslip angle ($\beta$)?Standard production cars do not have cheap sensors to measure sideslip angle directly. The EKF uses available sensors (like lateral acceleration and yaw rate) to estimate sideslip in real time.
• What causes a tire to saturate?A tire saturates when the slip angle becomes too large, causing the tire to transition from gripping to sliding. At this point, turning the steering wheel further will not generate more lateral force.
• Why is 4WS safer than FWS?FWS can only react to rear-wheel sliding. 4WS actively coordinates both axles, sharing the cornering load and keeping the vehicle balanced well within its physical safety limits.

Section 1: Building the Core – The 3-DOF Vehicle Dynamics Model

To design a real-time active controller, we must balance mathematical accuracy with computer processing speed. A full multi-body vehicle model is highly accurate, but its complex math is too heavy for standard on-board microcontrollers.

Because yaw rate ($r$) and sideslip angle ($\beta$) are the main variables for evaluating vehicle handling and lateral stability, we use a simplified 3-Degree-of-Freedom (3-DOF) vehicle dynamics model (representing longitudinal, lateral, and yaw motions).

Continuous-Time System Equations

By ignoring the steering-induced tire slip linearization errors for state-estimation purposes, we write the continuous equations of motion as follows:$$\dot{v}_x = r v_y + a_x$$$$\dot{v}_y = -r v_x + a_y$$$$\dot{r} = \frac{a F_{yf} \cos\delta_f – b F_{yr} \cos\delta_r}{I_z}$$

Where:

• $v_x, v_y$ are the longitudinal and lateral velocities at the vehicle’s Center of Gravity (CG).
• $r$ is the yaw rate.
• $a_x, a_y$ represent the longitudinal and lateral accelerations.
• $\delta_f, \delta_r$ are the steering angles of the front and rear wheels, respectively.
• $F_{yf}, F_{yr}$ represent the lateral tire forces of the front and rear axles.
• $a, b$ are the distances from the CG to the front and rear axles.
• $I_z$ is the vehicle’s yaw moment of inertia.
• $m$ is the total vehicle mass.

The vehicle’s lateral acceleration $a_y$ is modeled as:$$a_y = \frac{F_{yf} \cos\delta_f + F_{yr} \cos\delta_r}{m}$$

Ignoring small non-linear steering angle terms ($\cos\delta \approx 1$), the simplified lateral acceleration output equation is:$$a_y = \frac{F_{yf} + F_{yr}}{m}$$

Using the small-angle approximation for the vehicle sideslip angle, we define:$$\beta \approx \frac{v_y}{v_x}$$

Differentiating $\beta$ with respect to time yields:$$\dot{\beta} \approx \frac{\dot{v}_y}{v_x} – \frac{v_y \dot{v}_x}{v_x^2} \approx \frac{a_y}{v_x} – r = \frac{F_{yf} + F_{yr}}{m v_x} – r$$

Section 2: Estimating Unmeasurable States with the Extended Kalman Filter (EKF)

Many safety variables—such as the sideslip angle $\beta$—cannot be directly measured using standard, budget-friendly sensors. Instead, we must estimate them.

The Extended Kalman Filter (EKF) acts as the real-time observer for our 4WS vehicle. Traditional Kalman Filtering (KF) only works on linear systems. The EKF bypasses this limit by linearizing the non-linear dynamics about the current state estimate using a Taylor series expansion, discarding the high-order terms.

       [ Previous State x_{k-1} ]
                   │
                   ▼ (Prediction Step)
      Linearize system via Jacobian A_k
      Predict State:  x̂_{k|k-1} = f(x̂_{k-1}, u_{k-1})
      Predict Covar:  P_{k|k-1} = A_k P_{k-1} A_k^T + Q
                   │
                   ▼ (Measurement Update)
      Linearize output via Jacobian C_k
      Compute Kalman Gain: K_k = P_{k|k-1} C_k^T (C_k P_{k|k-1} C_k^T + R)^{-1}
      Correct State:  x̂_k = x̂_{k|k-1} + K_k (z_k - h(x̂_{k|k-1}, u_k))
      Correct Covar:  P_k = (I - K_k C_k) P_{k|k-1}
                   │
                   ▼
         [ Estimated State x_k ]

Figure 1: Extended Kalman Filter state estimation workflow.

