As an engineer deeply involved in the development of next-generation electric vehicles, my research focuses on harnessing the unique capabilities of battery EV cars to achieve superior dynamic performance, especially under demanding conditions. The emergence of four-wheel independent drive systems, particularly those utilizing in-wheel motors, presents a revolutionary opportunity. Unlike conventional centralized drives, a distributed drive battery EV car offers direct and independent torque control at each wheel. This fundamental characteristic allows us to rethink vehicle stability management from the ground up, moving beyond the limitations of traditional electronic stability control (ESC) systems.
The core advantage of a wheel-hub motor driven battery EV car lies in its drastically shortened drive chain. The motor is integrated directly into the wheel, leading to a compact, efficient, and highly responsive architecture. This configuration provides an unprecedented level of actuation bandwidth and precision. Each wheel’s torque can be modulated independently and rapidly, enabling not just propulsion but also direct intervention in the vehicle’s yaw and lateral dynamics. For a battery EV car, this means the potential to implement stability control that works preemptively during normal driving, subtly correcting deviations without intrusive braking or noticeable driver disturbance, thereby enhancing both safety and driving feel.
However, this powerful capability also introduces a key challenge: functional overlap. In a traditional vehicle, longitudinal traction control and lateral yaw stability control are typically managed by a single system like ESC, which often uses differential braking as its primary actuator. In a distributed drive battery EV car, both differential driving torque (Torque Vectoring Control, TVC) and differential braking (e.g., via a Brake-by-Wire system) can generate the corrective yaw moment needed to stabilize the vehicle. Without coordination, these systems could conflict, leading to suboptimal performance or reduced efficiency. Therefore, the central problem becomes the design of an integrated collaborative control strategy that seamlessly unifies drive and brake control layers to maximize the stability envelope of the battery EV car under extreme operating conditions.
Architecture of the Integrated Control Strategy
To address this, a hierarchical control architecture is adopted. This structure clearly separates the high-level stability decision-making from the low-level optimal torque allocation, ensuring coherent operation between the drive and brake systems.
- Upper-Layer Controller (Stability Decision): This layer is responsible for monitoring the vehicle’s state, calculating the desired vehicle dynamics based on driver input and physical limits, and determining the total additional yaw moment (\( \Delta M_z \)) required to maintain stability. Its objective is to track a reference yaw rate while suppressing the vehicle sideslip angle.
- Lower-Layer Controller (Optimal Torque Allocation): This layer receives the desired yaw moment (\( \Delta M_z \)) from the upper layer. Its role is to optimally distribute this demand between the four in-wheel motors and the brake-by-wire system. The allocation prioritizes using drive torque from the motors to preserve kinetic energy and driving feel, only engaging friction brakes when the motor torque capacity is insufficient. It must also consider actuator limits, tire slip constraints, and overall efficiency.
The entire control framework for the distributed drive battery EV car can be summarized in the following table:
| Control Layer | Primary Objective | Key Inputs | Key Output | Core Algorithm |
|---|---|---|---|---|
| Upper Layer | Vehicle Stability (Yaw rate tracking, Sideslip suppression) | Driver steering angle, Vehicle speed, Measured yaw rate & sideslip angle | Total required corrective yaw moment (\( \Delta M_z \)) | Explicit Model Predictive Control (MPC) |
| Lower Layer | Optimal Actuation (Drive/Brake torque分配) | Desired \( \Delta M_z \), Driver torque demand, Wheel speeds, Motor/brake limits | Individual wheel motor torque (\( \Delta T_i \)) and brake pressure commands | Quadratic Programming-based MPC |
Upper-Layer Controller Design via Model Predictive Control
The upper controller employs an explicit Model Predictive Control approach to calculate the necessary stabilizing yaw moment. This requires a vehicle model for prediction.
Vehicle Reference Model
To define the desired vehicle behavior, a linear two-degree-of-freedom (2-DOF) bicycle model is used as a reference. This model describes the fundamental relationship between steering input and the vehicle’s lateral/yaw response in the linear region. The equations of motion are:
$$ m V_x (\dot{\beta} + \gamma) = F_{yf} + F_{yr} $$
$$ I_z \dot{\gamma} = l_f F_{yf} – l_r F_{yr} $$
Where \( m \) is vehicle mass, \( V_x \) is longitudinal velocity, \( \beta \) is the vehicle sideslip angle, \( \gamma \) is the yaw rate, \( I_z \) is the yaw inertia, \( l_f \) and \( l_r \) are distances from the center of gravity to the front and rear axles. \( F_{yf} \) and \( F_{yr} \) represent the total lateral tire forces on the front and rear axles, respectively. For the reference model, these forces are assumed to be linear functions of tire slip angles (\( \alpha_f, \alpha_r \)): \( F_{yf} = C_f \alpha_f \), \( F_{yr} = C_r \alpha_r \), where \( C_f, C_r \) are cornering stiffnesses. The slip angles are: \( \alpha_f = \beta + \frac{l_f \gamma}{V_x} – \delta \), \( \alpha_r = \beta – \frac{l_r \gamma}{V_x} \).
From this model, the steady-state desired yaw rate \( \gamma_{des} \) under a given steering angle \( \delta \) and speed \( V_x \) can be derived. However, this value must be limited by the available tire-road friction \( \mu \):
$$ \gamma_{max} = \frac{\mu g}{V_x} $$
$$ \gamma_{des,lim} = \min(|\gamma_{des}|, \gamma_{max}) \cdot \text{sign}(\gamma_{des}) $$
This \( \gamma_{des,lim} \) serves as the primary tracking target for the stability controller of the battery EV car.
Prediction Model and Control Formulation
For the MPC’s prediction model, a more detailed 2-DOF model is used, but the lateral tire forces are calculated using a non-linear Fiala tire model to better capture behavior near the friction limit. The Fiala model is expressed as:
$$ F_y = \begin{cases}
-C_\alpha \tan \alpha + \frac{C_\alpha^2}{3 \mu F_z} | \tan \alpha | \tan \alpha – \frac{C_\alpha^3}{27 \mu^2 F_z^2} \tan^3 \alpha, & |\alpha| < \arctan\left(\frac{3 \mu F_z}{C_\alpha}\right) \\
-\mu F_z \text{sign}(\alpha), & \text{otherwise}
\end{cases} $$
Where \( C_\alpha \) is the tire cornering stiffness (\(C_f\) or \(C_r\)), \( \alpha \) is the tire slip angle, \( \mu \) is the friction coefficient, and \( F_z \) is the tire vertical load. This model provides a more realistic force prediction as the battery EV car approaches its handling limits.
The state-space model for the MPC is formulated with the vehicle states and the control input:
$$ \mathbf{x} = [\beta, \gamma]^T, \quad u = \Delta M_z $$
The discrete-time state update equation is:
$$ \mathbf{x}(k+1) = \mathbf{A} \mathbf{x}(k) + \mathbf{B}_1 \delta(k) + \mathbf{B}_2 u(k) $$
Where matrices \(\mathbf{A}, \mathbf{B}_1, \mathbf{B}_2\) are derived from the discretized vehicle dynamics equations, incorporating the non-linear tire force calculations linearized at each operating point.
The MPC optimization problem at each time step is designed to minimize a cost function \(J\) over a prediction horizon:
$$ \min_{\Delta M_z} J = \sum_{i=1}^{N_p} \left\| \gamma(k+i) – \gamma_{des,lim}(k+i) \right\|^2_{Q_\gamma} + \sum_{i=1}^{N_p} \left\| \beta(k+i) \right\|^2_{Q_\beta} + \sum_{i=1}^{N_c} \left\| \Delta M_z(k+i-1) \right\|^2_{R} $$
Subject to: \( \Delta M_{z,min} \leq \Delta M_z \leq \Delta M_{z,max} \)
Where \( N_p \) is the prediction horizon, \( N_c \) is the control horizon, and \( Q_\gamma, Q_\beta, R \) are weighting matrices that balance yaw rate tracking accuracy, sideslip angle suppression, and control effort.
By leveraging the structure of this problem and applying Karush–Kuhn–Tucker (KKT) conditions, an explicit solution for the optimal corrective yaw moment \( \Delta M_z^* \) can be derived, avoiding online iterative optimization. This solution takes the form:
$$ \Delta M_z^* = f_1(\mathbf{x}) \cdot \gamma_{des} + f_2(\mathbf{x}) \cdot \beta + f_3(\mathbf{x}) $$
The functions \( f_1, f_2, f_3 \) depend on the current vehicle state and parameters. The final control output is saturated according to the instantaneous actuator limits: \( \Delta M_z = \text{sat}(\Delta M_z^*, \Delta M_{z,min}, \Delta M_{z,max}) \). This computed yaw moment is then passed to the lower-layer controller for execution.