Discretized State-Space Formulation

To run the estimator on an electronic control unit (ECU), we convert the 3-DOF vehicle state equations into this discrete-time state-space representation:$$x_k = f(x_{k-1}, u_{k-1}) + w_{k-1}$$$$z_k = h(x_k, u_k) + v_k$$

Where:

• $x_k = [\beta_k, r_k, v_{x,k}]^T$ is the state vector.
• $u_k = [\delta_{f,k}, \delta_{r,k}]^T$ is the control input vector.
• $z_k = [a_{y,k}, r_k]^T$ is the observation vector.
• $w_k, v_k$ represent the process and measurement white noise vectors with covariance matrices $Q$ and $R$.

To compute the prediction-update equations, we linearize the system at each time step using these Jacobians:$$A_k = \left. \frac{\partial f}{\partial x} \right|_{\hat{x}_{k-1|k-1}, u_{k-1}}, \quad C_k = \left. \frac{\partial h}{\partial x} \right|_{\hat{x}_{k|k-1}, u_k}$$

The real-time state estimate $\hat{x}_k$ then feeds directly into the safety boundary controller.

Section 3: Defining Safety Limits – The Brush Tire Model

Even a vehicle with modern 4WS cannot defy physics: its handling capability is governed by the friction at the tire-road interface.

To map these limits, we evaluate tire behavior using the physical Brush Tire Model. Under this model, the tire slip angles of the front and rear axles are defined geometrically as:$$\alpha_f = \beta + \frac{a \cdot r}{v_x} – \delta_f$$$$\alpha_r = \beta – \frac{b \cdot r}{v_x} – \delta_r$$

Understanding Tire Saturation

As slip angle increases, the tire transitions from elastic stretching to sliding, eventually reaching a point where no additional lateral force can be generated. This threshold is defined as the saturation tire slip angle ($\alpha_{sat}$):$$\alpha_{sat} = \arctan \left( \frac{3 \mu F_z \kappa}{k} \right)$$

Where:

• $\mu$ is the tire-road friction coefficient.
• $k$ is the tire cornering stiffness.
• $F_z$ is the dynamic tire vertical load.
• $\kappa$ is the longitudinal attenuation coefficient.

Factoring in Real-world Dynamics: $\kappa$ and $F_z$

To accurately locate the boundaries, we must account for two real-world dynamic factors:

1. The Longitudinal Force Discount ($\kappa$): Longitudinal forces (braking or acceleration) consume tire adhesion, leaving less grip available for turning. We integrate this interaction via the attenuation coefficient $\kappa$: $$\kappa = \sqrt{1 – \left(\frac{F_x}{\mu F_z}\right)^2}$$ As longitudinal tire force $F_x$ increases, $\kappa$ drops, shrinking the available lateral saturation slip angle $\alpha_{sat}$.
2. Dynamic Weight Transfer ($F_z$): During braking or acceleration, weight shifts between the front and rear axles. We define the dynamic vertical load $F_{zi}$ on each axle as: $$F_{zi} = F_{z, \text{static}} \pm \Delta F_z$$$$F_{zf} = \frac{m g b}{l} – \frac{m a_x h}{l}$$$$F_{zr} = \frac{m g a}{l} + \frac{m a_x h}{l}$$ Where $l = a + b$ is the wheelbase, $h$ is the center of gravity height, and $g$ is gravity. The second term represents longitudinal load transfer, which dynamically changes $F_{zf}$ and $F_{zr}$, reshaping tire saturation boundaries in real-time.

Section 4: Mapping the Closed Stability Envelope ($S_c$)

To guarantee high-speed stability, we construct a closed boundary polygon in the $(\beta, r)$ phase-plane. This is the stability envelope $S_c$. It is made of six boundary lines: four sideslip angle boundaries and two yaw rate boundaries.

                  ▲ Yaw Rate (r)
                  │          L1 (Max Yaw Rate)
         ┌────────┴────────┐
         │                 │ 
   L4    │        •        │   L3
 (Front  │      (Origin)   │ (Front
  Min)   │                 │  Max)
         └────────┬────────┘
                  │          L2 (Min Yaw Rate)
   L5 (Rear Min)  │  L6 (Rear Max)
                  ▼ Sideslip Angle (β)

Figure 2: Conceptual visualization of the bounded 4WS stability phase-plane.