Lower-Layer Controller: Coordinated Drive & Brake Torque Allocation
The lower layer’s task is to generate the upper layer’s demanded yaw moment (\( \Delta M_{z,req} \)) through optimal coordination of the four in-wheel motors and the brake-by-wire system in the battery EV car. The allocation philosophy is to prioritize the use of drive torque from the motors (TVC) because it is more efficient and does not decelerate the vehicle unintentionally. The hydraulic brake system (IBC) is engaged only when the TVC capacity is exceeded.
TVC Yaw Moment Capacity
First, the available capacity for generating yaw moment via drive torque is calculated. The maximum possible additional torque at each wheel (\( \Delta T_{i,max} \)) is the difference between the motor’s peak torque at the current speed and the driver’s demanded torque at that wheel. The corresponding maximum and minimum yaw moments achievable through TVC alone are:
$$ \Delta M_{z,TVC}^{max} = \frac{d}{2R} (\Delta T_{FL,max} – \Delta T_{FR,max} + \Delta T_{RL,max} – \Delta T_{RR,max}) $$
$$ \Delta M_{z,TVC}^{min} = \frac{d}{2R} (\Delta T_{FL,min} – \Delta T_{FR,min} + \Delta T_{RL,min} – \Delta T_{RR,min}) $$
Where \( d \) is the track width and \( R \) is the tire effective radius.
Drive/Brake Coordination Logic
The coordination logic between TVC and IBC is as follows:
$$ \text{If } |\Delta M_{z,req}| \leq |\Delta M_{z,TVC}^{limit}|: \quad \Delta M_{z,TVC} = \Delta M_{z,req}, \quad \Delta M_{z,IBC} = 0 $$
$$ \text{If } |\Delta M_{z,req}| > |\Delta M_{z,TVC}^{limit}|: \quad \Delta M_{z,TVC} = 0.8 \cdot \Delta M_{z,TVC}^{limit} \cdot \text{sign}(\Delta M_{z,req}), \quad \Delta M_{z,IBC} = \Delta M_{z,req} – \Delta M_{z,TVC} $$
Here, \( \Delta M_{z,TVC}^{limit} \) is the maximum yaw moment TVC can generate reliably (often set below the absolute physical maximum to retain a torque reserve). This logic ensures the drive system handles the bulk of the stabilization, with braking providing supplemental support in extreme situations, a key feature for managing the energy resources of a battery EV car.
Wheel Dynamics and MPC-Based Allocation
A detailed model including wheel rotational dynamics is used for precise torque allocation:
$$ I_w \dot{\omega}_i = T_{d,i} + \Delta T_i – T_{b,i} – F_{x,i} R $$
Where for each wheel \( i \), \( I_w \) is the rotational inertia, \( \omega_i \) is the angular speed, \( T_{d,i} \) is the driver’s torque demand, \( \Delta T_i \) is the additional control torque from the motor (the optimization variable), \( T_{b,i} \) is the brake torque, and \( F_{x,i} \) is the longitudinal tire force.
A second MPC controller is formulated in this lower layer. Its states are the four wheel speeds (\( \mathbf{x} = [\omega_1, \omega_2, \omega_3, \omega_4]^T \)), and its control inputs are the four motor torque adjustments and the brake torques for selected wheels (\( \mathbf{u} = [\Delta T_1, \Delta T_2, \Delta T_3, \Delta T_4, T_{b1}, T_{b2}]^T \)). The cost function for this MPC balances multiple objectives:
$$ \min_{\mathbf{u}} J_{low} = \sum \left\| \Delta M_{z,TVC}(k) – \Delta M_{z,TVC,cmd}(k) \right\|^2 + \sum \left\| \Delta M_{z,IBC}(k) – \Delta M_{z,IBC,cmd}(k) \right\|^2 + \sum \left\| \omega_i(k) – \omega_{i,des}(k) \right\|^2 + \sum \left\| \Delta T_i(k) \right\|^2 $$
The first two terms ensure tracking of the TVC and IBC yaw moment commands from the coordination logic. The third term helps maintain wheel speed stability and prevent excessive slip. The fourth term minimizes control effort. This optimization is subject to critical constraints:
- Motor Torque Limits: \( \Delta T_{i,min} \leq \Delta T_i \leq \Delta T_{i,max} \)
- Brake Torque Limits: \( 0 \leq T_{b,i} \leq T_{b,max} \)
- Tire Slip Ratio Constraints: \( s_{min} \leq s_i \leq s_{max} \) (implemented via slack variables to ensure solver feasibility).
This quadratic programming problem is solved at each control step to find the optimal set of motor and brake torques that fulfill the upper layer’s stability demand while respecting all physical limits of the battery EV car.