1. Sideslip Angle Boundaries ($\beta$-Limits)

By setting the tire slip angles equal to their saturation thresholds ($\alpha_i = \pm \alpha_{i, sat}$), we solve for the sideslip angle $\beta$. This gives four boundary lines ($L_3, L_4$ for the front axle, and $L_5, L_6$ for the rear axle):

• Front Axle Saturation Boundaries: $$L_3: \beta = \alpha_{f, sat} – \frac{a \cdot r}{v_x} + \delta_f \quad \text{(Front Max Limit)}$$$$L_4: \beta = -\alpha_{f, sat} – \frac{a \cdot r}{v_x} + \delta_f \quad \text{(Front Min Limit)}$$
• Rear Axle Saturation Boundaries: $$L_5: \beta = -\alpha_{r, sat} + \frac{b \cdot r}{v_x} + \delta_r \quad \text{(Rear Min Limit)}$$$$L_6: \beta = \alpha_{r, sat} + \frac{b \cdot r}{v_x} + \delta_r \quad \text{(Rear Max Limit)}$$

The Front-Wheel Steering (FWS) Comparison

For a standard FWS vehicle, the rear wheels cannot steer ($\delta_r = 0$). Because the rear wheels have no steering capability, they saturate first. Therefore, FWS stability analysis only evaluates the rear tire boundaries ($L_7, L_8$):

• FWS Rear Saturation Boundaries: $$L_7: \beta = \alpha_{r, sat} + \frac{b \cdot r}{v_x} \quad \text{(FWS Rear Max)}$$$$L_8: \beta = -\alpha_{r, sat} + \frac{b \cdot r}{v_x} \quad \text{(FWS Rear Min)}$$

2. Yaw Rate Boundaries ($r$-Limits)

The absolute maximum yaw rate is bounded by peak road friction, preventing the vehicle from pulling lateral forces that exceed physical tire adhesion limits:$$L_1: r = r_{max} = \frac{\mu g}{v_x} \quad \text{(Max Yaw Rate Limit)}$$$$L_2: r = -r_{max} = -\frac{\mu g}{v_x} \quad \text{(Min Yaw Rate Limit)}$$

The Final 4WS Stability Region Set ($S_c$)

The intersection of these six boundary lines establishes the closed, stable operating region $S_c$:$$S_c = \left\{ (\beta, r) \mid -r_{max} \le r \le r_{max}, \max(L_4, L_5) \le \beta \le \min(L_3, L_6) \right\}$$

Section 5: Why 4WS Outperforms FWS: Three Key Lessons

To demonstrate the real-world advantages of 4WS, we examine a vehicle traveling at $60 \text{ km/h}$ under an abrupt direction change with a steering wheel angle of $86^\circ$. The 4WS system runs an active dual-critical-speed mode-switching strategy.

                          ▲ Yaw Rate r
                          │
             L1 ┌─────────┴─────────┐
                │                   │
                │        • (4WS)    │
                │                   │
     L4 (Front) │                   │ L3 (Front)
                │         X (FWS)   │
                └─────────┬─────────┘ 
             L2           │           
                          ▼ Sideslip β

Figure 3: Phase-portrait comparison showing the cohesive 4WS operating point (•) vs. the outer FWS operating point (X).

Lesson 1: Stronger Operating Point Convergence

In the $(\beta, r)$ phase-plane, the 4WS operating point (red circle) converges much closer to the origin $(0,0)$ compared to the FWS operating point (blue cross). In vehicle dynamics, proximity to the origin indicates a smaller vehicle slip angle and more stable, predictable handling. The FWS vehicle, being much further from the center, runs dangerously close to its stability limit.

Lesson 2: Lower Demands on Rear Tire Grip

Under identical high-speed maneuvers, 4WS distributes the steering demand across both axles. Consequently, the lateral force required from the rear wheels is substantially lower on the 4WS vehicle than on the FWS counterpart. This prevents the rear tires from reaching their saturation threshold, delaying slip and keeping a wider safety buffer.

Lesson 3: Thorough Risk Modeling via Dual-Axle Saturation

In FWS, boundary lines are calculated solely based on rear-tire saturation ($L_7, L_8$). 4WS, however, accounts for both front and rear wheel saturation. This introduces two key triangular instability zones in the phase-plane:

• Upper-Right Triangle: Defined by $\beta > L_6 \cap \beta > L_3$
• Lower-Left Triangle: Defined by $\beta < L_5 \cap \beta < L_4$

By defining these regions as unstable due to front-wheel saturation, 4WS provides a much more complete and physically realistic safety net.

Moving Forward: Steering Within the Lines

Operating within the boundary set $S_c$ guarantees vehicle controllability. Crossing these boundary lines does not mean immediate mechanical failure, but it signals that the tires have saturated, and that an unrecoverable drift or high-speed spin is imminent.

By modeling these boundaries in real-time and feeding them to an active steering controller, a 4WS-equipped vehicle can dynamically adjust its rear steering angle to prevent the state from ever crossing $S_c$. This pre-emptive intervention represents a significant step forward for vehicle safety—controlling stability at the limits of physics before the ESP brakes even need to intervene.