Simulation Analysis & Validation
The proposed integrated control strategy was rigorously tested through co-simulation using a high-fidelity vehicle dynamics model in Carsim and the controller model in MATLAB/Simulink. The test vehicle was modeled as a distributed drive battery EV car with independent in-wheel motors on all four wheels and a brake-by-wire system. Key tests were the Double Lane Change and Emergency Obstacle Avoidance maneuvers on low-friction surfaces, following standardized procedures.
Double Lane Change on Low-μ Surface
A simulation was conducted with an initial speed of 90 km/h on a snowy surface (friction coefficient μ ≈ 0.35). The results starkly demonstrate the necessity and effectiveness of the integrated control system for the battery EV car.
| Performance Metric | Controller OFF | Controller ON | Improvement |
|---|---|---|---|
| Max Yaw Rate Tracking Error (deg/s) | 37.0 | 0.8 | ~98% reduction |
| Peak Vehicle Sideslip Angle (deg) | 22.7 | 1.2 | ~95% reduction |
| Peak Tire Slip Ratio | 0.47 | 0.03 | ~94% reduction |
| Maneuver Completion | Failed (Spin-out) | Stable & Successful | Fundamental stability restored |
With the controller off, the vehicle quickly lost stability, exhibiting large yaw rate errors and a dangerous sideslip angle, leading to a spin. With the integrated controller active, the battery EV car maintained excellent composure. The yaw rate closely tracked the friction-limited desired value, the sideslip angle was kept near zero, and tire slip was minimized, allowing the vehicle to complete the maneuver safely. This shows the controller’s ability to optimally utilize the combined tire-road friction through coordinated torque allocation.
Emergency Obstacle Avoidance on Low-μ Surface
This more aggressive maneuver tests transient stability. The test velocity for a stable pass increased from 45 km/h with the controller off to 48 km/h with it on, indicating an expanded stability boundary. The quantitative improvements at 45 km/h are summarized below:
| Performance Metric | Controller OFF | Controller ON | Improvement |
|---|---|---|---|
| Average Steering Wheel Angle (deg) | 92.46 | 49.58 | ~46% reduction |
| Peak Vehicle Sideslip Angle (deg) | > 5 (Unstable) | 2.14 | Effectively suppressed |
| Peak Yaw Rate (deg/s) | 40.13 | 23.09 | ~42% reduction (closer to desired) |
| Steering Effort/Work | High | Substantially Lower | Driver burden significantly reduced |
The data shows that with the controller active, the driver needed significantly less steering input to navigate the obstacle, the vehicle’s yaw motion was more controlled, and the critical sideslip angle was kept within a safe, stable range. The battery EV car’s response became more predictable and less taxing for the driver, confirming that the integrated control system effectively assists in maintaining stability without being intrusive.
The actuator response plots further illustrate the coordination. During the maneuvers, the controller primarily used differential drive torque from the motors to generate the stabilizing yaw moment. The brake-by-wire system activated only during the most demanding phases when the yaw moment demand exceeded the pre-set TVC threshold, providing a seamless and powerful stabilizing effect. This synergy between the drive and brake systems is the hallmark of the proposed integrated approach.
Conclusion
This research presents a comprehensive hierarchical integrated control strategy for a four-wheel distributed drive battery EV car, designed to enhance handling stability under extreme conditions. By employing Model Predictive Control at both the vehicle stability layer and the torque allocation layer, the system achieves coherent, optimal coordination between the independent drive motors and the brake-by-wire system. The upper-layer MPC explicitly computes the required stabilizing yaw moment to track a friction-limited reference while minimizing sideslip. The lower-layer MPC optimally distributes this demand, prioritizing efficient drive torque and engaging friction brakes only as necessary, all while respecting actuator and tire slip constraints.
Simulation results for severe double lane change and emergency avoidance maneuvers on low-friction surfaces demonstrate the profound effectiveness of this approach. Compared to the uncontrolled vehicle, the integrated system drastically reduces yaw rate error, suppresses sideslip angles to safe levels, minimizes tire slip, and significantly reduces required driver steering effort. Most importantly, it prevents loss of control in situations where the conventional vehicle would fail. The strategy successfully expands the stable operating envelope of the battery EV car, showcasing the tangible benefits of deeply integrating drive and brake control. This work confirms that the inherent actuation advantages of distributed drive battery electric vehicles can be fully harnessed through sophisticated control algorithms to deliver unprecedented levels of active safety and dynamic performance.